title
stringlengths 1
827
⌀ | uuid
stringlengths 36
36
| pmc_id
stringlengths 5
8
| search_term
stringclasses 18
values | text
stringlengths 0
8.42M
|
|---|---|---|---|---|
Autonomous Establishment of CT‐Independent Sections to Obtain Optimum Pedicle‐Screw Axis in Direction, Length, and Safety Margin | 49570c22-53bd-43a0-9e11-d1fa8f04e823 | 11892341 | Surgical Procedures, Operative[mh] | Introduction Pedicle‐screw placement is one of the most common surgeries. The surgery is performed to restore spinal stability, relieve pain, and prevent further deterioration by preventing the rubbing of vertebral bones. Additionally, this surgery is performed to correct the deformities along the longitudinal axis of the spine . The pedicle is in close proximity to the spinal canal, which stores an integral portion of the central nervous system. Adjacent to the spine are numerous vital organs. Accuracy in accessing the desired path of the pedicle‐screw, both in direction and length, is of utmost importance in pedicle‐screw placement surgery as the margin‐for‐error is very small . The unguided free‐hand procedure of pedicle‐screw implantation is still a largely prevalent practice. Such practices have recorded a poor success rate, with incidences of pedicle‐screw misplacement ranging from 10% to as high as 40% . It is further recorded that 2%–8% of free hand surgeries leave the patient with critical injuries. There could be multiple reasons for inaccurate placement. One of the issues is presented in emphasises the importance of preoperative planning in aspects of pedicle diameter, pedicle angle, and screw length. The typical spinal pedicle‐screw placement is as shown in Figure . The transverse length (width) of the spinal pedicle in an axial view varies depending on vertebra orientation (see Figure ) with respect to the reference coordinates system of CT‐machine. The pedicle volume has an asymmetric structure, and its ‘major axis’ is not aligned with any of the reference axes of a CT‐machine. A surgical planner uses a set of two‐dimensional (2D) multiplanar reconstructed and/or a 3D rendering object to annotate a line. Manual annotation of the path is time‐consuming and also a largely unacceptable addition to the standard workflow. It is difficult to determine precise 3D lines that pass through the mid‐line of the asymmetric pedicle volumes . As a result, the manual annotation of the desired path is prone to human error and inconsistency due to intra‐ and interoperable variability . In recent years, there has been significant research and innovation to automate preoperative planning to estimate the pedicle axis and reduce the risk of screw misplacement. Machine‐learning‐based methods has been proposed to solve the pedicle segmentation problem and pre‐operative planning of the pedicle‐screw insertion path. The major drawback of the evolving method is that it requires a large amount of CT data, the right balance of age group and gender, and accurate curation of data to predict the pedicle contours. Also, ML does not consider the rotation of vertebrae, in the DICOM axial plane, as shown in Figure , while curating the training data sets. In this article, a novel method is presented to provide objective‐insight into the pedicle, irrespective of the vertebral orientation in the image space. Additionally, autonomous preoperative planning of pedicle‐screw insertion is devised. The objective is to define a pedicle‐medial axis which serves as optimum pedicle‐screw axis in direction, length and margin for spinal fusion surgery. The novelty of this work begins by establishing the pose of the ROI with respect to the local body coordinate frame for surgery in medical image space. A local coordinate system is uniquely defined for each vertebra based on the body features. Further, a coordinate system is attached to each pedicle of the vertebra to generate coordinate‐based true‐section of the pedicle called machine‐independent MPRs. The approach involves utilising image processing techniques and geometry‐based mathematical algorithms. The pose of the pedicle‐medial axis in relation to the vertebral coordinate system, {V} is established with digital accuracy. This eliminates subjective biases and human error. The remainder of the paper is organised such that Section provides an overview of data, the framework of the methodology for determination of pedicle‐medial axis and Section outlines the validation and assessment of results. The conclusion and discussion are made in Section , giving an account of the benefits of the new approach and the method.
Methods Image data preparation involves accusation of CT imaging data sets from individuals aged between 25 and 55 years (mean age 37.2 and median age 30). The data of axial sections were sequentially arranged, with an in‐plane voxel size vary between 0.385 and 0.791 mm and a slice thickness vary between 0.725 and 1.530 mm. Each image plane has 512 × 512 voxels and is acquired under a voltage of 120 kV with radiation ranging from 5 to 151 mAs. The slices that were stacked to form each vertebra in the lumbar region were segmented. Ethical approval for the use of anonymised data was obtained from the hospital. This article does not include any direct experiments involving animals or human participants conducted by the authors. During pre‐processing, the CT volumes were analysed, with each slice consisting of pixel intensities representing various structures of the human body. One such slice consisting of L2 vertebrae is illustrated in Figure . In this article, such an image is referred to as a raw CT image. The steps of pre‐processing the raw image are shown in Figure . The bounding of ROI, as shown in the Figure , is achieved by applying various filters and developing an algorithm. Detailed discussion on the method of bounding of ROI is kept out of the scope of this article. A CT image coordinate system, {I} is attached to the top‐left corner of the image, as shown in Figure , where +u‐axis represents the increasing columns and +v‐axis represents the increasing rows of the image pixels. In medical imaging, there is a standard practice of describing image planes in terms that relate to imaging device coordinate systems. In this article, the reference coordinate system of a CT‐machine is termed as the machine coordinate system, denoted by { C } , with C 0 representing the origin of the image space of the CT. The unit vectors X ^ C , Y ^ C a n d Z ^ C define the principal directions of { C } . When discussing the orientation of image planes, the XY plane is commonly understood as axial, the YZ plane as sagittal, and the XZ plane as coronal. However, it’s important to recognise that these descriptions—axial, sagittal, and coronal—originally refer with respect to body of the patient rather than specific orientations within the machine's coordinate system. For instance, what is clinically referred to as a mid‐sagittal corresponds to a YZ plane of the frame { C } , not the medial section of the body. This subtle difference in perception, if not accounted for, can lead to analytical errors. The challenge lies in reconciling these differing perspectives. While CT axial images may not explicitly account for the orientation of each vertebra, understanding and correcting for this discrepancy is somewhat subjective. It requires a careful consideration of both the anatomical context and the technical orientation of the images. In a pedicle‐screw placement surgery, a body feature‐based coordinate system becomes extremely beneficial for accurate perception. Kaushik et al. illustrated the importance of an ROI‐specific coordinate system in the context of robot assisted neurosurgery. They introduced a body affixed fiducial coordinate system to estimate the pose of the ROI image space . The aim is to make the robot take an optimum posture so as to keep the ROI in its best manipulation zone of the workspace. Therefore, a localised coordinate system for each vertebra and pedicle, based on its body feature will improve the visual perception by providing objectively accurate geometry, which is important for robotic assisted spine surgery and for manual surgery. The cross‐sectional images produced by employing a local body coordinate system, are invariant across different imaging platforms and are said to be Machine‐independent sections. A framework is formulated to establish pedicle‐medial axis, and create two coordinate systems, (a) vertebral coordinate system, denoted by {V}, and (b) pedicle coordinate system. It is designed in three stages: (i) inputting a stack of axial slices of a vertebra to obtain the Initial Pedicle Region (IPR), (ii) partitioned boundary of the pedicle region as an input and direction of the pedicle‐medial axis as an output, and (iii) generating machine‐independent multiPlanar reconstructions (MiMPRs) of the pedicle region while the pedicle‐medial axis is input. The first stage of our framework segments vertebral body boundary (VBB) and spinal canal boundary (SCB) using a delineating algorithm (Algorithm ). It begins with the Hough Transform , which identifies a closest circle approximating the centre of the lower vertebral body. A 2D working coordinate frame, labelled BO ( U ^ B , V ^ B ), is established at the centre of the circle. A distinct working coordinate frame, SC ( U ^ S , V ^ S ), is set near the centre of the spinal canal, with its origin, S C 0 , positioned by offsetting vector ‘ D ‾ ’ from B O 0 along the negative V ^ direction, as shown in the Figure . This detachment of body features from frame { C } enables the algorithm to pinpoint the extreme boundaries essential for the segmentation of the pedicle region in the local body specific frames. This is highlighted in Figure , showcasing the feature detection and precise separation. Algorithm 1 Marking of boundary 1 The orientation of each pedicle is not symmetric to the central line of the vertebra. Therefore, it needs to be segmented and isolated individually from the vertebra. The steps for isolation are explained below. An axis aligned rectangular box (see Figure ) passing through the extreme pixels E, and bounding the desired IPR, is placed. Further, a line segment is constructed joining centre of SCB and the centre of the incircle, as depicted in Figure . This line segment divides the region into the left and the right pedicles in accordance with the patient's left and right. The division is as shown in Figure using colour code. This procedure is applied to each of axial section of the vertebrae under consideration to obtain the bilateral partition. The coloured locations in Figure represent a ‘planar point cluster’ of the pedicle in XY plane of frame { C } . These cluster, when stacked together along the Z ^ c ‐axis, forms a 3D pedicle point cloud, as shown in the Figure . The module of the algorithm independently gives the left IPR and the right IPR, each shown in a distinguished colour code. The first stage of the framework is focused on the assessment and the measurement of the data in the images. It is based on the intensity and count of pixels in 2D image frame {I}. Transiting to the stage 2, the description and assessment will shift focus to the study of geometry. 2.1 Pedicle‐Medial Axis Estimation Henceforth, the discussion will be on stage 2 of the framework, where the development of a novel Decision‐Tree algorithm is elaborated for estimating the pedicle‐medial axis (P). The pedicle‐medial axis gives a true‐perspective of the pedicle geometry. Stage 2 also details the advanced planning strategy for lumbar and thoracic spinal fusion. To the best of our knowledge, no prior work has addressed the issue of identifying the optimal pedicle‐screw axis in both position on a critical section (pedicle isthmus) and orientation, based on reconstructed geometry. The research works suggested a path based on the safety margin and pull‐out strength, while suggested a method based on a coefficient assigned to different levels to obtain the path. However, the asymmetrical structure of the pedicle allows for more possibilities of unique solutions for an optimum path. Unlike the methods mentioned in , and , the Decision‐Tree algorithm integrates two distinct techniques and is generally patient‐specific. The first technique is termed as centroidal distance method (CDM). This technique is designed to mark the axis of symmetry of the pedicle represented in frame { C } as PC = P C x P C y P C z T where, P C x , P C y , P C z are direction cosines. The optimization is based on the following criteria (a) pivoting the axis through the area‐centroid of the directional minimal area enclosed by the cortical wall of the pedicle, (b) aligning the axis with the area‐centroid of neighbourhood sections to achieve optimal orientation, and (c) achieving the least degree of unevenness of margin along the axis. The CDM operates on (a) a stacked set of 2‐D point cluster of pedicle region (IPR), (b) seed‐axis, and (c) marked central region of IPR, as inputs. All inputs were obtained as described in the step 1 to step 3. The resulting vector is an optimal candidate for the pedicle‐medial axis (P), as described in step 4. The method for establishing the axis consists of the following sequential steps: establishment of a seed axis in pedicle region and regeneration of the planar point cluster. technique for determination of geometrical properties of the pedicle. a novel approach for defining the pivot point for the axis. provide a solution scheme to discover the pedicle‐medial axis. Step 1 Establishment of seed axis: The algorithm is initiated by establishing a seed axis. The seed axis is designed to pass through the IPR and is taken parallel to Y ^ C ‐axis of frame { C } . To analyse the geometric nature of the IPR volume, multiple layers are dissected that are parallel to each other and perpendicular to the axis. Figure illustrates the normal planes along the seed axis (a dashed line in green). The layers are discrete and equidistant by predetermined spacing. Understanding the geometry of the pedicle region is critical in finding the pedicle‐medial axis and accurately identifying its physical and geometrical features. One such dissection is shown in Figure . Step 2 Determination of geometrical properties of the pedicle: A planar surface or volume would be evaluated by connecting the simplices using a general formula (given by Equation ) (1) V n ( P ) = ∑ σ ∈ T 1 n ! det v 1 ( σ ) … … … … v n ( σ ) In this context, consider a layer (as described in step‐1) enclosing the boundary of the pedicle. The boundary is divided into smaller pieces called ( n − 1 ) simplices, where n is dimension of space. Now, for a given layer of IPR, boundary points of dissected layer lie in a plane and therefore, simplex will be a line segment, displayed in Figure . The important aspect is that these simplices are oriented to form a closed chain, which in this case, represents the wall of the pedicle. For n = 2, Equation can be modified to a shoe formula given by Equation . (2) 2 A i j Π j = x 1 x 2 y 1 y 2 + x 2 x 3 y 2 y 3 + x 3 x 4 y 3 y 4 + … . + x m x 1 y m y 1 here, A i j is enclosed area by boundary of j th dissected layer Π j and (X i , Y i ) are the vertices of the ( n − 1)‐simplex that is line segment. A further modification in the Equation leads to the development of a formula for the calculation of area‐centre, displayed in Figure . The area and its centroid are stored, for both, left and right pedicle, separately. The list is illustrated in Figure . Step 3 Defining anchor‐point for axis: Steps 1 and 2 describe the volume discretisation of IPR into layers and evaluation of the area enclosed by the pedicle boundary and its area‐centroid. An anchor point on an axis is uniquely defined as the directional minimal area (smallest cross‐sectional area normal to the given axis) within a ROI. It is crucial to do that because that minimal area has the least available margin. To achieve this, the area‐centroid, G i n (x, y, z) , is localised as given in Equation . (3) G i n ( x , y , z ) = arg min Y ∈ [ a , b ] A i j Π j where, — G i n symbolises the anchored‐point (temporary pivot) for subsequent axis, with left‐superscript indicating i th candidate of PC‐axis and subscript ‘n’ represents that unique plane which encloses directional minimal area. — A i j Π j represents the area, as function of temporary coordinate system (x, y, z), embedded in the j th dissected planes — Π j represent j th dissected plane which is normal to i th axis. The temporary coordinate system (x, y, z) can be replaced by specific frame as per requirement; for instance, when i = 0, machine frame {C} is selected and the location of the anchored‐point, G 0 n is marked,shown in Figure . This is the anchored‐point, which also acts as the origin of a successive frame that is to be used to establish next axis, i = 1. Similarly, for i = 2, a frame fixed at G 1 n is used and so forth. It is important to note that the level at which the transverse‐sectional area becomes minimum is not the same for both the pedicles. The Figure shows the variation of the left (in red) and right (in green) pedicle coronal areas as the axis descends into the pedicle. The result suggests the possibilities that: (a) the geometry of left and the right pedicle is not mirror image about Y ^ c ‐axis, or (b) the vertebra is rotated about axes of frame { C } , (c) combined effect of both (a) and (b). It is critical to identify and account for such differences during surgery. Step 4 Establishment of pedicle‐medial axis: The pedicle volume has an asymmetric structure, and the difference between the pedicle's anatomical alignment with respect to frame { C } introduces a challenge in identifying its medial axis. In order to overcome this difficulty and to provide better perspective, an axis of symmetry of pedicle volume is termed the PC‐axis, incorporating local symmetries inherent in the pedicle structure. To isolate the local symmetry and for a more precise analysis, the IPR is segmented into three distinct regions: (a) the top diverging region, (b) the central region, and (c) the bottom diverging region. The central region, being narrowest, exhibits close symmetry, whereas the other two regions on either side have asymmetric diverging regions. The depth of the central region of IPR about anchored points ranges as follows: T11 and T12 vertebrae: −1.8 mm to +2.4 mm, L1 and L2 vertebrae: −2.1 mm to +2.4 mm, and L3 vertebra: −1.2 mm to +1.8 mm showing vertebral anatomical variations. The solution step initiates with a seed axis aligning parallel to the Y ^ c ‐axis, which serves as an initial approximation for the PC‐axis. This setup occurs in the first iteration and serves as a basal step for subsequent refinements, shown as the converging axis ( P C i ) in Figure . A converging axis is anchored at the area‐centroid ( G i n ) of minimal area. A new volume of the central region is reconstructed, using discrete layers normal to the‐axis (see Figure ), maintaining proximity to the central region. The essence of an iterative solution lies in employing a covariance matrix. This mathematical approach binds the centroidal location within the vicinity of the minimal area while refining the axis by calculating the least perpendicular distances. Solving the covariance matrix yields three principal vectors, with the one exhibiting the highest variance designated as the axis of convergence. Further, the weighted root mean square (wrms) of the shortest Euclidian distance between the axis and the area‐centroid (axis‐centroid offset) is calculated. The centroid is the average point representation of boundary points and the axis with the minimum value is considered PC‐axis. The non‐uniform geometry of the pedicle provides multiple solution towards optimal axis with the same set of critical constraints. We find the other possible optimal solution maintaining critical constraints. The closeness of the solution not only validates the first method but also provides the optimal axis with different geometrical approaches. The second technique is the least area method (LAM), which establishes area‐optimum axis (PA‐axis). The variation of the optimal axis between the two techniques will reveal the difference due to geometrical criteria while serving the critical constraints. The PA‐axis is normal to plane enclosing the narrowest possible area, also termed as omnidirectional‐minimal area, enclosed by a pedicle cortical boundary. This method begins with the establishment of a seed axis, aligned parallel to the Y ^ C ‐axis, serving as the preliminary approximation of the PA‐axis. The IPR is discretised along the seed axis, as outlined in the step , and subsequently each plane is rotated within a range specifically tailored to the geometry of the pedicle under examination. A central to this methodology is the identification of the plane orientation that encloses the omnidirectional‐minimal area of the pedicle. Further, an axis normal to this minimal area is considered as PA‐axis. The corresponding flowchart detailing this algorithm is shown in Figure . It is important to note that the PA‐axis marks the critical insertion path along which the area of the pedicle is at its minimum, positioning the PA‐axis as a prominent and critical candidate for pedicle‐medial axis (P). The criticality of the PA‐axis lies in its precision; any linear deviation towards the medial wall may lead to catastrophic damage to the spinal cord. Also, any angular displacement about the anchored‐point increases the volume around the axis, thereby providing a greater pedicle isthmus area, but there should be check on the embedding optimum screw diameter. The decision‐tree algorithm integrates the two methods (CDM and LAM) to select the most suitable candidates between the PC‐axis and PA‐axis for the pedicle‐medial axis (P), based on the calculated values of (a) wrms of axis‐centroid offset, (b) wrms of unevenness of Buffer Margin (uBM) for each section (Π j ) along both axes. The flowchart is shown in Figure . 2.2 Optimization of Size of Insertion Path of Pedicle‐Screw In the context of pedicle‐screw placement, the diameter of the screw is calculated considering the designed safety margin (SM), defined at a plane containing the pedicle isthmus, which is the narrowest section of the pedicle. The SM is crucial as it accounts for allowable deviation in orientation and translation during screw insertion into the pedicle without any risk of failure. This concept is mathematically expressed as in Equation . (4) S M = D P n − D S where, D P n is diameter of pedicle isthmus and D s is diameter of Pedicle‐Screw. The SM value is pre‐decided by the surgeon as it depends on factors like bone mineral density and bone mass, and the direction of axis chosen. Therefore, the diameter of the‐screw is dependent on the morphological as well as physical properties of the pedicle, leading to Equation : (5) D S = D P n − S M To ascertain D P n , an approach known as the expanding circle method is formulated. This method starts with the formation of a circle that originates from a predetermined area‐centroid within the pedicle isthmus and expands outward until it tangentially contacts the pedicle walls. This process is visually demonstrated in Figure . Optimization proceeds in a stepwise manner, with each iteration recalculating the ashes of margin (AOM), as explicated in Equation . To do that, in the pedicle isthmus. An expanding circle initially make contact with only one side of the pedicle wall, the centre of circle is displaced in the radially opposite direction (see Figure ) by amount A O M 2 . This is an iterative process of expansion and translation of centre that is performed until the circle tangentially contacts the pedicle walls on both sides. This process converges if A O M ≤ 0.05 m m thus, ensuring an accurate measurement of the isthmus diameter. Moreover, this method facilitates the identification of a pivotal point by ensuring that the expansion of the circle continues symmetrically on both sides and refines the parameters for determining the medial axis (P). (6) A O M = | min D G P n k ( x , y , z ) , V n ( x , y , z ) − min D G P n k ( x , y , z ) , V n ′ ( x , y , z ) | P where, — G P n k symbolises the k t h anchored‐point(pivot), with left superscript P indicating medial path (axis) and subscript ‘n’ represents that unique plane Π n .which encloses enclosing pedicle isthmus A P min . — V n (x, y, z) belongs to coordinate of pedicle wall at a n th plane enclosing pedicle isthmus A P min that is normal to P‐axis. — V n ′ (x, y, z) belongs to coordinate of pedicle wall that are radially opposite to V n (x, y, z), at a n th plane enclosing pedicle isthmus A P min that is normal to P‐axis. — D is the distance operator that returns the distance between a 3‐D axis and a point. Once the pivotal point's location is optimised, the computation of optimum D P n is given in Equation (7) D P n 2 = min D I P ∩ Π n ( x , y , z ) , V n ( x , y , z ) where, — I P ∩ Π n belongs to intersection point of P‐axis and n th normal plane ( Π n ). After that, uBM is computed, as detailed in expression‐8. (8) u B M P j = min D I P ∩ Π j ( x , y , z ) , V j ′ ( x , y , z ) − D P j 2 where, D P j symbolise the maximum diameter of j th dissected plane Π j . In Figure , the j th dissected plane ( Π j ), represents a section where the condition BM > 0 indicates a positive margin surrounding the screw at its maximum available diameter. This surplus margin is crucial as it enhances pull‐out strength, reinforcing structural stability and ensuring secure anchoring. In figure, the uBM quantifies the degree of unevenness of the margin by measuring the absolute difference between the distances from the pedicle medial axis (P) to the medial cortex wall and the outer lateral cortex wall. The results of this analysis are discussed in Section of the article. Determination of the insertion depth of the‐screw depends on the path. As of now, there are majorly three ways for insertion of Pedicle‐screws, i.e. TT approach, CBT technique and MC approach . The medially directed traditional trajectory (TT approach) has gained popularity during the past 2 decades . If the surgeon follows the TT approach for screw insertion, the Decision Tree Algorithm (see Figure ), optimises the pedicle‐screw insertion depth based on quantitative measures. The measure of the insertion depth of a pedicle‐screw is well explained in .
Pedicle‐Medial Axis Estimation Henceforth, the discussion will be on stage 2 of the framework, where the development of a novel Decision‐Tree algorithm is elaborated for estimating the pedicle‐medial axis (P). The pedicle‐medial axis gives a true‐perspective of the pedicle geometry. Stage 2 also details the advanced planning strategy for lumbar and thoracic spinal fusion. To the best of our knowledge, no prior work has addressed the issue of identifying the optimal pedicle‐screw axis in both position on a critical section (pedicle isthmus) and orientation, based on reconstructed geometry. The research works suggested a path based on the safety margin and pull‐out strength, while suggested a method based on a coefficient assigned to different levels to obtain the path. However, the asymmetrical structure of the pedicle allows for more possibilities of unique solutions for an optimum path. Unlike the methods mentioned in , and , the Decision‐Tree algorithm integrates two distinct techniques and is generally patient‐specific. The first technique is termed as centroidal distance method (CDM). This technique is designed to mark the axis of symmetry of the pedicle represented in frame { C } as PC = P C x P C y P C z T where, P C x , P C y , P C z are direction cosines. The optimization is based on the following criteria (a) pivoting the axis through the area‐centroid of the directional minimal area enclosed by the cortical wall of the pedicle, (b) aligning the axis with the area‐centroid of neighbourhood sections to achieve optimal orientation, and (c) achieving the least degree of unevenness of margin along the axis. The CDM operates on (a) a stacked set of 2‐D point cluster of pedicle region (IPR), (b) seed‐axis, and (c) marked central region of IPR, as inputs. All inputs were obtained as described in the step 1 to step 3. The resulting vector is an optimal candidate for the pedicle‐medial axis (P), as described in step 4. The method for establishing the axis consists of the following sequential steps: establishment of a seed axis in pedicle region and regeneration of the planar point cluster. technique for determination of geometrical properties of the pedicle. a novel approach for defining the pivot point for the axis. provide a solution scheme to discover the pedicle‐medial axis. Step 1 Establishment of seed axis: The algorithm is initiated by establishing a seed axis. The seed axis is designed to pass through the IPR and is taken parallel to Y ^ C ‐axis of frame { C } . To analyse the geometric nature of the IPR volume, multiple layers are dissected that are parallel to each other and perpendicular to the axis. Figure illustrates the normal planes along the seed axis (a dashed line in green). The layers are discrete and equidistant by predetermined spacing. Understanding the geometry of the pedicle region is critical in finding the pedicle‐medial axis and accurately identifying its physical and geometrical features. One such dissection is shown in Figure . Step 2 Determination of geometrical properties of the pedicle: A planar surface or volume would be evaluated by connecting the simplices using a general formula (given by Equation ) (1) V n ( P ) = ∑ σ ∈ T 1 n ! det v 1 ( σ ) … … … … v n ( σ ) In this context, consider a layer (as described in step‐1) enclosing the boundary of the pedicle. The boundary is divided into smaller pieces called ( n − 1 ) simplices, where n is dimension of space. Now, for a given layer of IPR, boundary points of dissected layer lie in a plane and therefore, simplex will be a line segment, displayed in Figure . The important aspect is that these simplices are oriented to form a closed chain, which in this case, represents the wall of the pedicle. For n = 2, Equation can be modified to a shoe formula given by Equation . (2) 2 A i j Π j = x 1 x 2 y 1 y 2 + x 2 x 3 y 2 y 3 + x 3 x 4 y 3 y 4 + … . + x m x 1 y m y 1 here, A i j is enclosed area by boundary of j th dissected layer Π j and (X i , Y i ) are the vertices of the ( n − 1)‐simplex that is line segment. A further modification in the Equation leads to the development of a formula for the calculation of area‐centre, displayed in Figure . The area and its centroid are stored, for both, left and right pedicle, separately. The list is illustrated in Figure . Step 3 Defining anchor‐point for axis: Steps 1 and 2 describe the volume discretisation of IPR into layers and evaluation of the area enclosed by the pedicle boundary and its area‐centroid. An anchor point on an axis is uniquely defined as the directional minimal area (smallest cross‐sectional area normal to the given axis) within a ROI. It is crucial to do that because that minimal area has the least available margin. To achieve this, the area‐centroid, G i n (x, y, z) , is localised as given in Equation . (3) G i n ( x , y , z ) = arg min Y ∈ [ a , b ] A i j Π j where, — G i n symbolises the anchored‐point (temporary pivot) for subsequent axis, with left‐superscript indicating i th candidate of PC‐axis and subscript ‘n’ represents that unique plane which encloses directional minimal area. — A i j Π j represents the area, as function of temporary coordinate system (x, y, z), embedded in the j th dissected planes — Π j represent j th dissected plane which is normal to i th axis. The temporary coordinate system (x, y, z) can be replaced by specific frame as per requirement; for instance, when i = 0, machine frame {C} is selected and the location of the anchored‐point, G 0 n is marked,shown in Figure . This is the anchored‐point, which also acts as the origin of a successive frame that is to be used to establish next axis, i = 1. Similarly, for i = 2, a frame fixed at G 1 n is used and so forth. It is important to note that the level at which the transverse‐sectional area becomes minimum is not the same for both the pedicles. The Figure shows the variation of the left (in red) and right (in green) pedicle coronal areas as the axis descends into the pedicle. The result suggests the possibilities that: (a) the geometry of left and the right pedicle is not mirror image about Y ^ c ‐axis, or (b) the vertebra is rotated about axes of frame { C } , (c) combined effect of both (a) and (b). It is critical to identify and account for such differences during surgery. Step 4 Establishment of pedicle‐medial axis: The pedicle volume has an asymmetric structure, and the difference between the pedicle's anatomical alignment with respect to frame { C } introduces a challenge in identifying its medial axis. In order to overcome this difficulty and to provide better perspective, an axis of symmetry of pedicle volume is termed the PC‐axis, incorporating local symmetries inherent in the pedicle structure. To isolate the local symmetry and for a more precise analysis, the IPR is segmented into three distinct regions: (a) the top diverging region, (b) the central region, and (c) the bottom diverging region. The central region, being narrowest, exhibits close symmetry, whereas the other two regions on either side have asymmetric diverging regions. The depth of the central region of IPR about anchored points ranges as follows: T11 and T12 vertebrae: −1.8 mm to +2.4 mm, L1 and L2 vertebrae: −2.1 mm to +2.4 mm, and L3 vertebra: −1.2 mm to +1.8 mm showing vertebral anatomical variations. The solution step initiates with a seed axis aligning parallel to the Y ^ c ‐axis, which serves as an initial approximation for the PC‐axis. This setup occurs in the first iteration and serves as a basal step for subsequent refinements, shown as the converging axis ( P C i ) in Figure . A converging axis is anchored at the area‐centroid ( G i n ) of minimal area. A new volume of the central region is reconstructed, using discrete layers normal to the‐axis (see Figure ), maintaining proximity to the central region. The essence of an iterative solution lies in employing a covariance matrix. This mathematical approach binds the centroidal location within the vicinity of the minimal area while refining the axis by calculating the least perpendicular distances. Solving the covariance matrix yields three principal vectors, with the one exhibiting the highest variance designated as the axis of convergence. Further, the weighted root mean square (wrms) of the shortest Euclidian distance between the axis and the area‐centroid (axis‐centroid offset) is calculated. The centroid is the average point representation of boundary points and the axis with the minimum value is considered PC‐axis. The non‐uniform geometry of the pedicle provides multiple solution towards optimal axis with the same set of critical constraints. We find the other possible optimal solution maintaining critical constraints. The closeness of the solution not only validates the first method but also provides the optimal axis with different geometrical approaches. The second technique is the least area method (LAM), which establishes area‐optimum axis (PA‐axis). The variation of the optimal axis between the two techniques will reveal the difference due to geometrical criteria while serving the critical constraints. The PA‐axis is normal to plane enclosing the narrowest possible area, also termed as omnidirectional‐minimal area, enclosed by a pedicle cortical boundary. This method begins with the establishment of a seed axis, aligned parallel to the Y ^ C ‐axis, serving as the preliminary approximation of the PA‐axis. The IPR is discretised along the seed axis, as outlined in the step , and subsequently each plane is rotated within a range specifically tailored to the geometry of the pedicle under examination. A central to this methodology is the identification of the plane orientation that encloses the omnidirectional‐minimal area of the pedicle. Further, an axis normal to this minimal area is considered as PA‐axis. The corresponding flowchart detailing this algorithm is shown in Figure . It is important to note that the PA‐axis marks the critical insertion path along which the area of the pedicle is at its minimum, positioning the PA‐axis as a prominent and critical candidate for pedicle‐medial axis (P). The criticality of the PA‐axis lies in its precision; any linear deviation towards the medial wall may lead to catastrophic damage to the spinal cord. Also, any angular displacement about the anchored‐point increases the volume around the axis, thereby providing a greater pedicle isthmus area, but there should be check on the embedding optimum screw diameter. The decision‐tree algorithm integrates the two methods (CDM and LAM) to select the most suitable candidates between the PC‐axis and PA‐axis for the pedicle‐medial axis (P), based on the calculated values of (a) wrms of axis‐centroid offset, (b) wrms of unevenness of Buffer Margin (uBM) for each section (Π j ) along both axes. The flowchart is shown in Figure .
Optimization of Size of Insertion Path of Pedicle‐Screw In the context of pedicle‐screw placement, the diameter of the screw is calculated considering the designed safety margin (SM), defined at a plane containing the pedicle isthmus, which is the narrowest section of the pedicle. The SM is crucial as it accounts for allowable deviation in orientation and translation during screw insertion into the pedicle without any risk of failure. This concept is mathematically expressed as in Equation . (4) S M = D P n − D S where, D P n is diameter of pedicle isthmus and D s is diameter of Pedicle‐Screw. The SM value is pre‐decided by the surgeon as it depends on factors like bone mineral density and bone mass, and the direction of axis chosen. Therefore, the diameter of the‐screw is dependent on the morphological as well as physical properties of the pedicle, leading to Equation : (5) D S = D P n − S M To ascertain D P n , an approach known as the expanding circle method is formulated. This method starts with the formation of a circle that originates from a predetermined area‐centroid within the pedicle isthmus and expands outward until it tangentially contacts the pedicle walls. This process is visually demonstrated in Figure . Optimization proceeds in a stepwise manner, with each iteration recalculating the ashes of margin (AOM), as explicated in Equation . To do that, in the pedicle isthmus. An expanding circle initially make contact with only one side of the pedicle wall, the centre of circle is displaced in the radially opposite direction (see Figure ) by amount A O M 2 . This is an iterative process of expansion and translation of centre that is performed until the circle tangentially contacts the pedicle walls on both sides. This process converges if A O M ≤ 0.05 m m thus, ensuring an accurate measurement of the isthmus diameter. Moreover, this method facilitates the identification of a pivotal point by ensuring that the expansion of the circle continues symmetrically on both sides and refines the parameters for determining the medial axis (P). (6) A O M = | min D G P n k ( x , y , z ) , V n ( x , y , z ) − min D G P n k ( x , y , z ) , V n ′ ( x , y , z ) | P where, — G P n k symbolises the k t h anchored‐point(pivot), with left superscript P indicating medial path (axis) and subscript ‘n’ represents that unique plane Π n .which encloses enclosing pedicle isthmus A P min . — V n (x, y, z) belongs to coordinate of pedicle wall at a n th plane enclosing pedicle isthmus A P min that is normal to P‐axis. — V n ′ (x, y, z) belongs to coordinate of pedicle wall that are radially opposite to V n (x, y, z), at a n th plane enclosing pedicle isthmus A P min that is normal to P‐axis. — D is the distance operator that returns the distance between a 3‐D axis and a point. Once the pivotal point's location is optimised, the computation of optimum D P n is given in Equation (7) D P n 2 = min D I P ∩ Π n ( x , y , z ) , V n ( x , y , z ) where, — I P ∩ Π n belongs to intersection point of P‐axis and n th normal plane ( Π n ). After that, uBM is computed, as detailed in expression‐8. (8) u B M P j = min D I P ∩ Π j ( x , y , z ) , V j ′ ( x , y , z ) − D P j 2 where, D P j symbolise the maximum diameter of j th dissected plane Π j . In Figure , the j th dissected plane ( Π j ), represents a section where the condition BM > 0 indicates a positive margin surrounding the screw at its maximum available diameter. This surplus margin is crucial as it enhances pull‐out strength, reinforcing structural stability and ensuring secure anchoring. In figure, the uBM quantifies the degree of unevenness of the margin by measuring the absolute difference between the distances from the pedicle medial axis (P) to the medial cortex wall and the outer lateral cortex wall. The results of this analysis are discussed in Section of the article. Determination of the insertion depth of the‐screw depends on the path. As of now, there are majorly three ways for insertion of Pedicle‐screws, i.e. TT approach, CBT technique and MC approach . The medially directed traditional trajectory (TT approach) has gained popularity during the past 2 decades . If the surgeon follows the TT approach for screw insertion, the Decision Tree Algorithm (see Figure ), optimises the pedicle‐screw insertion depth based on quantitative measures. The measure of the insertion depth of a pedicle‐screw is well explained in .
Results The proposed method was implemented in PythonV3.7 and executed on an HP workstation operated under windows 10, 64‐bit operating system. The workstation is embedded with an Intel i7‐9800X processor at 3.80 GHz, 32 GB memory, and graphics processing unit (GPU) acceleration (Nvidia Quadro 4000 with CUDA). As mentioned above, the proposed method consisted of three stages, that is, (i) IPR segmenting, (ii) establishing the pedicle‐medial axis(P), and (iii) MiMPRs of pedicle region. Visualisation of MPRs is implemented using the open 3D library, while other major libraries used are scikit‐image, SciPy and NumPy. 3.1 Validation and Assessment The proposed framework to formulate CT‐independent sections and to obtain optimum pedicle‐screw axis in direction, length, and safety margin is evaluated by optimising three criteria: guidance based on the narrowest section of the pedicle isthmus, the non‐uniformity of the margin around the axis (referred as ‘unevenness of margin or uBM’), and aligning the axis with the area‐centroid of neighbourhood sections to achieve optimal orientation. The (a)‐series of figures from Figures , , , provide a holistic visualisation of the results for the left pedicle and the (b)‐series from Figures , , , show the results for the right pedicle through violin plots. The input data are generated from medical images of T11‐L3 vertebrae obtained from a CT scan of the patient in DICOM format. The slice thickness of 0.625 mm along Z ^ C , equal Pixel spacing of 0.404297 mm along X ^ C a n d Y ^ C . The plots offer a comparative analysis of the central pedicle region, as discussed in step of Section . Results and comparison of attributes such as the sectional area, the axis‐centroid offset, the degree of uBM and the available screw diameter in central pedicle region are demonstrated through violin plots. This analyses the results of both the CDM and LAM methods on both the left and right pedicles of vertebrae. The axis determined through CDM is referred to as the PC‐axis, while the axis determined through LAM is termed the PA‐axis. The violin plots Figure highlights significant observation about the sectional areas embedded in the dissected planes. The sectional areas are measured along the depth of pedicle region normal to PC‐axis and PA‐axis and for the left and right pedicle across various vertebral levels. The sectional areas of T11, T12, L1, L2 demonstrate low variance and a strong tendency to cluster within a narrow range, indicating effective identification of central regions. However, it differs for L3 due to the morphology. Therefore, in cases like L3, the analysis of area variations meets the basic objective of increasing diameter but does not conclusively guarantee symmetricity along the length of the pedicle‐medial axis (P). It lays a foundational basis for further evaluation of the offset between the axis and the area‐centroid with the understanding that (i) the axis is medial or close to medial to pedicle volume (ii) uBM should also be least for the axis. Figure provide a comparison of the offset for PC‐axis and PA‐axis along the depth of pedicle region. The graph shows how the spread of offset is distributed along PC‐axis and PA‐axis. Low variance (spread) signifies symmetric bone morphology along the depth of the axis. In general, the variance along the PC‐axis is significantly lower than along the PA‐axis ( except T11 vertebra). This is further supported by the wrms value of the offset, where the PC‐axis has a lower value than the PA‐axis (Table ), indicating that the PC‐axis has closer alignment with the area‐centroid. The ideal medial path should have low variance and low wrms value of the offset. Figure examine the degree of uBM for PC‐axis and PA‐axis, along the depth of pedicle region. Unlike offset measurement, uBM measurement is based on the local variation of boundary points within dissected plane. In general cases, the axis obtained is considered as pedicle‐medial axis (P) if the wrms values of offset and uBM are low. However in the cases exception to the general, that is, if the wrms value for offset are lower along the PC‐axis and uBM are lower along the PA‐axis, the decision‐tree algorithm selects the pedicle‐medial axis (P) based on lower unevenness of margin or uBM. The left pedicle of the L1 vertebra is the case in study exhibiting such a selection. Figure presents a graph that compares the trends of diameter and area along the PC‐axis for the left and right pedicles of the L2 vertebrae. The results highlight a critical narrowing region along the axis. Interestingly, the minimum diameter does not coincide with the minimum area (pedicle isthmus), as indicated by the markers in the legend. This suggests that the narrowest diameter does not necessarily correspond to the region with the smallest cross‐sectional area because the irregular geometry of the area may not embed the circular diameter proportionally. The study of the sections of area along the length of the axis is important because it has implications on structural assessments, such as load‐bearing capacity or vulnerability to stress along the depth of the pedicle. Figure shows the machine‐independent vertebral coordinate system and pedicle coordinate system along with the seed axis and pedicle‐medial axis (P). Figure illustrates machine‐independent coronal section that is defined in the space of unit axes ( X ^ L i , Z ^ L i ) while third axes, Y ^ L i , is cross product of first two axes and along the medial axis. Similarly, the true‐sagittal section is contained by ( Y ^ L i , Z ^ L i ). Krag et al. reported that increasing the depth of screw insertion can improve the strength of the transpedicular screw‐vertebra interface . Daemi et al. shows the result that using the conventional path planning method, pull‐out strength is a quadratic function of insertion depth . However, the recent introduction of Cortical Bone Trajectory (CBT)/Medial‐Lateral Screw Trajectory (MLST) method has shown to provide superior stability and grip compared to traditional trajectories, in case of osteoporotic vertebrae. Therefore, depending on physical properties such as BMD and morphological characteristics of pedicles such as diameter and geometry , and availability of technology such as screw augmented technology, clinicians decide the most optimum path for screw insertion. Regardless of the path utilised, the current Decision‐Tree algorithm accurately reconstructs the pedicle walls along the pedicle‐medial axis. The structure of machine independent and accurate reconstruction is demonstrated in Figure . In all the cases, the algorithm is successful in securing the optimal diameter of the screw satisfying the safety margin, uniformity of the margin and closeness to the centroid of the sectional area. The contribution of this paper is also in the ability to suggest the optimal screw insertion path based on local geometry. The Decision‐Tree Algorithm, as explained in Figure , opts for a longer axis length if the PC‐axis and PA‐axis are close. The MiMPRs facilitate the detection of any potential penetrations to the medial or lateral walls while planning manually.
Validation and Assessment The proposed framework to formulate CT‐independent sections and to obtain optimum pedicle‐screw axis in direction, length, and safety margin is evaluated by optimising three criteria: guidance based on the narrowest section of the pedicle isthmus, the non‐uniformity of the margin around the axis (referred as ‘unevenness of margin or uBM’), and aligning the axis with the area‐centroid of neighbourhood sections to achieve optimal orientation. The (a)‐series of figures from Figures , , , provide a holistic visualisation of the results for the left pedicle and the (b)‐series from Figures , , , show the results for the right pedicle through violin plots. The input data are generated from medical images of T11‐L3 vertebrae obtained from a CT scan of the patient in DICOM format. The slice thickness of 0.625 mm along Z ^ C , equal Pixel spacing of 0.404297 mm along X ^ C a n d Y ^ C . The plots offer a comparative analysis of the central pedicle region, as discussed in step of Section . Results and comparison of attributes such as the sectional area, the axis‐centroid offset, the degree of uBM and the available screw diameter in central pedicle region are demonstrated through violin plots. This analyses the results of both the CDM and LAM methods on both the left and right pedicles of vertebrae. The axis determined through CDM is referred to as the PC‐axis, while the axis determined through LAM is termed the PA‐axis. The violin plots Figure highlights significant observation about the sectional areas embedded in the dissected planes. The sectional areas are measured along the depth of pedicle region normal to PC‐axis and PA‐axis and for the left and right pedicle across various vertebral levels. The sectional areas of T11, T12, L1, L2 demonstrate low variance and a strong tendency to cluster within a narrow range, indicating effective identification of central regions. However, it differs for L3 due to the morphology. Therefore, in cases like L3, the analysis of area variations meets the basic objective of increasing diameter but does not conclusively guarantee symmetricity along the length of the pedicle‐medial axis (P). It lays a foundational basis for further evaluation of the offset between the axis and the area‐centroid with the understanding that (i) the axis is medial or close to medial to pedicle volume (ii) uBM should also be least for the axis. Figure provide a comparison of the offset for PC‐axis and PA‐axis along the depth of pedicle region. The graph shows how the spread of offset is distributed along PC‐axis and PA‐axis. Low variance (spread) signifies symmetric bone morphology along the depth of the axis. In general, the variance along the PC‐axis is significantly lower than along the PA‐axis ( except T11 vertebra). This is further supported by the wrms value of the offset, where the PC‐axis has a lower value than the PA‐axis (Table ), indicating that the PC‐axis has closer alignment with the area‐centroid. The ideal medial path should have low variance and low wrms value of the offset. Figure examine the degree of uBM for PC‐axis and PA‐axis, along the depth of pedicle region. Unlike offset measurement, uBM measurement is based on the local variation of boundary points within dissected plane. In general cases, the axis obtained is considered as pedicle‐medial axis (P) if the wrms values of offset and uBM are low. However in the cases exception to the general, that is, if the wrms value for offset are lower along the PC‐axis and uBM are lower along the PA‐axis, the decision‐tree algorithm selects the pedicle‐medial axis (P) based on lower unevenness of margin or uBM. The left pedicle of the L1 vertebra is the case in study exhibiting such a selection. Figure presents a graph that compares the trends of diameter and area along the PC‐axis for the left and right pedicles of the L2 vertebrae. The results highlight a critical narrowing region along the axis. Interestingly, the minimum diameter does not coincide with the minimum area (pedicle isthmus), as indicated by the markers in the legend. This suggests that the narrowest diameter does not necessarily correspond to the region with the smallest cross‐sectional area because the irregular geometry of the area may not embed the circular diameter proportionally. The study of the sections of area along the length of the axis is important because it has implications on structural assessments, such as load‐bearing capacity or vulnerability to stress along the depth of the pedicle. Figure shows the machine‐independent vertebral coordinate system and pedicle coordinate system along with the seed axis and pedicle‐medial axis (P). Figure illustrates machine‐independent coronal section that is defined in the space of unit axes ( X ^ L i , Z ^ L i ) while third axes, Y ^ L i , is cross product of first two axes and along the medial axis. Similarly, the true‐sagittal section is contained by ( Y ^ L i , Z ^ L i ). Krag et al. reported that increasing the depth of screw insertion can improve the strength of the transpedicular screw‐vertebra interface . Daemi et al. shows the result that using the conventional path planning method, pull‐out strength is a quadratic function of insertion depth . However, the recent introduction of Cortical Bone Trajectory (CBT)/Medial‐Lateral Screw Trajectory (MLST) method has shown to provide superior stability and grip compared to traditional trajectories, in case of osteoporotic vertebrae. Therefore, depending on physical properties such as BMD and morphological characteristics of pedicles such as diameter and geometry , and availability of technology such as screw augmented technology, clinicians decide the most optimum path for screw insertion. Regardless of the path utilised, the current Decision‐Tree algorithm accurately reconstructs the pedicle walls along the pedicle‐medial axis. The structure of machine independent and accurate reconstruction is demonstrated in Figure . In all the cases, the algorithm is successful in securing the optimal diameter of the screw satisfying the safety margin, uniformity of the margin and closeness to the centroid of the sectional area. The contribution of this paper is also in the ability to suggest the optimal screw insertion path based on local geometry. The Decision‐Tree Algorithm, as explained in Figure , opts for a longer axis length if the PC‐axis and PA‐axis are close. The MiMPRs facilitate the detection of any potential penetrations to the medial or lateral walls while planning manually.
Discussion The MiMPRs of a pedicle‐geometry are based on the orientation of the pedicle‐medial axis, which provides objective perception. The framework considers the uniformity of margin as primary and area‐centroid as secondary requisite in planning direction, length, and safety margin pedicle‐screw axis. The steps discovered are logic‐based, which rely on the geometry of the local region and make decisions without the need for manual data input. This mathematical and logical approach enhances the interpretability of the results, making it straightforward to understand the solution outcomes. The proposed frame also serves as an advanced surgical planner for pedicle‐screw placement that aligns with the midline of the pedicle volume, based on various parameters determined from multi‐planar reconstructions along the axis. The methods provide an average success rate of 100% among T11‐L3 vertebrae. The method provides patient‐specific and geometric‐specific results. No, general perception or knowledge were used to formulae the path. Considering that the error chain of the surgical system reaches up to 2 mm, the derived path for the pedicle‐screw can reduce the risk of screw misplacement or pedicle penetration, even when errors occur during registration or actual screw insertion. The stepwise analytical methodology of this study provides a thorough understanding of the effectiveness of the methods in accurately identifying the medial axis of the pedicle and supports the proposed selection criteria.
Amit Kumar: conceptualisation, data curation, methodology, visualisation, writing–original draft, writing–review & editing. Dwarakanath Thambihalli Aswath: conceptualisation, data curation, methodology, supervision, validation, writing–review & editing. Gaurav Bhutani: supervision, validation, writing–review & editing. Dwarakanath Srinivas: data curation, supervision, validation.
Ethical approval for the use of anonymised data was obtained from the hospital. This article does not include any direct experiments involving animals or human participants conducted by the authors.
Patient informed consent was not necessary for this investigation.
The authors declare no conflicts of interest.
|
Arthroscopic treatment for femoroacetabular impingement yields favourable patient‐reported outcomes and method survivorship at 10‐year follow‐up | 22513105-aaeb-4ba6-95a4-ca22e6bb2c19 | 11848980 | Surgical Procedures, Operative[mh] | Femoroacetabular impingement syndrome (FAIS) is a common cause of hip and groin pain among young and physically active individuals . In patients with cam morphology, the labral and cartilage injury caused by shearing forces can lead to osteoarthritis . Treatment of FAIS can be either nonsurgical with supervised personalized physiotherapy, patient education and activity modifications or surgical with the use of hip arthroscopy together with rehabilitation . Randomized controlled studies (RCTs) have reported superior outcomes of hip arthroscopy compared with physiotherapy; however, the majority have a short follow‐up period of 1 year . The arthroscopic treatment for FAIS is a well‐established treatment option . The aim of the surgery is to restore the hip joint anatomy by treating the bony irregularities (cam/pincer) and addressing any soft tissue damage that is associated with this condition . The long‐term outcomes of hip arthroscopic treatment for FAIS have not been extensively reported. Previous studies have evaluated predictive factors of hip arthroscopy results for FAIS, concluding that high grade of cartilage injury, older age and longer symptom duration are associated with worse outcomes . Several studies have reported good 2‐ and 5‐year outcomes of hip arthroscopic treatment for FAIS . However, there are only a few studies evaluating the long‐term outcomes of hip arthroscopy using modern validated patient‐reported outcome measures (PROMs) . The purpose of this study was to report the 10‐year patient‐reported hip function outcomes of hip arthroscopic treatment for FAIS using PROMs. The secondary aim was to investigate the survivorship of hip arthroscopy defined as nonconversion to total hip arthroplasty (THA) at a 10‐year follow‐up. The hypothesis was that at a 10‐year follow‐up, a statistically and clinically meaningful improvement in PROMs would be observed. Patients with FAIS who underwent hip arthroscopy between January 2011 and December 2013 were prospectively registered in a local hip arthroscopy registry. Surgeries were performed by three experienced surgeons at two hospitals in Gothenburg, Sweden. Patients were eligible for inclusion if they had a diagnosis of FAIS and had failed nonoperative treatment before proceeding to surgery. Exclusion criteria for PROMs analysis were previous hip surgery and withdrawn consent as well as conversion to THA. A total of 128 patients (151 hips) were included in the final PROMs analysis (Figure ). The demographic and perioperative information, including age, sex, uni‐ or bilateral surgery, operated side, surgical and traction time and Konan classification grade of acetabular chondral damage, was reported by the treating surgeon. Information regarding reoperations and conversion to THA was retrieved from patient journals. Descriptive statistics were used to present demographic data. The data were reported as mean, median, standard deviation (SD) and range for ordinal data. For nominal data, relative and absolute frequencies were used. Nonparametric statistical testing was used to compare paired means for continuous PROM data not normally distributed. To compare PROMs between preoperative and 2‐year follow‐up, the Wilcoxon signed rank test was used. The level of significance was set at p ˂ 0.05. Considering that 10 points of score change in International Hip Outcome Tool (iHOT‐12) are clinically relevant, an SD of 21 points (as Jonasson et al. has previously suggested) and an α ‐value of 0.05, the sample size calculation revealed that a power of >90% would be reached with 75 patients . The survivorship rate was calculated by dividing the number of hips not receiving THA within 10 years by the number of hips at risk of receiving THA. The Statistical Package for the Social Sciences (IBM SPSS statistics, version 28.0.1.1) was used to statistically analyse the patient data and PROMs. The patients completed the following PROMs preoperatively and at the 10‐year follow‐up: International Hip Outcome Tool short version (iHOT‐12), a shorter version of the iHOT‐33, which measures both health‐related quality of life and changes after treatment in young, active patients with hip disorders . Copenhagen Hip and Groin Outcome Score (HAGOS). The HAGOS questionnaire consists of six subscales: Symptoms, Pain, Function in daily living (ADL), Function in sport and recreation (Sport/Rec), Participation in Physical Activities (PA) and hip and/or groin‐related Quality of Life (QOL). The scores of each subscale are summed up and transformed to a 0–100 scale. A score of 0 represents severe hip and/or groin problems, while a score of 100 represents no hip and/or groin problems. With an interclass correlation coefficient (ICC) between 0.81 and 0.89 for the six subscales, the test–retest reliability of the Swedish version of HAGOS is found to be very good . European Quality of Life–5 Dimensions Questionnaire (EQ‐5D) and European Quality of Life–Visual Analogue Scale (EQ VAS), a standardized instrument evaluating health‐related quality of life . This PROM is a descriptive system that comprises five dimensions: mobility, self‐care, usual activities, pain/discomfort and anxiety/depression. Each dimension has five levels: no problems, slight problems, moderate problems, severe problems and extreme problems. Previous studies have shown that this instrument is valid and reliable . Hip Sports Activity Scale (HSAS), a PROM measuring the level of physical activity in patients with FAIS. The HSAS scale consists of nine different levels, from lowest being ‘No recreational or competitive sports’ to highest ‘Competitive sports (elite level)' . The Swedish version of HSAS has been found to be reliable (test‐retest reliability, ICC 0.930) and to have good content validity when compared to the Tegner score ( r = 0.794) . VAS scale for hip function. A question regarding satisfaction with surgery (yes/no). Patients exceeding the minimal important change (MIC) were reported, with the use of a distribution‐based technique, setting the cut‐off value at 0.5 times the SD of the score change . The number of patients achieving the acceptable symptomatic state (PASS) for the six HAGOS subscales and the iHOT‐12 was reported. The PASS is defined as the highest level of symptom beyond which patients consider themselves well . In previous studies, the PASS values are estimated and set for the iHOT‐12 63, HAGOS pain 68.8, HAGOS symptoms 62.5, HAGOS‐ADL 82.5, HAGOS‐PA 43.8, HAGOS‐Sport/Rec 60.9 and HAGOS‐QoL 42.5 . Informed consent was obtained from each patient in the study. Ethical approval for the study was granted by the Regional Ethical Review Board in Gothenburg (registration number EPN 2019‐06050). The hip arthroscopy procedures were performed in a supine position using one anterolateral and one midanterior portal. Axial traction was used to gain access to the central compartment. The peripheral compartment was approached through a capsulotomy made longitudinal to the capsule fibres and the iliofemoral ligament. This approach may potentially decrease the risk of iatrogenic‐induced hip laxity postoperatively. To assess the correct reshaping of the femoral head‐neck junction cam morphologies were resected under the guidance of intraoperative fluoroscopy. An ‘over‐the‐top’ technique was used for the correction of pincer morphology . Labral tears were treated with either debridement or suture, depending on the type and size of the tear. Depending on the location and morphology, the cartilage lesions were treated with either debridement or microfracture. No capsular closures were performed. Postoperatively full weight‐bearing adjusted in relation to pain was allowed immediately, but patients were recommended crutches for up to 4 weeks when walking outdoors. All patients were assigned to supervised rehabilitation. The patients were furthermore prescribed diclofenac 50 mg orally three times a day for three weeks postoperatively to prevent heterotopic ossification. There were 128 patients (151 hips) included in the final PROMs analysis. The follow‐up rate, defined as the percentage of eligible patients who were not lost to follow‐up was 48% (173/361). The method survivorship rate was 77%. Demographic data are presented in Table . Of the included 151 hips, 42 had isolated cam morphology (28%). Two hips had isolated pincer morphology (1%) and 79 hips had both cam and pincer morphology (52%). The labrum was sutured in 16 hips (Table ). Ten patients had exposed bare bone in the acetabulum corresponding to Konan type 4 (Table ). When comparing the preoperative PROMs results to those obtained at 10‐year follow‐up a statistically significant increase in all outcome measures except for HSAS was found. At the 10‐year follow‐up, 104 patients (83%) reported that they were satisfied with the surgery (Table ). The median preoperative iHOT‐12 score was 44.5 compared with 85.8 at 10‐year follow‐up. The calculated MIC value for iHOT‐12 was 10.0. Of all patients, 76% (97/128) exceeded MIC, while 69% (88/128) exceeded the PASS level of 63.0 postoperatively. A statistically significant improvement ( p < 0.001) of all HAGOS subscales was observed. For HAGOS subscales, the calculated MIC values were 8.5 for symptoms, 9.2 for pain, 6.9 for ADL, 11.8 for Sports/Rec, 15.0 for PA and 14.2 for QoL (Table ). No significant change in HSAS levels was observed at the 10‐year follow‐up, compared with baseline. The median HSAS levels preoperatively and at 10‐year follow‐up was 3.1 versus 3.4. The most important finding of this study was the statistically significant improvement of PROMs at 10‐year follow‐up after hip arthroscopic treatment for FAIS compared with preoperative values. Clinically meaningful results were achieved as demonstrated by the number of patients exceeding the MIC and PASS values for iHOT‐12 and HAGOS scores. The survivorship of the method was durable (77%) at the end of the follow‐up period. Previous studies have reported similar PROMs at medium‐ and long‐term follow‐up . In a study with 177 athletes treated with hip arthroscopy for FAIS, the included patients were followed for a mean period of 127 months. The patients demonstrated a significant improvement in PROMs at a minimum 10‐year follow‐up with an 85.7% survivorship of the method, defined as no conversion to THA . Another study, including 393 hips followed patients treated for FAIS for a mean period of 7.5 years . The hip joint preservation rate was 90.4% (SD ± 1.7%; 95% confidence interval [CI]: 87.1–93.7), with a durable improvement of PROMs at the final follow‐up. Interestingly the authors performed a subgroup analysis investigating the surgical results of three distinct groups treated with different techniques. Both the survivorship rate and PROMs were better for the group treated with modern arthroscopic techniques compared with open or semi‐open techniques. That is in concordance with the results of the present study, where all included patients were treated with arthroscopic techniques. Carton et al. reported on a group of 119 patients with FAIS treated with arthroscopic osteochondroplasty with a minimum of 10‐year follow‐up. Statistically significant improvements were seen in modified Harris Hip Score (mHHS), 36‐Item Short Form Health Survey (SF‐36) and Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) scores, with high satisfaction rate (90%) observed at 10 years after surgery. The hip arthroscopy survivorship rate for this group was 91.4%, and the satisfaction rate was 90%. Tönnis grade 1 was associated with a lower survivorship rate of 80%, and the authors could not find any association between the increased risk of THA conversion and the age and sex of the patients. However, the PROMs used in that study do not satisfactorily reflect the level of function and activity for the young and active population. In this study, all PROMs except for HSAS were improved at 10‐year follow‐up compared with baseline. Similar PROMs were reported by Kaldau et al. for a group of 84 patients with FAIS treated with hip arthroscopy with a median follow‐up time of 82.9 months. At the end of the follow‐up period, 18% of the patients had undergone THA, which is in concordance with the present study. In the present study, there was a statistically significant improvement of EQ‐5D and HAGOS values while HSAS was unchanged. This can possibly be explained by the fact that in the long term, many of these patients do not continue at the same level of physical activity they had before the FAIS symptoms presented because of their increased age. However, the HAGOS subscales for sport and physical activity exhibited a significant improvement compared with baseline (sport 41.7 vs. 75.1, physical activity 31.9 vs. 74.5), thus making it difficult to draw any substantial conclusions. The 2‐ and 5‐year follow‐ups for the same group of patients included in this study have previously been reported . The PROMs did not decline over time and the patient satisfaction rate at 10‐year follow‐up is at the same level compared with 5 years after surgery (83% vs. 84.6%). Survivorship at the 10‐year follow‐up had marginally declined compared to the previous follow‐up. The results in this study are comparable to other studies in which capsular repair and labral suturing are more commonly used . Although the study is not designed to evaluate the effect of these procedures, the present study shows that good results can be achieved with different techniques . The strengths of this study are the long follow‐up period, the large number of patients as well as the use of PROMs suitable for a young, active population. There are, however, several limitations to this study. There is no control group, thus not allowing a direct comparison to the PROMs of the treatment group included in this study. Hence, there is a risk of confounding factors affecting the reported PROMs. However, this is a case series study with prospective enrolment of patients where the whole group of patients treated has been eligible for inclusion. Another limitation of this study is the follow‐up rate of 48%, which could be attributed to the long follow‐up period and could affect the internal validity of the study. However, the relatively low response rate (RR) is not uncommon in registry studies . In a scoping review of RRs in clinical quality registries, the authors found that the average RR in Scandinavian registry studies at 10‐year follow‐up was about 51% (±10.7) . A previous study compared patient reported outcomes between responders and initial nonresponders in the hip arthroscopy registry also used for the purpose of this study; the authors concluded that there was no difference between the two groups except for the patient satisfaction rate . Hence, the results reported seem to be in concordance with previous studies. Moreover, there is no radiographic 10‐year follow‐up of the patients included in this study. That depends on the fact that patient data used for this study were extracted from a local registry that does not include a radiographic examination. However, it would be interesting to investigate if the patients who are less satisfied at the final follow‐up have more radiographic signs of cartilage injury than the patients who are satisfied. Patients undergoing arthroscopic treatment for FAIS reported statistically significant and clinically relevant improved outcomes at 10‐year follow‐up. All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by Sarantos Nikou and Joel Sturesson. The first draft of the manuscript was written by Sarantos Nikou and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript. The study was approved by the Swedish Ethical Review Authority. All the procedures being performed were part of the routine care. Informed consent was obtained from all individual participants included in the study. The authors declare no conflict of interest. |
Genetic context modulates aging and degeneration in the murine retina | 847fb129-ca5b-41e7-b24c-d35ccdd5515c | 11744848 | Biochemistry[mh] | Retinal degeneration is a leading cause of vision loss and is among the most common forms of neurodegenerative disease. These conditions, including age-related macular degeneration (AMD), diabetic retinopathy (DR), glaucoma, and retinitis pigmentosa (RP), pose a major economic burden , and loss of vision due to these conditions severely reduces quality of life . Aging is one of the most important risk factors for the development of retinal degenerative diseases . Thus, it is vital to understand the mechanisms of healthy and pathological retinal aging. Further, given the shared biological features of the retina and brain, mechanisms driving retinal degenerative diseases are likely relevant to neurodegenerative diseases of the brain, such as Alzheimer’s disease and related dementias (ADRD). There is growing interest in leveraging the accessibility of the retina to track the progression of neurodegeneration within the brain . Yet, age-related and pathological changes in retinal function may obscure this insight into brain pathology. The prevalence of aging-associated retinal impairments in people over 65 has been estimated to range from 5–20% of this increasingly prevalent population . These impairments can include decline of visual acuity and color perception . The molecular processes that underlie the development of these common impairments are not entirely clear. Pathological retinal aging, including the development of AMD and glaucoma, is associated with moderate to severe loss of central and/or peripheral vision. Foundational understanding of common and pathological retinal aging-associated changes will help foster the use of the eye as a biomarker for other CNS diseases, including ADRD. Visual impairment has been associated with worse outcomes on the commonly utilized Mini-Mental State Examinations cognitive screening tool, even on tasks not requiring sight—suggesting links between retinal dysfunction and cognitive decline . Only a few studies have investigated aging-associated transcriptomic changes of the human retina. One such study showed evidence of changes in cellular metabolism, cell cycle regulation, and cell adhesion in aged retinas compared to young controls . Mouse models have also aided in understanding retinal aging and degeneration . Previous retinal microarray analyses revealed aging-associated complement activation concomitant with changes in stress response pathways in humans and C57BL/6J (B6) mice . B6 rod photoreceptors exhibit age-associated changes to metabolic pathways—including lipid metabolism . These initial studies suggest human and mouse retinas exhibit similarities in aging responses. The development of higher throughput transcriptomics and proteomics has made it possible to profile many more transcripts and proteins in a single sample . This is critically important for retinal tissue, in which rods comprise ~ 70% of cells . In-depth profiling of the retina transcriptome and proteome may reveal changes that are smaller in magnitude or are driven by changes in less abundant cell types—identifying novel therapeutic targets to explore. There is a substantial genetic component influencing susceptibility to retinal degeneration. For instance, there are growing numbers of single nucleotide polymorphisms linked to retinal neurodegenerative diseases . In addition, genetic contributions to retinal degeneration are likely to be underestimated as, historically, the inclusion of genetically diverse individuals is lacking . One of the only studies investigating aging in the human retina was limited to Caucasians, and another study did not report ethnicities . While there is a similar gap in preclinical studies of retinal neurodegenerative diseases (the majority incorporate only one strain–commonly B6), a few studies show genetic background strongly influences retinal disease phenotypes. For example, in retinal degeneration, the rd7 phenotypes on a B6 background were suppressed by protective alleles present in CAST and AKR/J genetic backgrounds . Furthermore, a quantitative trait locus mapping study identified multiple loci that modified rd3 -associated retinal phenotypes . For glaucoma, mutations known to drive ocular hypertension and consequential retinal ganglion cell (RGC) loss in DBA/2J mice failed to cause glaucoma-relevant phenotypes in B6 mice . Thus, the inclusion of genetically diverse mouse strains will be critical in the study of retinal aging and degeneration. Mouse resource populations consisting of genetically diverse strains or individuals (e.g. Collaborative Cross (CC) and Diversity Outbred (DO) populations) have been leveraged to link genetic context to specific phenotypes and gene expression . CC and DO mice originate from crosses of eight founder strains, which encapsulate the majority of known genetic differences within laboratory mice . Aging has been linked to numerous cellular perturbations across many tissues . We have documented the importance of incorporating diverse mouse strains into studies of aging and neurodegenerative diseases of the brain , yet their use in understanding retinal aging has been limited. Therefore, we profiled molecular signatures of retinal aging across nine genetically diverse mouse strains: standard pigmented strains (B6, 129S1, NZO), standard albino strains (BALBc, NOD, AJ), and wild-derived strains (WSB, CAST, PWK). These strains include all founder strains of the CC and DO mouse panels with the addition of commonly utilized BALBc mice. Our initial analyses of these data unveiled a common aging signature among mice which was strongly modulated by genetic background. Analyses of cell type-specific marker genes predicted NZO and WSB strains may exhibit human-relevant retinal neurodegeneration with age. In vivo and ex vivo ophthalmologic assays indicated hallmarks of RP- and AMD-relevant photoreceptor degeneration in WSB retinas, while NZO mice displayed many features relevant to DR. Collectively, our work elucidates the mechanisms underlying retinal aging and degeneration and offers a publicly available database to accelerate research into mechanisms relevant to aging and degeneration of both the retina and brain. Mouse strains and cohort generation All research was approved by the Institutional Animal Care and Use Committee (IACUC) at The Jackson Laboratory. All mice were bred and housed in a 12-h light/dark cycle and received a standard 6% LabDiet Chow (Cat# 5K52) and water ad libitum. Cohorts of the following mouse strains were established and aged together for multi-omics analyses: C57BL/6J (B6; JAX Stock# 000664), 129S1/SvlmJ (129S1; JAX Stock# 002488), NZO/HlLtJ (NZO; JAX Stock# 002105), WSB/EiJ (WSB; JAX Stock# 001145), CAST/EiJ (CAST; JAX Stock# 000928), PWK/PhJ (PWK; JAX Stock# 003715), A/J (AJ; JAX Stock# 000646), NOD/ShiLtj (NOD; JAX Stock# 001976), BALB/cJ (BALBc; JAX Stock# 000651). Importantly, none of these strains carry any known rd mutation . Additional cohorts of B6, NZO, and WSB mice were generated for all other experiments. Tissue collection for RNA-sequencing and proteomics Mice were anesthetized with ketamine/xylazine and then underwent cardiac perfusion with PBS. Eye globes were gently excised, and retinas dissected. Retinas were quickly minced with a fresh razor blade on ice and tissue dispersed between two DNase/RNase free microcentrifuge tubes. Samples were flash frozen in liquid nitrogen and stored at −80 °C until further processing. RNA Isolation and RNA-sequencing Total RNA was isolated from retinal tissue aliquots using RNeasy Micro kits (Qiagen), according to the manufacturers’ protocols, including the optional DNase digest step. Tissues were homogenized in RLT buffer (Qiagen) using a Pellet Pestle Motor (Kimbal). RNA concentration and quality were assessed using the Nanodrop 2000 spectrophotometer (Thermo Scientific) and the RNA 6000 Nano and Pico Bioanalyzer Assay (Agilent Technologies). Libraries were constructed using the KAPA mRNA HyperPrep Kit (Roche Sequencing and Life Science), according to the manufacturer’s protocol. Libraries were assessed using D5000 ScreenTape (Agilent Technologies) and the Qubit dsDNA HS Assay (ThermoFisher) according to the manufacturers’ instructions. Approximately, 44 M 150 bp paired-end reads were sequenced per sample on an Illumina NovaSeq 6000 using the S4 Reagent Kit v1.5 (Illumina, Cat#20,028,313) by the Genome Technologies Core at The Jackson Laboratory. Raw FASTQ files were processed using standard quality control practices. High quality read pairs were aligned to a custom pseudogenome produced by incorporating strain-specific variants (single nucleotide variants and short insertions/deletions derived from Mouse Genomes Project REL-1505) into the mouse reference genome (GRCm38/mm10) using g2gtools version 0.2.7 ( https://github.com/churchill-lab/g2gtools ). Reads were aligned using STAR version 2.7.10a. Transcript counts were generated using RSEM version 1.3.3 . Differential expression analyses were completed using edgeR v3.40.2 within the R environment v4.2.3 using the glmQLFtest function . filterByExpr within edgeR was used to filter out very lowly expressed genes prior to TMM normalization . Differential expression was analyzed using generalized linear models via edgeR, with either 1) groups defined as a combination of strain, sex, age and a library batch term (~ 0 + Group + Batch) and contrasts designed to test for average strain effect, average age effect within each strain; or 2) age as a continuous variable to control and test for differential expression related to strain differences or determining the impact of aging while controlling for strain and sex differences. The clusterProfiler v4.6.2 R package was used to test for overrepresentation of gene ontology (GO) Biological Process gene sets within the list of differentially expressed genes (DEGs) with a false-discovery rate (FDR) of less than 0.05 and enrichPlot v1.18.4 and EnhancedVolcano v1.16.0 were used to visualize the enrichment results . Proteomics Sample Preparation Each half retina sample had ice-cold 600 µL of extraction buffer (2:2:1 methanol:acetonitrile:water) added, along with a pre-chilled 5 mm stainless steel bead (QIAGEN). Tubes were then added to the pre-chilled (at −20 °C) cassettes for the Tissue Lyser II in batches of 48 samples, then were lysed in the Tissue Lyser II for 2 min at 30 1/s. The stainless steel beads were removed with a magnet and the samples were placed at −20 °C for 16 h overnight. Sample extracts were then centrifuged at 21,000 × g at 4 °C for 15 min. The supernatant was removed, and the protein pellet was reconstituted in 100 µL of 50 mM HEPES buffer, pH 8.0. Reconstituted samples were vortexed at max speed for 30 s, a chilled steel bead was added, and they were reconstituted using the Tissue Lyser II as described for the extraction. Samples were then waterbath sonicated (sweep at 37 Hz at 100% power) for 5 min (30 s on, 30 s off for five cycles) with ice added to keep the temperature from rising. Samples were assessed for clarity and if necessary, the sonication was repeated. Samples were then centrifuged at 21,000 × g for 15 min at 4 °C and protein supernatant was transferred to a clean microcentrifuge tube. Protein quantification was then performed using a microBCA assay (Thermo, Cat# 23,235) on a single well for each sample with a 1:25 dilution. The remaining sample was snap-frozen and stored at −80 °C until further use. Protein digestion and peptide purification After protein quantification, 20 µg of each sample were aliquoted to a new tube and brought to an equal volume in 50 mM HEPES, pH 8.0. Samples with less than 20 µg total required use of the full sample for digest and the trypsin ratio was adjusted at the digestion stage. Samples were reduced with 10 mM dithiothreitol at 42 °C for 30 min on a ThermoMixer with 500 rpm agitation, alkylated with 15 mM iodoacetamide for 20 min at room temperature on a ThermoMixer with 500 rpm agitation, and trypsin-digested (Sequence Grade Modified; Promega) with a 1:50 trypsin:protein ratio at 37 °C for 20 h. Following the digest, all peptide samples were C18-purified using Millipore C18 zip-tips (Millipore, Cat#: ZTC18S096) according to the manufacturer protocol’s and as previously reported . Eluted purified peptides were dried with a vacuum centrifuge and stored at −20 °C until tandem mass tag (TMT) labeling was performed. TMTpro peptide labeling The dried peptide samples were reconstituted in 20 µL of 50 mM HEPES, pH 8.5) and quantified using the Quantitative Colorimetric Peptide Assay (Thermo, Cat# 23,275) according to the manufacturer’s protocol. Samples were diluted to a concentration of ~ 0.5 µg/µL in 10 µL of 50 mM HEPES, pH 8.5, and mixed on a ThermoMixer for 10 min at 25 °C (500 rpm). To create a quality control pool for a carrier channel, equal amounts of all samples were used to create enough pool that could be used in each multiplex for normalization. While the peptide solutions mixed, TMTpro reagents (Thermo, Cat# A44520) were mixed according to manufacturer protocol. TMTpro labels were added to a 19.37 mM final concentration to the appropriate peptide sample, followed by a 1 h incubation at 25 °C with 500 rpm agitation on a ThermoMixer. As the study required many multiplexes the samples were first randomized for the multiplex assignment and then were randomized for the TMT tag assignment within a multiplex through a list randomization in Random.org. Reactions were then quenched by adding a final concentration of 0.5% hydroxylamine and incubating for 15 min at 25 °C with 400 rpm agitation on a ThermoMixer. All samples were added to the appropriate TMT multiplex group, acidified by adding 20 µL of 10% formic acid, snap-frozen, and stored at −80 °C overnight. The next day samples were dried in a vacuum centrifuge. Liquid Chromatography Tandem Mass Spectrometry Analysis (LC–MS/MS) Each of the dried multiplexed peptide samples were reconstituted in 25 µL of LC–MS grade water with 0.1% TFA and zip-tipped using Millipore C18 zip-tips according to the manufacturer protocol’s and as previously reported . Purified multiplexes were then dried in a vacuum centrifuge and reconstituted in 20 µL of 98% H2O/2% ACN with 0.1% formic acid via pipetting and vortexing. Reconstituted samples were then transferred to a mass spec vial and placed in the autosampler at 4 °C. LC–MS/MS was then performed on a Thermo Eclipse Tribrid Orbitrap with a FAIMS coupled to an UltiMate 3000 nano-flow LC system in The Jackson Laboratory Mass Spectrometry and Protein Chemistry Service. The method duration was 180 min at a flow rate of 300 nL/min. Buffer A (100% H2O with 0.1% formic acid) and Buffer B (100% acetonitrile with 0.1% formic acid) were utilized for the gradient. The full gradient consisted of 98% A/2% B from 0–2 min, 98% A/2% B at 2 min to 92.5% A/7.5% B at 15 min, 92.5% A/7.5% B at 15 min to 70% A/30% B at 145 min, and 70% A/30% B at 145 min to 10% A/90% B at 155 min. The gradient was held at 10% A/90% B until 160 min and brought to 98% A/2% B by 162 min, where it was then held for 15 min to equilibrate the column. The TMT SP3 MS3 real-time search (RTS) method was used on the instrument. Global parameters included a default charge = 2, expected peak width = 30 s, advanced peak determination, spray voltage = 2000 V, mode = positive, FAIMS carrier gas = 4.6 L/min, and an ion transfer tube temperature = 325 °C. The instrument method utilized the FAIMS voltages of −40 V, −55 V, and −65 V. Settings for precursor spectra detection (MS1) in each node included: cycle time = 1 s (each node), detector = Orbitrap, Orbitrap resolution = 120,000, scan range = 400–1600 m/z, RF lens % = 30, normalize AGC target = 250, maximum inject time (ms) = auto, microscans = 1, data type = profile, polarity = positive, monoisotopic precursor selection = peptide, minimum intensity threshold = 5.0e3 (lower because of the FAIMS), charge states = 2–7, and a dynamic exclusion of a n = 1 for 60 s. Peptide fragment analysis (MS2) was performed in the ion trap and settings included: isolation window (m/z) = 0.7, collision energy = 35 (fixed), activation type = CID, CID activation time = 10 ms, quadrupole isolation mode, ion trap scan rate = turbo, maximum inject time = 35 ms, and data type = centroid. Prior to data-dependent MS3 (ddMS3) analysis, RTS was utilized with the UniProtKB Mus musculus (sp_tr_incl_isoforms TaxID = 10,090) protein database including cysteine carbamidomethylation (+ 57.0215 Da), TMTpro16plex (+ 304.20171 on Kn), and methionine oxidation (+ 15.9949). Additional parameters in the RTS included maximum missed cleavages = 2, Xcorr threshold of 1, dCn threshold of 0.1, and a precursor ppm of 10. SP3 MS3 was performed in the Orbitrap and settings included SPS precursors = 20, isolation window = 0.7 m/z, activation type = HCD, HCD energy normalized at 45%, resolution = 60,000, scan range = 100–500 m/z, normalized AGC target = 500%, maximum injection time = 118 ms, and centroid data collection. LC–MS/MS data analysis All of the Thermo Eclipse RAW data files from the Eclipse Tribrid Orbitrap mass spectrometer were searched against the UniProtKB Mus musculus (sp_tr_incl_isoforms TaxID = 10,090) protein database in Proteome Discoverer (Thermo Scientific, version 2.5.0.400) using Sequest HT according to standard manufacturer recommended workflows. The Sequest protein database search parameters included trypsin cleavage, precursor mass tolerance = 20 ppm, fragment mass tolerance = 0.5 Da, static cysteine carbamidomethylation (+ 57.021 Da), static TMTpro modification on any N-terminus (+ 304.207 Da), and a variable methionine oxidation (+ 15.995 Da). Other setting included a maximum number of missed cleavages = 2, minimum peptide length = 6, a maximum of 144 amino acids, and a fragment mass tolerance = 0.6 Da. Percolator was used and in this module in the software, the target/decoy selection was concatenated, q-value validation was utilized, a maximum delta Cn = 0.05 was set, and FDR < 0.05 for all matches was used as the threshold. Default Minora node parameters were used. Abundance values were then normalized to the total ion signal in the samples using the Reporter Ions Quantifier Node for TMT-based MS3 events. Differential proteomics analyses Identified peptides with FDR < 0.05 confidence and detected in at least three samples were used for further investigation. log 2 (scaled abundance + 1) values were used for differential expression analyses. Differentially expressed proteins (DEPs) were determined using generalized linear models and a repeated measures model via limma and DEqMS , which takes into account the number of unique peptide spectrum matches (PSMs) used for each peptide to enhance the statistical power of differential expression, with either 1) groups defined as a combination of strain, sex, age and a TMT multiplex batch term (~ 0 + Group + Multiplex) and contrasts designed to test for strain effects, age effects within each strain; or 2) age as a continuous variable to control and test for differential expression related to strain differences or determining the impact of aging while controlling for strain and sex differences. As samples were run in triplicate, we utilized the duplicateCorrelation function within limma to block on “mouseID” . As for the transcriptomic analyses, clusterProfiler v4.6.2 R package was used to test for overrepresentation of Gene Ontology (GO) Biological Process gene sets within the list of differentially expressed proteins with a FDR of less than 0.05 and enrichPlot v1.18.4 and EnhancedVolcano v1.16.0 were used to visualize the enrichment results . STRING-dB v 2.14.3 R package was used to visualize protein–protein interaction networks with the settings “score_threshold = 200” and “network_type = physical” and to calculate enrichment of GO gene sets within the list of DEPs identified between 4 M NZO or WSB with the other 5 pigmented strains . In vivo ophthalmological investigations All in vivo investigations were performed in B6, NZO, and WSB mice at 4, 8, 12, and/or 18 M. In vivo techniques included: slit lamp imaging, fundus exams, fluorescein angiography, and electroretinography (full-field and pattern). Slit lamp imaging Mice were anesthetized with ketamine/xylazine. Animals were placed on the apparatus with eyes aligned to the camera. Bright light images were obtained using Topcon DC-4 digital slit lamp and EZ Capture image software. Electroretinography (ERG) Mice were placed in an anesthesia induction chamber infused with 3–4% isoflurane. Once fully anesthetized, mice were transferred to the heated platform of the Celeris-Diagnosys ERG system (Diagnosys LLC, MA, USA) to maintain body temperature at approximately 37 °C, where anesthesia was sustained using 1–2% isoflurane. 1% Tropicamide and 2.5% phenylephrine were topically applied to dilate the pupils of both eyes, while a thin layer of 2.5% hypromellose was used to keep the eyes moist. Two corneal electrodes with integrated stimulators were placed on the surfaces of lubricated corneas to facilitate full-field ERG recordings. The Scotopic ERG was conducted initially, with nine steps of stimulus intensity (0.001, 0.002, 0.01, 0.0316, 0.316, 1,3, 10, and 31.6 cd.s/m 2 ). The final ERG waveform was the average of 10 individual waveforms. Following a 10 min light adaptation interval, the photopic ERG was evoked by five steps of stimulus intensity (0.316, 1, 3, 10, 31.6 cd.s/m 2 ). The final ERG waveform was an average of 20 individual waveforms. The amplitude of the a-wave was measured as the difference between the pre-stimulus baseline and the trough of the a-wave, and the b-wave amplitude was measured from the trough of the a-wave to the peak of the b-wave. OP amplitudes were measured from the pre-stimulus baseline to the highest OP peak. The highest amplitude values across strains were elicited by 10 cd.s/m 2 luminance, thus, amplitudes generated at 10 cd.s/m 2 were compared. Pattern electroretinography (PERG) All PERGs were performed using a JORVEC PERG system (Intelligent Hearing Systems, Miami, Florida) as described by the manufacturer. Briefly, mice were anesthetized with ketamine/xylazine and slit lamp was used to inspect each mouse eye prior to starting experiment. Mice were kept on a warming stage throughout the experiment. Electrodes were placed subcutaneously such that the active electrode is placed between the eyes with the tip just before the snout and the reference electrode is placed in line with the active electrode just between the ears. The ground electrode was placed approximately 2 cm in front of the base of the tail. Scans were collected and waveforms were averaged using default settings. PERG amplitude was calculated as P1-N2 for each eye per mouse. Optical Coherence Tomography (OCT), fundus, and fluorescein angiography OCT, fundus exams, and fluorescein angiographies were performed using Micron IV and Reveal OCT and Discover 2.4 software (Phoenix-Micron, Bend, OR). Mice were given 1 drop (20-30µl) of 0.5% Tropicamide (Somerset Therapeutics, NDC# 70,069–121-01) in both eyes. After 10 min, one drop (20-30µl) of 2.5% phenylephrine (Bausch & Lomb, NDC# 42,702–0103-05) was applied to both eyes. Mice were then anesthetized using 4% isoflurane until a proper plane of anesthesia was achieved. Mice were then transferred to a nose cone on a heated holding cradle. Anesthesia was then reduced to 2% isoflurane. Eyes were kept moist using GenTeal Tears Lubricant Eye Gel Drops (Alcon). Respiration rate was monitored, and the camera lens was adjusted to be perpendicular to the cornea. For OCT imaging, focus and brightness of the image guided OCT was adjusted until optimal image is previewed. OCT images were captured from the temporal to the nasal retina in a plane that included the optic nerve. To improve resolution, 10–40 images were taken and averaged. For fundus exams, a white light fundus image was taken of each eye. Brightness adjustments were made, and the focal plane was optimally adjusted. Immediately following the image-guided OCT and fundus imaging, mice were injected intraperitoneally with 1% Fluorescite® Fluorescein Sodium at a dose of 10 mg/kg (10 mg/mL solution, Akorn). Using a GFP filter, images of the right eye were taken every 30 s for 6 min. Images were taken of the left eye 30 s after the right eye imaging period. Retinal layers from OCT images were measured using the FIJI distribution of ImageJ . Each layer was measured 200, 400, and 600 µm from the edge of the optic nerve using the line measurement tool, and the three respective measurements were averaged and used for statistical analyses. Fluorescein isothiocyanate–dextran (FITC-dextran) leakage assay Mice were anesthetized with tribromoethanol and then transcardially perfused with 4% paraformaldehyde (PFA) (Electron Microscopy Services, Cat#15,714) and 3 mg/mL 70 kDa fluorescein isothiocyanate–dextran (Millipore-Sigma, Cat#FD70S) in PBS. Eye globes were excised, and retinas were immediately dissected and flat mounted. Flat mounts were imaged with a Leica DMi8 inverted microscope within 30 min of harvest using a GFP filter by taking 10 µm step z-stacked imaged to encompass the entire tissue at 10 × magnification. The presence of leaks and avascular areas were recorded. Tissue integrity was assessed by bright field microscopy. Histological analyses At the time of sacrifice, animals were anesthetized with tribromoethanol and were transcardially perfused with PBS. Eyes were enucleated and fixed in 37.5% methanol and 12.5% acetic acid in PBS for 18 h at 4 °C. The Histology Core at The Jackson Laboratory performed paraffin embedding, sectioning, and hematoxylin and eosin and Prussian blue staining of optic nerve head-containing sections. Slides containing 2–3 sections per eye were imaged using a Hamamatsu S210 NanoZoomer Digital Slide Scanner at 40X magnification by the Microscopy Core at The Jackson Laboratory. Images were processed using custom FIJI scripts. Briefly, 400 µm x 400 µm regions of interest (ROIs) were selected on both sides of the optic nerve head at approximate distances of 500, 1000 and 1500 µm representing the central, middle, and peripheral retina. After which, additional ROIs of the ONL per image were generated for automated thresholding, watershed segmentation, and quantification. The default thresholding was utilized with a size threshold of 30 µm 2 for automated counting for all ONL cells. To determine the length of the retina, we first generated a mask of the retina by applying 1) gaussian blur, 2) color thresholding, 3) filling holes in the mask, and finally calculated the length of the mask. Vascular network isolation, staining, and analysis At the time of sacrifice, animals were anesthetized with tribromoethanol and were transcardially perfused with PBS. Eyes were enucleated and fixed in 4% PFA in PBS for 18 h at 4 °C. The retinal vascular network was isolated as previously described with a few modifications. Briefly, retinas were gently dissected and washed 3 × with ddH20 for 5 min and then left overnight in ddH20 at RT. The following day, retinas were incubated in 2.5% Trypsin (ThermoFisher Scientific, Cat#15,090,046) for 2 h at 37 °C. After this, the retinas were washed 5 × with ddH20. Next, the vessels were gently washed from the neural retina with additional ddH20. Retinal vascular networks were then mounted onto glass slides and allowed to dry before staining. The vascular networks were stained with hematoxylin and eosin according to the manufacturer’s instructions (Abcam, Cat#ab245880). Stained networks were dehydrated with 3 washes with reagent alcohol and then mounted using organic mounting medium (Organo/Limonene Mount, Sigma-Aldrich, Cat#O8015). Dried slides were imaged using a Hamamatsu S210 NanoZoomer Digital Slide Scanner at 40X magnification by the Microscopy Core at The Jackson Laboratory. Acellular capillaries were manually counted across 4–7 400–500 µm × 400–500 µm ROIs of each retina, and the average count was used for statistical analyses. Immunohistochemistry (IHC) analyses Whole mount IHC At the time of sacrifice, animals were anesthetized with tribromoethanol and were transcardially perfused with PBS. Eyes were enucleated and fixed in 4% PFA in PBS for 18h at 4°C. Retinas were gently dissected from the eye cup and washed 3 times in PBS for 5 min, then washed 3 times in 0.3% Triton-X 100 in PBS for 5 min. Retinas were subsequently blocked in 10% Donkey Serum (Sigma-Aldrich, Cat#D9663) in 0.3% TritonX PBS for 24h at 4°C. Retinas were incubated in primary antibodies diluted in blocking buffer for 72h at 4°C and then washed 3× with PBS for 5 min. Retinas were then incubated with secondary antibodies in PBS for 24 hours at 4°C. Finally, retinas were washed 4 times with PBS for 5 min prior to whole mounting onto slides with fluorescent mounting medium (Polysciences, Cat#18606-20). Slides were allowed to dry prior to imaging. Primary antibodies utilized include: 1µg/mL Rabbit anti-RNA Binding Protein, MRNA Processing Factor (RBPMS) (GeneTex Cat#GTX118619). 0.5 μg/mL Donkey anti-Rabbit Alexa Fluor 568 (Invitrogen, Cat#A10042) was used as a secondary antibody. Multicolor wide-field tiled images were taken on a Leica DMi8 microscope a20× magnification across the entire retina. Retina images were stitched together in FIJI. All microscope settings were kept identical for each experiment. RBPMS + RGC counts were determined by generating 4 separate 500µm x 500µm ROIs from the peripheral and central retina for each mouse. These ROIs were then subjected to automated thresholding using the moments algorithm , and watershed segmentation before counting all RBPMS + objects above 100µm 2 . All analyses were performed with the FIJI distribution of ImageJ (NIH) . Retinal cross-sections IHC Paraffin embedded sections were de-paraffinized using 3 × 5 min washes with Clear-Rite 3 (Thermofisher, Cat#6901TS) and then rehydrated with a series of 5 min Ethanol washes (2 × 100%, 1 × 90%, 1 × 70%) and then washed 3 × with ddH20. Antigen retrieval was performed by steaming in Epitope Retrieval Solution (IHC World, Cat#IW-1100) for 1 h. Sections were permeabilized for 10 min with 0.3% TritionX100 in PBS then washed 3 × PBS. Sections were subsequently blocked in 10% Donkey Serum (Sigma-Aldrich, Cat#D9663) in PBS for 24 h at 4 °C. Sections were incubated in primary antibodies diluted in blocking buffer for 72 h at 4 °C and then washed 3 × with PBS for 5 min. Sections were then incubated with secondary antibodies in PBS for 2 h at room temperature. Finally, sections were incubated with DAPI for 5 min and then washed 4 × PBS for 5 min prior to cover slipping with fluorescent mounting medium (Polysciences, Cat#18,606–20). Slides were allowed to dry prior to imaging. Primary antibodies utilized include: 5 µg/mL Rabbit anti-Red/Green Opsin (Sigma-Aldrich, Cat#AB5405), 2 µg/mL Mouse anti-Rhodopsin [clone:1D4](Abcam,Cat#AB5417). Secondary antibodies utilized include: 0.5 µg/mL Donkey anti-Rabbit Alexa Fluor 647 (Invitrogen, Cat#A32795), and 0.5 µg/mL Donkey anti-Mouse Alexa Fluor 647 (Abcam, Cat#AB150107). Multicolor wide-field tiled images were taken on a Leica DMi8 microscope at 40 × magnification across the entire cross-section. All microscope settings were kept identical for each experiment. Images were stitched together in FIJI. Images were processed using custom FIJI scripts. Briefly, 400 µm x 400 µm regions of interest (ROIs) were selected on both sides of the optic nerve head at approximate distances of 500, 1000 and 1500 µm representing the central, middle, and peripheral retina. After which, additional ROIs of the ONL per image were generated for background removal (20 rolling ball radius), automated thresholding, segmentation, and quantification. For cones: the automated thresholding method “Triangle” was applied to each image, and objects with areas between 2–200 µm 2 were counted using FIJI. The resulting cone counts per image were normalized to the length of the IS/OS imaged and the percent change in retinal length over age for each strain as determined from histology. For the IS/OS analysis, ROIs containing the IS/OS were manually outlined, skeletonized, and the resulting skeleton was analyzed for length and area to calculate width. For rhodopsin staining: an ROI containing the IS/OS was used to evaluate rhodopsin signal by thresholding using the “Otsu” method and the resulting mask was used to calculate the area of positive rhodopsin. All analyses were performed with the FIJI distribution of ImageJ (NIH) . Statistics Analyses were performed using GraphPad Prism 10 software. Comparisons of percent of instances were analyzed using a Chi-square test. Data from experiments designed to test differences between two groups (e.g., one measurement across two ages within strain) were subjected to a Shapiro–Wilk test to test normality and an F test to compare variance. For normally distributed data with equal variance, a two-tailed independent samples t test was utilized. For normally distributed data with unequal variance, a Welch’s t test was used. For non-normally distributed data, a Mann–Whitney test was used. Data from experiments designed to test differences among more than two groups across one condition (e.g. one measurement across strains at one timepoint) were subjected to a Shapiro–Wilk test to test normality and a Brown-Forsythe test to compare variance. Normally distributed data with equal variance were analyzed using a one-way ANOVA followed by Tukey’s post-hoc test. Data from experiments designed to detect differences among multiple groups and across two conditions (e.g. measurements across ages and sexes within strains) were analyzed using a two-way ANOVA followed by Holm-Sidak’s post-hoc test. P values < 0.05 were considered statistically significant. Throughout the manuscript, results are reported as mean ± standard error of the mean (SEM). All research was approved by the Institutional Animal Care and Use Committee (IACUC) at The Jackson Laboratory. All mice were bred and housed in a 12-h light/dark cycle and received a standard 6% LabDiet Chow (Cat# 5K52) and water ad libitum. Cohorts of the following mouse strains were established and aged together for multi-omics analyses: C57BL/6J (B6; JAX Stock# 000664), 129S1/SvlmJ (129S1; JAX Stock# 002488), NZO/HlLtJ (NZO; JAX Stock# 002105), WSB/EiJ (WSB; JAX Stock# 001145), CAST/EiJ (CAST; JAX Stock# 000928), PWK/PhJ (PWK; JAX Stock# 003715), A/J (AJ; JAX Stock# 000646), NOD/ShiLtj (NOD; JAX Stock# 001976), BALB/cJ (BALBc; JAX Stock# 000651). Importantly, none of these strains carry any known rd mutation . Additional cohorts of B6, NZO, and WSB mice were generated for all other experiments. Mice were anesthetized with ketamine/xylazine and then underwent cardiac perfusion with PBS. Eye globes were gently excised, and retinas dissected. Retinas were quickly minced with a fresh razor blade on ice and tissue dispersed between two DNase/RNase free microcentrifuge tubes. Samples were flash frozen in liquid nitrogen and stored at −80 °C until further processing. Total RNA was isolated from retinal tissue aliquots using RNeasy Micro kits (Qiagen), according to the manufacturers’ protocols, including the optional DNase digest step. Tissues were homogenized in RLT buffer (Qiagen) using a Pellet Pestle Motor (Kimbal). RNA concentration and quality were assessed using the Nanodrop 2000 spectrophotometer (Thermo Scientific) and the RNA 6000 Nano and Pico Bioanalyzer Assay (Agilent Technologies). Libraries were constructed using the KAPA mRNA HyperPrep Kit (Roche Sequencing and Life Science), according to the manufacturer’s protocol. Libraries were assessed using D5000 ScreenTape (Agilent Technologies) and the Qubit dsDNA HS Assay (ThermoFisher) according to the manufacturers’ instructions. Approximately, 44 M 150 bp paired-end reads were sequenced per sample on an Illumina NovaSeq 6000 using the S4 Reagent Kit v1.5 (Illumina, Cat#20,028,313) by the Genome Technologies Core at The Jackson Laboratory. Raw FASTQ files were processed using standard quality control practices. High quality read pairs were aligned to a custom pseudogenome produced by incorporating strain-specific variants (single nucleotide variants and short insertions/deletions derived from Mouse Genomes Project REL-1505) into the mouse reference genome (GRCm38/mm10) using g2gtools version 0.2.7 ( https://github.com/churchill-lab/g2gtools ). Reads were aligned using STAR version 2.7.10a. Transcript counts were generated using RSEM version 1.3.3 . Differential expression analyses were completed using edgeR v3.40.2 within the R environment v4.2.3 using the glmQLFtest function . filterByExpr within edgeR was used to filter out very lowly expressed genes prior to TMM normalization . Differential expression was analyzed using generalized linear models via edgeR, with either 1) groups defined as a combination of strain, sex, age and a library batch term (~ 0 + Group + Batch) and contrasts designed to test for average strain effect, average age effect within each strain; or 2) age as a continuous variable to control and test for differential expression related to strain differences or determining the impact of aging while controlling for strain and sex differences. The clusterProfiler v4.6.2 R package was used to test for overrepresentation of gene ontology (GO) Biological Process gene sets within the list of differentially expressed genes (DEGs) with a false-discovery rate (FDR) of less than 0.05 and enrichPlot v1.18.4 and EnhancedVolcano v1.16.0 were used to visualize the enrichment results . Sample Preparation Each half retina sample had ice-cold 600 µL of extraction buffer (2:2:1 methanol:acetonitrile:water) added, along with a pre-chilled 5 mm stainless steel bead (QIAGEN). Tubes were then added to the pre-chilled (at −20 °C) cassettes for the Tissue Lyser II in batches of 48 samples, then were lysed in the Tissue Lyser II for 2 min at 30 1/s. The stainless steel beads were removed with a magnet and the samples were placed at −20 °C for 16 h overnight. Sample extracts were then centrifuged at 21,000 × g at 4 °C for 15 min. The supernatant was removed, and the protein pellet was reconstituted in 100 µL of 50 mM HEPES buffer, pH 8.0. Reconstituted samples were vortexed at max speed for 30 s, a chilled steel bead was added, and they were reconstituted using the Tissue Lyser II as described for the extraction. Samples were then waterbath sonicated (sweep at 37 Hz at 100% power) for 5 min (30 s on, 30 s off for five cycles) with ice added to keep the temperature from rising. Samples were assessed for clarity and if necessary, the sonication was repeated. Samples were then centrifuged at 21,000 × g for 15 min at 4 °C and protein supernatant was transferred to a clean microcentrifuge tube. Protein quantification was then performed using a microBCA assay (Thermo, Cat# 23,235) on a single well for each sample with a 1:25 dilution. The remaining sample was snap-frozen and stored at −80 °C until further use. Protein digestion and peptide purification After protein quantification, 20 µg of each sample were aliquoted to a new tube and brought to an equal volume in 50 mM HEPES, pH 8.0. Samples with less than 20 µg total required use of the full sample for digest and the trypsin ratio was adjusted at the digestion stage. Samples were reduced with 10 mM dithiothreitol at 42 °C for 30 min on a ThermoMixer with 500 rpm agitation, alkylated with 15 mM iodoacetamide for 20 min at room temperature on a ThermoMixer with 500 rpm agitation, and trypsin-digested (Sequence Grade Modified; Promega) with a 1:50 trypsin:protein ratio at 37 °C for 20 h. Following the digest, all peptide samples were C18-purified using Millipore C18 zip-tips (Millipore, Cat#: ZTC18S096) according to the manufacturer protocol’s and as previously reported . Eluted purified peptides were dried with a vacuum centrifuge and stored at −20 °C until tandem mass tag (TMT) labeling was performed. TMTpro peptide labeling The dried peptide samples were reconstituted in 20 µL of 50 mM HEPES, pH 8.5) and quantified using the Quantitative Colorimetric Peptide Assay (Thermo, Cat# 23,275) according to the manufacturer’s protocol. Samples were diluted to a concentration of ~ 0.5 µg/µL in 10 µL of 50 mM HEPES, pH 8.5, and mixed on a ThermoMixer for 10 min at 25 °C (500 rpm). To create a quality control pool for a carrier channel, equal amounts of all samples were used to create enough pool that could be used in each multiplex for normalization. While the peptide solutions mixed, TMTpro reagents (Thermo, Cat# A44520) were mixed according to manufacturer protocol. TMTpro labels were added to a 19.37 mM final concentration to the appropriate peptide sample, followed by a 1 h incubation at 25 °C with 500 rpm agitation on a ThermoMixer. As the study required many multiplexes the samples were first randomized for the multiplex assignment and then were randomized for the TMT tag assignment within a multiplex through a list randomization in Random.org. Reactions were then quenched by adding a final concentration of 0.5% hydroxylamine and incubating for 15 min at 25 °C with 400 rpm agitation on a ThermoMixer. All samples were added to the appropriate TMT multiplex group, acidified by adding 20 µL of 10% formic acid, snap-frozen, and stored at −80 °C overnight. The next day samples were dried in a vacuum centrifuge. Liquid Chromatography Tandem Mass Spectrometry Analysis (LC–MS/MS) Each of the dried multiplexed peptide samples were reconstituted in 25 µL of LC–MS grade water with 0.1% TFA and zip-tipped using Millipore C18 zip-tips according to the manufacturer protocol’s and as previously reported . Purified multiplexes were then dried in a vacuum centrifuge and reconstituted in 20 µL of 98% H2O/2% ACN with 0.1% formic acid via pipetting and vortexing. Reconstituted samples were then transferred to a mass spec vial and placed in the autosampler at 4 °C. LC–MS/MS was then performed on a Thermo Eclipse Tribrid Orbitrap with a FAIMS coupled to an UltiMate 3000 nano-flow LC system in The Jackson Laboratory Mass Spectrometry and Protein Chemistry Service. The method duration was 180 min at a flow rate of 300 nL/min. Buffer A (100% H2O with 0.1% formic acid) and Buffer B (100% acetonitrile with 0.1% formic acid) were utilized for the gradient. The full gradient consisted of 98% A/2% B from 0–2 min, 98% A/2% B at 2 min to 92.5% A/7.5% B at 15 min, 92.5% A/7.5% B at 15 min to 70% A/30% B at 145 min, and 70% A/30% B at 145 min to 10% A/90% B at 155 min. The gradient was held at 10% A/90% B until 160 min and brought to 98% A/2% B by 162 min, where it was then held for 15 min to equilibrate the column. The TMT SP3 MS3 real-time search (RTS) method was used on the instrument. Global parameters included a default charge = 2, expected peak width = 30 s, advanced peak determination, spray voltage = 2000 V, mode = positive, FAIMS carrier gas = 4.6 L/min, and an ion transfer tube temperature = 325 °C. The instrument method utilized the FAIMS voltages of −40 V, −55 V, and −65 V. Settings for precursor spectra detection (MS1) in each node included: cycle time = 1 s (each node), detector = Orbitrap, Orbitrap resolution = 120,000, scan range = 400–1600 m/z, RF lens % = 30, normalize AGC target = 250, maximum inject time (ms) = auto, microscans = 1, data type = profile, polarity = positive, monoisotopic precursor selection = peptide, minimum intensity threshold = 5.0e3 (lower because of the FAIMS), charge states = 2–7, and a dynamic exclusion of a n = 1 for 60 s. Peptide fragment analysis (MS2) was performed in the ion trap and settings included: isolation window (m/z) = 0.7, collision energy = 35 (fixed), activation type = CID, CID activation time = 10 ms, quadrupole isolation mode, ion trap scan rate = turbo, maximum inject time = 35 ms, and data type = centroid. Prior to data-dependent MS3 (ddMS3) analysis, RTS was utilized with the UniProtKB Mus musculus (sp_tr_incl_isoforms TaxID = 10,090) protein database including cysteine carbamidomethylation (+ 57.0215 Da), TMTpro16plex (+ 304.20171 on Kn), and methionine oxidation (+ 15.9949). Additional parameters in the RTS included maximum missed cleavages = 2, Xcorr threshold of 1, dCn threshold of 0.1, and a precursor ppm of 10. SP3 MS3 was performed in the Orbitrap and settings included SPS precursors = 20, isolation window = 0.7 m/z, activation type = HCD, HCD energy normalized at 45%, resolution = 60,000, scan range = 100–500 m/z, normalized AGC target = 500%, maximum injection time = 118 ms, and centroid data collection. LC–MS/MS data analysis All of the Thermo Eclipse RAW data files from the Eclipse Tribrid Orbitrap mass spectrometer were searched against the UniProtKB Mus musculus (sp_tr_incl_isoforms TaxID = 10,090) protein database in Proteome Discoverer (Thermo Scientific, version 2.5.0.400) using Sequest HT according to standard manufacturer recommended workflows. The Sequest protein database search parameters included trypsin cleavage, precursor mass tolerance = 20 ppm, fragment mass tolerance = 0.5 Da, static cysteine carbamidomethylation (+ 57.021 Da), static TMTpro modification on any N-terminus (+ 304.207 Da), and a variable methionine oxidation (+ 15.995 Da). Other setting included a maximum number of missed cleavages = 2, minimum peptide length = 6, a maximum of 144 amino acids, and a fragment mass tolerance = 0.6 Da. Percolator was used and in this module in the software, the target/decoy selection was concatenated, q-value validation was utilized, a maximum delta Cn = 0.05 was set, and FDR < 0.05 for all matches was used as the threshold. Default Minora node parameters were used. Abundance values were then normalized to the total ion signal in the samples using the Reporter Ions Quantifier Node for TMT-based MS3 events. Differential proteomics analyses Identified peptides with FDR < 0.05 confidence and detected in at least three samples were used for further investigation. log 2 (scaled abundance + 1) values were used for differential expression analyses. Differentially expressed proteins (DEPs) were determined using generalized linear models and a repeated measures model via limma and DEqMS , which takes into account the number of unique peptide spectrum matches (PSMs) used for each peptide to enhance the statistical power of differential expression, with either 1) groups defined as a combination of strain, sex, age and a TMT multiplex batch term (~ 0 + Group + Multiplex) and contrasts designed to test for strain effects, age effects within each strain; or 2) age as a continuous variable to control and test for differential expression related to strain differences or determining the impact of aging while controlling for strain and sex differences. As samples were run in triplicate, we utilized the duplicateCorrelation function within limma to block on “mouseID” . As for the transcriptomic analyses, clusterProfiler v4.6.2 R package was used to test for overrepresentation of Gene Ontology (GO) Biological Process gene sets within the list of differentially expressed proteins with a FDR of less than 0.05 and enrichPlot v1.18.4 and EnhancedVolcano v1.16.0 were used to visualize the enrichment results . STRING-dB v 2.14.3 R package was used to visualize protein–protein interaction networks with the settings “score_threshold = 200” and “network_type = physical” and to calculate enrichment of GO gene sets within the list of DEPs identified between 4 M NZO or WSB with the other 5 pigmented strains . Each half retina sample had ice-cold 600 µL of extraction buffer (2:2:1 methanol:acetonitrile:water) added, along with a pre-chilled 5 mm stainless steel bead (QIAGEN). Tubes were then added to the pre-chilled (at −20 °C) cassettes for the Tissue Lyser II in batches of 48 samples, then were lysed in the Tissue Lyser II for 2 min at 30 1/s. The stainless steel beads were removed with a magnet and the samples were placed at −20 °C for 16 h overnight. Sample extracts were then centrifuged at 21,000 × g at 4 °C for 15 min. The supernatant was removed, and the protein pellet was reconstituted in 100 µL of 50 mM HEPES buffer, pH 8.0. Reconstituted samples were vortexed at max speed for 30 s, a chilled steel bead was added, and they were reconstituted using the Tissue Lyser II as described for the extraction. Samples were then waterbath sonicated (sweep at 37 Hz at 100% power) for 5 min (30 s on, 30 s off for five cycles) with ice added to keep the temperature from rising. Samples were assessed for clarity and if necessary, the sonication was repeated. Samples were then centrifuged at 21,000 × g for 15 min at 4 °C and protein supernatant was transferred to a clean microcentrifuge tube. Protein quantification was then performed using a microBCA assay (Thermo, Cat# 23,235) on a single well for each sample with a 1:25 dilution. The remaining sample was snap-frozen and stored at −80 °C until further use. After protein quantification, 20 µg of each sample were aliquoted to a new tube and brought to an equal volume in 50 mM HEPES, pH 8.0. Samples with less than 20 µg total required use of the full sample for digest and the trypsin ratio was adjusted at the digestion stage. Samples were reduced with 10 mM dithiothreitol at 42 °C for 30 min on a ThermoMixer with 500 rpm agitation, alkylated with 15 mM iodoacetamide for 20 min at room temperature on a ThermoMixer with 500 rpm agitation, and trypsin-digested (Sequence Grade Modified; Promega) with a 1:50 trypsin:protein ratio at 37 °C for 20 h. Following the digest, all peptide samples were C18-purified using Millipore C18 zip-tips (Millipore, Cat#: ZTC18S096) according to the manufacturer protocol’s and as previously reported . Eluted purified peptides were dried with a vacuum centrifuge and stored at −20 °C until tandem mass tag (TMT) labeling was performed. The dried peptide samples were reconstituted in 20 µL of 50 mM HEPES, pH 8.5) and quantified using the Quantitative Colorimetric Peptide Assay (Thermo, Cat# 23,275) according to the manufacturer’s protocol. Samples were diluted to a concentration of ~ 0.5 µg/µL in 10 µL of 50 mM HEPES, pH 8.5, and mixed on a ThermoMixer for 10 min at 25 °C (500 rpm). To create a quality control pool for a carrier channel, equal amounts of all samples were used to create enough pool that could be used in each multiplex for normalization. While the peptide solutions mixed, TMTpro reagents (Thermo, Cat# A44520) were mixed according to manufacturer protocol. TMTpro labels were added to a 19.37 mM final concentration to the appropriate peptide sample, followed by a 1 h incubation at 25 °C with 500 rpm agitation on a ThermoMixer. As the study required many multiplexes the samples were first randomized for the multiplex assignment and then were randomized for the TMT tag assignment within a multiplex through a list randomization in Random.org. Reactions were then quenched by adding a final concentration of 0.5% hydroxylamine and incubating for 15 min at 25 °C with 400 rpm agitation on a ThermoMixer. All samples were added to the appropriate TMT multiplex group, acidified by adding 20 µL of 10% formic acid, snap-frozen, and stored at −80 °C overnight. The next day samples were dried in a vacuum centrifuge. Each of the dried multiplexed peptide samples were reconstituted in 25 µL of LC–MS grade water with 0.1% TFA and zip-tipped using Millipore C18 zip-tips according to the manufacturer protocol’s and as previously reported . Purified multiplexes were then dried in a vacuum centrifuge and reconstituted in 20 µL of 98% H2O/2% ACN with 0.1% formic acid via pipetting and vortexing. Reconstituted samples were then transferred to a mass spec vial and placed in the autosampler at 4 °C. LC–MS/MS was then performed on a Thermo Eclipse Tribrid Orbitrap with a FAIMS coupled to an UltiMate 3000 nano-flow LC system in The Jackson Laboratory Mass Spectrometry and Protein Chemistry Service. The method duration was 180 min at a flow rate of 300 nL/min. Buffer A (100% H2O with 0.1% formic acid) and Buffer B (100% acetonitrile with 0.1% formic acid) were utilized for the gradient. The full gradient consisted of 98% A/2% B from 0–2 min, 98% A/2% B at 2 min to 92.5% A/7.5% B at 15 min, 92.5% A/7.5% B at 15 min to 70% A/30% B at 145 min, and 70% A/30% B at 145 min to 10% A/90% B at 155 min. The gradient was held at 10% A/90% B until 160 min and brought to 98% A/2% B by 162 min, where it was then held for 15 min to equilibrate the column. The TMT SP3 MS3 real-time search (RTS) method was used on the instrument. Global parameters included a default charge = 2, expected peak width = 30 s, advanced peak determination, spray voltage = 2000 V, mode = positive, FAIMS carrier gas = 4.6 L/min, and an ion transfer tube temperature = 325 °C. The instrument method utilized the FAIMS voltages of −40 V, −55 V, and −65 V. Settings for precursor spectra detection (MS1) in each node included: cycle time = 1 s (each node), detector = Orbitrap, Orbitrap resolution = 120,000, scan range = 400–1600 m/z, RF lens % = 30, normalize AGC target = 250, maximum inject time (ms) = auto, microscans = 1, data type = profile, polarity = positive, monoisotopic precursor selection = peptide, minimum intensity threshold = 5.0e3 (lower because of the FAIMS), charge states = 2–7, and a dynamic exclusion of a n = 1 for 60 s. Peptide fragment analysis (MS2) was performed in the ion trap and settings included: isolation window (m/z) = 0.7, collision energy = 35 (fixed), activation type = CID, CID activation time = 10 ms, quadrupole isolation mode, ion trap scan rate = turbo, maximum inject time = 35 ms, and data type = centroid. Prior to data-dependent MS3 (ddMS3) analysis, RTS was utilized with the UniProtKB Mus musculus (sp_tr_incl_isoforms TaxID = 10,090) protein database including cysteine carbamidomethylation (+ 57.0215 Da), TMTpro16plex (+ 304.20171 on Kn), and methionine oxidation (+ 15.9949). Additional parameters in the RTS included maximum missed cleavages = 2, Xcorr threshold of 1, dCn threshold of 0.1, and a precursor ppm of 10. SP3 MS3 was performed in the Orbitrap and settings included SPS precursors = 20, isolation window = 0.7 m/z, activation type = HCD, HCD energy normalized at 45%, resolution = 60,000, scan range = 100–500 m/z, normalized AGC target = 500%, maximum injection time = 118 ms, and centroid data collection. All of the Thermo Eclipse RAW data files from the Eclipse Tribrid Orbitrap mass spectrometer were searched against the UniProtKB Mus musculus (sp_tr_incl_isoforms TaxID = 10,090) protein database in Proteome Discoverer (Thermo Scientific, version 2.5.0.400) using Sequest HT according to standard manufacturer recommended workflows. The Sequest protein database search parameters included trypsin cleavage, precursor mass tolerance = 20 ppm, fragment mass tolerance = 0.5 Da, static cysteine carbamidomethylation (+ 57.021 Da), static TMTpro modification on any N-terminus (+ 304.207 Da), and a variable methionine oxidation (+ 15.995 Da). Other setting included a maximum number of missed cleavages = 2, minimum peptide length = 6, a maximum of 144 amino acids, and a fragment mass tolerance = 0.6 Da. Percolator was used and in this module in the software, the target/decoy selection was concatenated, q-value validation was utilized, a maximum delta Cn = 0.05 was set, and FDR < 0.05 for all matches was used as the threshold. Default Minora node parameters were used. Abundance values were then normalized to the total ion signal in the samples using the Reporter Ions Quantifier Node for TMT-based MS3 events. Identified peptides with FDR < 0.05 confidence and detected in at least three samples were used for further investigation. log 2 (scaled abundance + 1) values were used for differential expression analyses. Differentially expressed proteins (DEPs) were determined using generalized linear models and a repeated measures model via limma and DEqMS , which takes into account the number of unique peptide spectrum matches (PSMs) used for each peptide to enhance the statistical power of differential expression, with either 1) groups defined as a combination of strain, sex, age and a TMT multiplex batch term (~ 0 + Group + Multiplex) and contrasts designed to test for strain effects, age effects within each strain; or 2) age as a continuous variable to control and test for differential expression related to strain differences or determining the impact of aging while controlling for strain and sex differences. As samples were run in triplicate, we utilized the duplicateCorrelation function within limma to block on “mouseID” . As for the transcriptomic analyses, clusterProfiler v4.6.2 R package was used to test for overrepresentation of Gene Ontology (GO) Biological Process gene sets within the list of differentially expressed proteins with a FDR of less than 0.05 and enrichPlot v1.18.4 and EnhancedVolcano v1.16.0 were used to visualize the enrichment results . STRING-dB v 2.14.3 R package was used to visualize protein–protein interaction networks with the settings “score_threshold = 200” and “network_type = physical” and to calculate enrichment of GO gene sets within the list of DEPs identified between 4 M NZO or WSB with the other 5 pigmented strains . All in vivo investigations were performed in B6, NZO, and WSB mice at 4, 8, 12, and/or 18 M. In vivo techniques included: slit lamp imaging, fundus exams, fluorescein angiography, and electroretinography (full-field and pattern). Slit lamp imaging Mice were anesthetized with ketamine/xylazine. Animals were placed on the apparatus with eyes aligned to the camera. Bright light images were obtained using Topcon DC-4 digital slit lamp and EZ Capture image software. Electroretinography (ERG) Mice were placed in an anesthesia induction chamber infused with 3–4% isoflurane. Once fully anesthetized, mice were transferred to the heated platform of the Celeris-Diagnosys ERG system (Diagnosys LLC, MA, USA) to maintain body temperature at approximately 37 °C, where anesthesia was sustained using 1–2% isoflurane. 1% Tropicamide and 2.5% phenylephrine were topically applied to dilate the pupils of both eyes, while a thin layer of 2.5% hypromellose was used to keep the eyes moist. Two corneal electrodes with integrated stimulators were placed on the surfaces of lubricated corneas to facilitate full-field ERG recordings. The Scotopic ERG was conducted initially, with nine steps of stimulus intensity (0.001, 0.002, 0.01, 0.0316, 0.316, 1,3, 10, and 31.6 cd.s/m 2 ). The final ERG waveform was the average of 10 individual waveforms. Following a 10 min light adaptation interval, the photopic ERG was evoked by five steps of stimulus intensity (0.316, 1, 3, 10, 31.6 cd.s/m 2 ). The final ERG waveform was an average of 20 individual waveforms. The amplitude of the a-wave was measured as the difference between the pre-stimulus baseline and the trough of the a-wave, and the b-wave amplitude was measured from the trough of the a-wave to the peak of the b-wave. OP amplitudes were measured from the pre-stimulus baseline to the highest OP peak. The highest amplitude values across strains were elicited by 10 cd.s/m 2 luminance, thus, amplitudes generated at 10 cd.s/m 2 were compared. Pattern electroretinography (PERG) All PERGs were performed using a JORVEC PERG system (Intelligent Hearing Systems, Miami, Florida) as described by the manufacturer. Briefly, mice were anesthetized with ketamine/xylazine and slit lamp was used to inspect each mouse eye prior to starting experiment. Mice were kept on a warming stage throughout the experiment. Electrodes were placed subcutaneously such that the active electrode is placed between the eyes with the tip just before the snout and the reference electrode is placed in line with the active electrode just between the ears. The ground electrode was placed approximately 2 cm in front of the base of the tail. Scans were collected and waveforms were averaged using default settings. PERG amplitude was calculated as P1-N2 for each eye per mouse. Optical Coherence Tomography (OCT), fundus, and fluorescein angiography OCT, fundus exams, and fluorescein angiographies were performed using Micron IV and Reveal OCT and Discover 2.4 software (Phoenix-Micron, Bend, OR). Mice were given 1 drop (20-30µl) of 0.5% Tropicamide (Somerset Therapeutics, NDC# 70,069–121-01) in both eyes. After 10 min, one drop (20-30µl) of 2.5% phenylephrine (Bausch & Lomb, NDC# 42,702–0103-05) was applied to both eyes. Mice were then anesthetized using 4% isoflurane until a proper plane of anesthesia was achieved. Mice were then transferred to a nose cone on a heated holding cradle. Anesthesia was then reduced to 2% isoflurane. Eyes were kept moist using GenTeal Tears Lubricant Eye Gel Drops (Alcon). Respiration rate was monitored, and the camera lens was adjusted to be perpendicular to the cornea. For OCT imaging, focus and brightness of the image guided OCT was adjusted until optimal image is previewed. OCT images were captured from the temporal to the nasal retina in a plane that included the optic nerve. To improve resolution, 10–40 images were taken and averaged. For fundus exams, a white light fundus image was taken of each eye. Brightness adjustments were made, and the focal plane was optimally adjusted. Immediately following the image-guided OCT and fundus imaging, mice were injected intraperitoneally with 1% Fluorescite® Fluorescein Sodium at a dose of 10 mg/kg (10 mg/mL solution, Akorn). Using a GFP filter, images of the right eye were taken every 30 s for 6 min. Images were taken of the left eye 30 s after the right eye imaging period. Retinal layers from OCT images were measured using the FIJI distribution of ImageJ . Each layer was measured 200, 400, and 600 µm from the edge of the optic nerve using the line measurement tool, and the three respective measurements were averaged and used for statistical analyses. Mice were anesthetized with ketamine/xylazine. Animals were placed on the apparatus with eyes aligned to the camera. Bright light images were obtained using Topcon DC-4 digital slit lamp and EZ Capture image software. Mice were placed in an anesthesia induction chamber infused with 3–4% isoflurane. Once fully anesthetized, mice were transferred to the heated platform of the Celeris-Diagnosys ERG system (Diagnosys LLC, MA, USA) to maintain body temperature at approximately 37 °C, where anesthesia was sustained using 1–2% isoflurane. 1% Tropicamide and 2.5% phenylephrine were topically applied to dilate the pupils of both eyes, while a thin layer of 2.5% hypromellose was used to keep the eyes moist. Two corneal electrodes with integrated stimulators were placed on the surfaces of lubricated corneas to facilitate full-field ERG recordings. The Scotopic ERG was conducted initially, with nine steps of stimulus intensity (0.001, 0.002, 0.01, 0.0316, 0.316, 1,3, 10, and 31.6 cd.s/m 2 ). The final ERG waveform was the average of 10 individual waveforms. Following a 10 min light adaptation interval, the photopic ERG was evoked by five steps of stimulus intensity (0.316, 1, 3, 10, 31.6 cd.s/m 2 ). The final ERG waveform was an average of 20 individual waveforms. The amplitude of the a-wave was measured as the difference between the pre-stimulus baseline and the trough of the a-wave, and the b-wave amplitude was measured from the trough of the a-wave to the peak of the b-wave. OP amplitudes were measured from the pre-stimulus baseline to the highest OP peak. The highest amplitude values across strains were elicited by 10 cd.s/m 2 luminance, thus, amplitudes generated at 10 cd.s/m 2 were compared. All PERGs were performed using a JORVEC PERG system (Intelligent Hearing Systems, Miami, Florida) as described by the manufacturer. Briefly, mice were anesthetized with ketamine/xylazine and slit lamp was used to inspect each mouse eye prior to starting experiment. Mice were kept on a warming stage throughout the experiment. Electrodes were placed subcutaneously such that the active electrode is placed between the eyes with the tip just before the snout and the reference electrode is placed in line with the active electrode just between the ears. The ground electrode was placed approximately 2 cm in front of the base of the tail. Scans were collected and waveforms were averaged using default settings. PERG amplitude was calculated as P1-N2 for each eye per mouse. OCT, fundus exams, and fluorescein angiographies were performed using Micron IV and Reveal OCT and Discover 2.4 software (Phoenix-Micron, Bend, OR). Mice were given 1 drop (20-30µl) of 0.5% Tropicamide (Somerset Therapeutics, NDC# 70,069–121-01) in both eyes. After 10 min, one drop (20-30µl) of 2.5% phenylephrine (Bausch & Lomb, NDC# 42,702–0103-05) was applied to both eyes. Mice were then anesthetized using 4% isoflurane until a proper plane of anesthesia was achieved. Mice were then transferred to a nose cone on a heated holding cradle. Anesthesia was then reduced to 2% isoflurane. Eyes were kept moist using GenTeal Tears Lubricant Eye Gel Drops (Alcon). Respiration rate was monitored, and the camera lens was adjusted to be perpendicular to the cornea. For OCT imaging, focus and brightness of the image guided OCT was adjusted until optimal image is previewed. OCT images were captured from the temporal to the nasal retina in a plane that included the optic nerve. To improve resolution, 10–40 images were taken and averaged. For fundus exams, a white light fundus image was taken of each eye. Brightness adjustments were made, and the focal plane was optimally adjusted. Immediately following the image-guided OCT and fundus imaging, mice were injected intraperitoneally with 1% Fluorescite® Fluorescein Sodium at a dose of 10 mg/kg (10 mg/mL solution, Akorn). Using a GFP filter, images of the right eye were taken every 30 s for 6 min. Images were taken of the left eye 30 s after the right eye imaging period. Retinal layers from OCT images were measured using the FIJI distribution of ImageJ . Each layer was measured 200, 400, and 600 µm from the edge of the optic nerve using the line measurement tool, and the three respective measurements were averaged and used for statistical analyses. Mice were anesthetized with tribromoethanol and then transcardially perfused with 4% paraformaldehyde (PFA) (Electron Microscopy Services, Cat#15,714) and 3 mg/mL 70 kDa fluorescein isothiocyanate–dextran (Millipore-Sigma, Cat#FD70S) in PBS. Eye globes were excised, and retinas were immediately dissected and flat mounted. Flat mounts were imaged with a Leica DMi8 inverted microscope within 30 min of harvest using a GFP filter by taking 10 µm step z-stacked imaged to encompass the entire tissue at 10 × magnification. The presence of leaks and avascular areas were recorded. Tissue integrity was assessed by bright field microscopy. At the time of sacrifice, animals were anesthetized with tribromoethanol and were transcardially perfused with PBS. Eyes were enucleated and fixed in 37.5% methanol and 12.5% acetic acid in PBS for 18 h at 4 °C. The Histology Core at The Jackson Laboratory performed paraffin embedding, sectioning, and hematoxylin and eosin and Prussian blue staining of optic nerve head-containing sections. Slides containing 2–3 sections per eye were imaged using a Hamamatsu S210 NanoZoomer Digital Slide Scanner at 40X magnification by the Microscopy Core at The Jackson Laboratory. Images were processed using custom FIJI scripts. Briefly, 400 µm x 400 µm regions of interest (ROIs) were selected on both sides of the optic nerve head at approximate distances of 500, 1000 and 1500 µm representing the central, middle, and peripheral retina. After which, additional ROIs of the ONL per image were generated for automated thresholding, watershed segmentation, and quantification. The default thresholding was utilized with a size threshold of 30 µm 2 for automated counting for all ONL cells. To determine the length of the retina, we first generated a mask of the retina by applying 1) gaussian blur, 2) color thresholding, 3) filling holes in the mask, and finally calculated the length of the mask. At the time of sacrifice, animals were anesthetized with tribromoethanol and were transcardially perfused with PBS. Eyes were enucleated and fixed in 4% PFA in PBS for 18 h at 4 °C. The retinal vascular network was isolated as previously described with a few modifications. Briefly, retinas were gently dissected and washed 3 × with ddH20 for 5 min and then left overnight in ddH20 at RT. The following day, retinas were incubated in 2.5% Trypsin (ThermoFisher Scientific, Cat#15,090,046) for 2 h at 37 °C. After this, the retinas were washed 5 × with ddH20. Next, the vessels were gently washed from the neural retina with additional ddH20. Retinal vascular networks were then mounted onto glass slides and allowed to dry before staining. The vascular networks were stained with hematoxylin and eosin according to the manufacturer’s instructions (Abcam, Cat#ab245880). Stained networks were dehydrated with 3 washes with reagent alcohol and then mounted using organic mounting medium (Organo/Limonene Mount, Sigma-Aldrich, Cat#O8015). Dried slides were imaged using a Hamamatsu S210 NanoZoomer Digital Slide Scanner at 40X magnification by the Microscopy Core at The Jackson Laboratory. Acellular capillaries were manually counted across 4–7 400–500 µm × 400–500 µm ROIs of each retina, and the average count was used for statistical analyses. Whole mount IHC At the time of sacrifice, animals were anesthetized with tribromoethanol and were transcardially perfused with PBS. Eyes were enucleated and fixed in 4% PFA in PBS for 18h at 4°C. Retinas were gently dissected from the eye cup and washed 3 times in PBS for 5 min, then washed 3 times in 0.3% Triton-X 100 in PBS for 5 min. Retinas were subsequently blocked in 10% Donkey Serum (Sigma-Aldrich, Cat#D9663) in 0.3% TritonX PBS for 24h at 4°C. Retinas were incubated in primary antibodies diluted in blocking buffer for 72h at 4°C and then washed 3× with PBS for 5 min. Retinas were then incubated with secondary antibodies in PBS for 24 hours at 4°C. Finally, retinas were washed 4 times with PBS for 5 min prior to whole mounting onto slides with fluorescent mounting medium (Polysciences, Cat#18606-20). Slides were allowed to dry prior to imaging. Primary antibodies utilized include: 1µg/mL Rabbit anti-RNA Binding Protein, MRNA Processing Factor (RBPMS) (GeneTex Cat#GTX118619). 0.5 μg/mL Donkey anti-Rabbit Alexa Fluor 568 (Invitrogen, Cat#A10042) was used as a secondary antibody. Multicolor wide-field tiled images were taken on a Leica DMi8 microscope a20× magnification across the entire retina. Retina images were stitched together in FIJI. All microscope settings were kept identical for each experiment. RBPMS + RGC counts were determined by generating 4 separate 500µm x 500µm ROIs from the peripheral and central retina for each mouse. These ROIs were then subjected to automated thresholding using the moments algorithm , and watershed segmentation before counting all RBPMS + objects above 100µm 2 . All analyses were performed with the FIJI distribution of ImageJ (NIH) . Retinal cross-sections IHC Paraffin embedded sections were de-paraffinized using 3 × 5 min washes with Clear-Rite 3 (Thermofisher, Cat#6901TS) and then rehydrated with a series of 5 min Ethanol washes (2 × 100%, 1 × 90%, 1 × 70%) and then washed 3 × with ddH20. Antigen retrieval was performed by steaming in Epitope Retrieval Solution (IHC World, Cat#IW-1100) for 1 h. Sections were permeabilized for 10 min with 0.3% TritionX100 in PBS then washed 3 × PBS. Sections were subsequently blocked in 10% Donkey Serum (Sigma-Aldrich, Cat#D9663) in PBS for 24 h at 4 °C. Sections were incubated in primary antibodies diluted in blocking buffer for 72 h at 4 °C and then washed 3 × with PBS for 5 min. Sections were then incubated with secondary antibodies in PBS for 2 h at room temperature. Finally, sections were incubated with DAPI for 5 min and then washed 4 × PBS for 5 min prior to cover slipping with fluorescent mounting medium (Polysciences, Cat#18,606–20). Slides were allowed to dry prior to imaging. Primary antibodies utilized include: 5 µg/mL Rabbit anti-Red/Green Opsin (Sigma-Aldrich, Cat#AB5405), 2 µg/mL Mouse anti-Rhodopsin [clone:1D4](Abcam,Cat#AB5417). Secondary antibodies utilized include: 0.5 µg/mL Donkey anti-Rabbit Alexa Fluor 647 (Invitrogen, Cat#A32795), and 0.5 µg/mL Donkey anti-Mouse Alexa Fluor 647 (Abcam, Cat#AB150107). Multicolor wide-field tiled images were taken on a Leica DMi8 microscope at 40 × magnification across the entire cross-section. All microscope settings were kept identical for each experiment. Images were stitched together in FIJI. Images were processed using custom FIJI scripts. Briefly, 400 µm x 400 µm regions of interest (ROIs) were selected on both sides of the optic nerve head at approximate distances of 500, 1000 and 1500 µm representing the central, middle, and peripheral retina. After which, additional ROIs of the ONL per image were generated for background removal (20 rolling ball radius), automated thresholding, segmentation, and quantification. For cones: the automated thresholding method “Triangle” was applied to each image, and objects with areas between 2–200 µm 2 were counted using FIJI. The resulting cone counts per image were normalized to the length of the IS/OS imaged and the percent change in retinal length over age for each strain as determined from histology. For the IS/OS analysis, ROIs containing the IS/OS were manually outlined, skeletonized, and the resulting skeleton was analyzed for length and area to calculate width. For rhodopsin staining: an ROI containing the IS/OS was used to evaluate rhodopsin signal by thresholding using the “Otsu” method and the resulting mask was used to calculate the area of positive rhodopsin. All analyses were performed with the FIJI distribution of ImageJ (NIH) . At the time of sacrifice, animals were anesthetized with tribromoethanol and were transcardially perfused with PBS. Eyes were enucleated and fixed in 4% PFA in PBS for 18h at 4°C. Retinas were gently dissected from the eye cup and washed 3 times in PBS for 5 min, then washed 3 times in 0.3% Triton-X 100 in PBS for 5 min. Retinas were subsequently blocked in 10% Donkey Serum (Sigma-Aldrich, Cat#D9663) in 0.3% TritonX PBS for 24h at 4°C. Retinas were incubated in primary antibodies diluted in blocking buffer for 72h at 4°C and then washed 3× with PBS for 5 min. Retinas were then incubated with secondary antibodies in PBS for 24 hours at 4°C. Finally, retinas were washed 4 times with PBS for 5 min prior to whole mounting onto slides with fluorescent mounting medium (Polysciences, Cat#18606-20). Slides were allowed to dry prior to imaging. Primary antibodies utilized include: 1µg/mL Rabbit anti-RNA Binding Protein, MRNA Processing Factor (RBPMS) (GeneTex Cat#GTX118619). 0.5 μg/mL Donkey anti-Rabbit Alexa Fluor 568 (Invitrogen, Cat#A10042) was used as a secondary antibody. Multicolor wide-field tiled images were taken on a Leica DMi8 microscope a20× magnification across the entire retina. Retina images were stitched together in FIJI. All microscope settings were kept identical for each experiment. RBPMS + RGC counts were determined by generating 4 separate 500µm x 500µm ROIs from the peripheral and central retina for each mouse. These ROIs were then subjected to automated thresholding using the moments algorithm , and watershed segmentation before counting all RBPMS + objects above 100µm 2 . All analyses were performed with the FIJI distribution of ImageJ (NIH) . Paraffin embedded sections were de-paraffinized using 3 × 5 min washes with Clear-Rite 3 (Thermofisher, Cat#6901TS) and then rehydrated with a series of 5 min Ethanol washes (2 × 100%, 1 × 90%, 1 × 70%) and then washed 3 × with ddH20. Antigen retrieval was performed by steaming in Epitope Retrieval Solution (IHC World, Cat#IW-1100) for 1 h. Sections were permeabilized for 10 min with 0.3% TritionX100 in PBS then washed 3 × PBS. Sections were subsequently blocked in 10% Donkey Serum (Sigma-Aldrich, Cat#D9663) in PBS for 24 h at 4 °C. Sections were incubated in primary antibodies diluted in blocking buffer for 72 h at 4 °C and then washed 3 × with PBS for 5 min. Sections were then incubated with secondary antibodies in PBS for 2 h at room temperature. Finally, sections were incubated with DAPI for 5 min and then washed 4 × PBS for 5 min prior to cover slipping with fluorescent mounting medium (Polysciences, Cat#18,606–20). Slides were allowed to dry prior to imaging. Primary antibodies utilized include: 5 µg/mL Rabbit anti-Red/Green Opsin (Sigma-Aldrich, Cat#AB5405), 2 µg/mL Mouse anti-Rhodopsin [clone:1D4](Abcam,Cat#AB5417). Secondary antibodies utilized include: 0.5 µg/mL Donkey anti-Rabbit Alexa Fluor 647 (Invitrogen, Cat#A32795), and 0.5 µg/mL Donkey anti-Mouse Alexa Fluor 647 (Abcam, Cat#AB150107). Multicolor wide-field tiled images were taken on a Leica DMi8 microscope at 40 × magnification across the entire cross-section. All microscope settings were kept identical for each experiment. Images were stitched together in FIJI. Images were processed using custom FIJI scripts. Briefly, 400 µm x 400 µm regions of interest (ROIs) were selected on both sides of the optic nerve head at approximate distances of 500, 1000 and 1500 µm representing the central, middle, and peripheral retina. After which, additional ROIs of the ONL per image were generated for background removal (20 rolling ball radius), automated thresholding, segmentation, and quantification. For cones: the automated thresholding method “Triangle” was applied to each image, and objects with areas between 2–200 µm 2 were counted using FIJI. The resulting cone counts per image were normalized to the length of the IS/OS imaged and the percent change in retinal length over age for each strain as determined from histology. For the IS/OS analysis, ROIs containing the IS/OS were manually outlined, skeletonized, and the resulting skeleton was analyzed for length and area to calculate width. For rhodopsin staining: an ROI containing the IS/OS was used to evaluate rhodopsin signal by thresholding using the “Otsu” method and the resulting mask was used to calculate the area of positive rhodopsin. All analyses were performed with the FIJI distribution of ImageJ (NIH) . Analyses were performed using GraphPad Prism 10 software. Comparisons of percent of instances were analyzed using a Chi-square test. Data from experiments designed to test differences between two groups (e.g., one measurement across two ages within strain) were subjected to a Shapiro–Wilk test to test normality and an F test to compare variance. For normally distributed data with equal variance, a two-tailed independent samples t test was utilized. For normally distributed data with unequal variance, a Welch’s t test was used. For non-normally distributed data, a Mann–Whitney test was used. Data from experiments designed to test differences among more than two groups across one condition (e.g. one measurement across strains at one timepoint) were subjected to a Shapiro–Wilk test to test normality and a Brown-Forsythe test to compare variance. Normally distributed data with equal variance were analyzed using a one-way ANOVA followed by Tukey’s post-hoc test. Data from experiments designed to detect differences among multiple groups and across two conditions (e.g. measurements across ages and sexes within strains) were analyzed using a two-way ANOVA followed by Holm-Sidak’s post-hoc test. P values < 0.05 were considered statistically significant. Throughout the manuscript, results are reported as mean ± standard error of the mean (SEM). Genetic background is a major source of transcriptomic and proteomic variation in the murine retina. Aging is an important risk factor for numerous vision-threatening diseases , yet, little is known about the interaction of age, sex, and genetic background. To better understand how aging across diverse genetic contexts may alter retinal function, we profiled the retinal transcriptome and proteome of nine genetically diverse mouse strains at the ages of 4, 12 and ≥ 18 M (Fig. A). Several strains had natural attrition , reducing the number of aged mice of particular strains available for tissue collection and data analysis (see Supplemental Table 1 for full details). Our initial analysis of the resulting transcriptomic and proteomic data included sex, strain, and age as covariates. After filtering the multi-omics data, we analyzed 22,928 genes and 4,854 proteins across these groups. Our transcriptomic analyses included many cell-type specific transcripts such as: Vwf (endothelial cells) , Rho (photoreceptors) , Gfap (astrocytes) , Glul (Müller glia) , Calb1 (horizontal cells) , Calb2 (amacrine cells) , Tmem119 (microglia) , and Rbpms (retinal ganglion cells). The detected proteins included uniquely expressed proteins in cell-types including rhodopsin (RHO, photoreceptors), glial fibrillary acidic protein (GFAP, astrocytes) and glutamine synthetase (GLUL, müller glia), Von Willebrand factor (VWF, endothelial cells), calbindin (CALB1, interneurons), and calretinin (CALB2, amacrine cells). First, we visualized transcriptomic and proteomic data using principal component analyses (PCA). This revealed that strain was the largest contributor to variation in the transcriptomic data (Fig. B) with more subtle patterns observed in the proteomic data (Fig. C). The less pronounced strain signature in proteomic data may be due to the more limited numbers of proteins analyzed but may also be due in part to buffering at the protein level . We next evaluated if the predictive strength of the transcriptome in determining protein abundance differed across strains and found that there were similar small (R = 0.19–0.28) but significant ( p < 0.0001) correlations of protein abundance and RNA expression regardless of strain or age (Supplemental Fig. ). As examples of the importance of strain variation in selected pathways regardless of age critical for retinal health, genes associated with the GO term “Regulation of complement”, including Cfh (Fig. D-E) or genes associated with the GO term “Detection of visible light”, including Crb1, displayed substantial strain-dependent variation (Fig. F-G). Proteins within the “Metabolic process” GO term, including acetyl-CoA acetyltransferase 1 (ACAT1), also displayed strain-dependent expression patterns (Fig. H-I). Thus, genetic context was a major source of variation in retinal molecular signatures. The aging retinal transcriptome and proteome are influenced by genetic context To define an overall retinal aging signature regardless of genetic context, we tested for differential gene and protein expression using age as the predictive variable across genetic backgrounds. We found that 4929 genes and 271 proteins were differentially expressed with aging across genetically diverse mice (Fig. A-B). Of these 271 differentially expressed proteins (DEPs), 131 were also significantly modulated at the gene level (Fig. C). Surprisingly, sex did not have a significant impact on the aging signatures across strains. Enrichment analyses revealed significant alteration of GO terms involving immune processes, photoreceptors, metabolism, and chromatin organization at both the gene and protein level (Fig. D-E). Differentially expressed photoreceptor genes and proteins decreased with age, while the differentially expressed genes (DEGs) and DEPs associated with immune processes largely increased with age (Supplemental Table 2). Next, we sought to examine how this aging signature was modified by genetic context. PCA visualization of the aging signature genes and proteins suggested differences in gene and protein expression between the strains of mice (Fig. A-B). For instance, genes associated with the GO term “Regulation of histone methylation”, including Mthfr (Fig. C-D), genes associated with the GO term “Antigen processing and peptide presentation”, including B2m (Fig. E-F) , and genes associated with the GO term “Retina homeostasis”, including Rho (Fig. G-H) all displayed substantial strain- and age-dependent variation. The GO pathway “Retina homeostasis” and Rho expression seemed to decrease with age in the albino strains and WSB mice more rapidly than the other 5 strains. Overall, these observed differences in the aging signature suggest genetic context dictates susceptibility to aging-associated retinal diseases. Common retinal neurodegenerative diseases involve RGC and/or photoreceptor loss . Single nucleus RNA-sequencing datasets have revealed several marker genes for specific retinal neurons . Therefore, we probed for changes in these neuronal populations by visualizing marker gene differences. We first profiled changes in photoreceptor-specific gene expression across strains using PCA (Fig. A). We found that WSB, NOD, and AJ mice showed age-associated shifts away from the other strains (Fig. A). In fact, these three strains exhibited reduced photoreceptor gene expression relative to the other strains at 18 M (Fig. B). In accordance with our transcriptomics data, AJ mice are known to develop age-associated loss of photoreceptors as shown by histological and functional analyses, while albino NOD mice can develop type 1 diabetes concomitant with neural and vascular damage by electron microscopy . In our analysis, WSB mice exhibited age-dependent differences in photoreceptor marker gene expression (Fig. C). WSB mice do not carry mutations in known retinal degeneration genes , and to date, there are no published reports of WSB photoreceptor loss. Thus, WSB mice may serve as a novel model of photoreceptor degeneration. In addition to photoreceptors, we profiled changes in RGC marker genes across strains. NZO, CAST, and BALBc mice exhibited a slight leftward shift in the PCA plot for RGC genes (Fig. D), and heatmap visualization of RGC marker genes revealed low expression of these genes in aged NZO, CAST and BALBc mice relative to other strains (Fig. E). A previous report suggested CAST mice develop significantly fewer RGCs (and thus have smaller optic nerves) than other strains, which may account for the reduced expression of RGC genes in this strain . Like many albino strains, BALBc mice have been shown to develop age-related retinal dysfunction exacerbated by vivarium light cycles by many measures including optical coherence tomography (OCT), electroretinography (ERG), histology and fundus imaging . However, there have been no previous reports of retinal phenotypes in NZO mice. To better understand how aging influences RGC gene expression in NZO mice, we visualized gene expression changes across age in NZO mice and found striking reductions in these genes (Fig. F). These data suggest NZO mice may be especially susceptible to RGC loss with age. Collectively, our multi-omics analysis suggested genetic context significantly contributes to retinal aging signatures. We created a user-friendly web application to increase data accessibility and foster hypothesis development and refinement, which can be accessed at: https://thejacksonlaboratory.shinyapps.io/Howell_AgingRetinaOmics/ . For example, one could search for an individual gene of interest (e.g., “ Sigmar1 ”) and visualize the RNA and/or protein expression (Supplemental Fig. A-B) across all strains. Alternatively, a set of genes/proteins can be visualized in a PCA style plot (Supplemental Fig. C). NZO and WSB mice exhibit disease-relevant retinal fundus abnormalities with age Our analyses suggested shared changes to photoreceptor markers across all nine strains, along with previously unreported age-dependent changes to retinal neurons in NZO and WSB mice. We therefore sought to verify these predictions using a battery of clinically-relevant in vivo ophthalmological exams at the ages of 4, 8, 12 and/or ≥ 18 M in NZO, WSB, and B6 mice of both sexes (Fig. A). Gross changes to the anterior segment of NZO, WSB, and B6 eyes were not detected using slit lamp imaging (Fig. B). In contrast, fundus imaging revealed age-associated increases of fundus abnormalities in WSB, NZO, and B6 mice (Fig. C-D). Fundus abnormalities in B6 mice included abnormal retinal spots (Fig. C-D). In NZO mice, these abnormalities included spots resembling cotton wool spots and exudates, which are commonly observed in human DR and other retinal vascular diseases . The presence of these phenotypes peaked by 12 M, however, what appeared to be an epiretinal membrane was uniquely observed in aged NZO eyes (7 of 38 NZO 18 M eyes; Supplemental Fig. ). WSB mice exhibited fundus abnormalities as early as 4 M, which became increasingly severe with age (Fig. C-D). These abnormalities included potential pigment disruption, retinal spots, and waxy optic disc pallor. Altogether, these data indicate age-related changes to DR-relevant fundus phenotypes in NZO mice and retinal degeneration-relevant phenotypes in WSB mice. OCT profiling unveils strain-specific patterns of retinal thinning To investigate whether fundus abnormalities corresponded with gross retinal degeneration, OCT was employed to measure retinal layer thicknesses across age in each strain (Fig. A). As total retinal thickness differed by strain in young animals (Fig. A-B), we determined the effect of age within each strain separately. Age reduced total retinal thickness across all three strains (Fig. B). There was a subtle but significant interaction effect between sex and age in NZO mice (Fig. B). The nerve fiber layer (NFL), ganglion cell layer (GCL), and inner plexiform layer (IPL) are largely composed of RGC axons, cell bodies, and dendritic processes/synapses, respectively. These layers were difficult to discern via OCT due to technical limitations, therefore, the NFL, GCL, and IPL were measured as one complex (NFL/IPL/GCL). For the NFL/IPL/GCL complex, there was a significant age effect in WSB mice (Fig. C) and an age-by-sex effect in NZO mice (Fig. C). We found significant age-associated reductions in the inner nuclear layer (INL) and outer nuclear layer (ONL) in NZO and B6 mice (Fig. D-E). WSB mice exhibited age- and sex-by-age driven reductions in the INL and ONL, respectively (Fig. D-E). In brief, 1) B6 mice lost 6% total retinal thickness by 18 M, which was due to a 19% loss of INL thickness and 12% loss of ONL thickness; 2) NZO mice also lost 6% total retinal thickness by 18 M, which was due to a 21% loss of INL thickness and 17% loss of ONL thickness; 3) WSB mice lost 22% total retinal thickness by 18 M, which was due to a 19% loss of INL thickness and 49% loss of ONL thickness. As sex was associated with such subtle effects, and no significant effects in the aging signatures within the multi-omics analyses were observed, we combined sexes for future analyses. Altogether, these data suggest that aging mice exhibit neural retinal thinning regardless of genetic context but that the specific age- and sex-affected disruptions are strongly influenced by genetics. Moreover, these perturbations were strongest in WSB mice, which may disproportionally lose photoreceptors within the ONL. WSB mice exhibit abnormal oscillatory potentials and exacerbated aging-associated reductions in photoreceptor function Given the extensive thinning of the ONL in WSB mice, we next sought to evaluate photoreceptor function with scotopic and photopic ERGs (dark-adapted and light-adapted, respectively) at 4 and 18 M (Fig. ). As was predicted by our omics analyses, age reduced these amplitudes across strains (Fig. A-I). At a luminance that elicited maximal ERG amplitudes (10 cd.s/m 2 ), WSB mice exhibited a 52% decline in scotopic a-wave amplitudes, while B6 and NZO mice had 15% and 26% reductions, respectively (Fig. B-C). These data suggest drastic rod dysfunction in aged WSB mice. Scotopic b-waves were reduced by 34% in WSB mice, 17% in B6 mice, and 27% in NZO mice (Fig. D-E). Both NZO and WSB mice had an age-associated decrease in scotopic oscillatory potentials (OPs; 24% and 37%, respectively), which was not observed in B6 mice (Fig. F-G). Strikingly, WSB mice appeared to exhibit nearly absent OPs even at young ages. In contrast, photopic b-wave amplitudes were similar across strains in young mice, and WSB and NZO mice exhibited mild age-associated reductions in photopic b-wave amplitudes (20% and 25%, respectively; Fig. H-I). These data suggest relatively conserved cone function across strains with age. Given drastic ONL thinning and rod photoreceptor dysfunction in WSB mice, we next sought to quantify photoreceptor cell loss more precisely, as OCT images may not capture changes outside of the central retina due to technical limitations. Thus, we next employed routine hematoxylin and eosin (H&E) staining on retinal sections at 4, 12, and 18 M in B6, NZO, and WSB mice (Fig. A). In accordance with OCT measurements, the number of ONL cells decreased with age across strains (Fig. B). B6 retinas had a 12–14% decrease in ONL cells by 12 M, which declined by an additional 10–14% by 18 M; netting a 23–25% overall loss. NZO retinas had a 17–15% loss of ONL cells by 12 M, which did not decline further by 18 M. WSB retinas had an 18–20% loss of ONL cells by 12 M, but lost an additional 24–50% of ONL cells by 18 M; netting a 37–60% overall loss. This loss was most drastic in the center retina. In fact, we observed nearly complete loss of ONL cells in the central retina of 18 M WSB animals alongside instances of abnormal retinal structures (Supplemental Fig. ). It is important to note that in many species, the retina thins and elongates with age, which could be falsely interpreted as neural loss . To address this possibility, we measured retinal lengths across our histological preparations and found that there was indeed age-related retinal elongation in B6 (9%) and NZO (14%) mice by 18 M with no observed change in 4 vs 18 M WSB mice (Fig. C-D). To add stringency to our ONL measurements, we normalized each measurement to the change in retinal length in each strain, and the overall ANOVA results were nearly identical (Fig. E). Our ERG analysis indicated that it was likely that rod photoreceptors were specifically degenerating in WSB mice with age. To confirm this, we performed immunohistochemistry using markers for rods (RHO, Supplemental Fig. A) and a subset of cones (opsin 1, long-wave (OPN1LW), Supplemental Fig. B). First, the width of the IS/OS was significantly altered by location and age across all three strains (Supplemental Fig. C). B6 IS/OS thickness declined by 27% in the center retina between 12 and 18 M and declined 22% in the periphery between 4 and 12 M. NZO IS/OS thickness declined by 26% in the periphery between 4 and 12 M. WSB IS/OS thickness declined by 47% in the central retina and 30% in the middle retina by 18 M. This was strongly mirrored by overall retinal area positive for RHO (Supplemental Fig. D). B6 RHO + area declined by 27% in the center retina between 12 and 18 M. NZO RHO + area declined by 41% in the peripheral retina by 18 M. WSB RHO + area declined by 47% in the central retina and 42% in the middle retina by 18 M. However, we found that there were no changes in OPN1LW + cell numbers with age in B6 and NZO mice (Supplemental Fig. B,E). There was a slight location-by-age effect in WSB mice with no significant changes in OPN1LW + cones for any age comparison identified by post-hoc analysis (Supplemental Fig. B,E). These assays indicate that aging reduces rod photoreceptor function across all three mouse strains, and WSB mice exhibit early retinal dysfunction with severe rod dysfunction and drastic loss of central ONL cells with age. Altogether, these considerations strongly indicate WSB animals exhibit a profound loss of photoreceptors. To begin to understand the mechanisms driving the early and progressive photoreceptor dysfunction in WSB mice, we compared proteome differences between 4 M WSB with other pigmented strains. This approach revealed that 258 proteins were altered in WSB mice relative to the other strains and were associated in pathways involving essential mitochondrial metabolism, including “Generation of precursor metabolites and energy”, “Purine ribonucleotide metabolic process”, and reductions to proteins associated with the BBSome (Supplemental Fig. A-B, Supplemental Table 3). The BBSome is critical for photoreceptor outer segment function, and loss of BBSome proteins have been shown to cause photoreceptor degeneration and dysfunction . Our data show reduced expression of WSB BBSome proteins relative to other strains even at a young age. Furthermore, Retinol Dehydrogenase 8 (the major outer segment enzyme responsible for converting all- trans -retinal to all- trans -retinol and preventing accumulation of toxic by-products) was significantly lower in WSB retinas . This is supported by our IHC analyses indicating that WSB mice exhibit much thinner IS/OS layers relative to B6 or NZO mice even at 4 M (Supplemental Fig. C). Thus, early WSB photoreceptor dysfunction may be due in part to cell-intrinsic deficits in outer segment functions which culminate in photoreceptor loss. NZO mice display profound loss of RGC function with age and exhibit retinal microvascular dysfunction Our transcriptomic data suggested NZO mice lose RGCs with age (Fig. D-F). Thus, we utilized PERG to probe for gross changes in RGC potentials . At 4 M, NZO mice exhibited blunted PERG amplitudes (Fig. A-B), which declined by 41% to near-noise levels by 18 M—indicating profound RGC dysfunction. Interestingly, B6 mice displayed a 28% reduction in PERG amplitudes by 18 M, while WSB mice were resilient to age-related PERG decline. To determine whether RGCs are lost (versus less functional), we performed whole-mount immunohistochemistry for the commonly utilized marker of RGCs: RNA Binding Protein, MRNA Processing Factor (RBPMS) to quantify this neuronal population across the entire retina. We found that 12 and 18 M NZO mice had a 35–56% decrease in surviving RBPMS + RGCs relative to 4 M NZO animals, compared to 5–13% and 25–26% for B6 and WSB, respectively (Fig. C-D). Collectively, these data indicate NZO mice display RGC dysfunction and loss. In addition to apparent RGC loss, aged NZO mice have been shown to exhibit most features of metabolic syndrome . It is well established that retinal microvascular dysfunction coincides with metabolic stress—including diabetes and hypertension . Therefore, we next evaluated retinal vascular changes using fluorescein angiography in vivo in B6, NZO, and WSB mice. We found differences in the mean intensity of fluorescein dye across strains (Fig. A-C). Interestingly, age significantly reduced the mean intensity across all three strains, with a dramatic effect observed in NZO mice (Fig. A-C). B6 and WSB mice had a 70% and 60% reduction in mean fluorescence intensity of fluorescein, respectively, while NZO mice had an 86% reduction in intensity. We also observed several instances of microaneurysms and hemorrhages in NZO animals by fundus and angiography (Supplemental Fig. A-E). Furthermore, we detected a few instances of Prussian Blue + deposits, indicating subtle intraretinal vascular leakage in aging NZO mice (Supplemental Fig. F). Collectively, these data suggested that NZO animals exhibit reduced retinal perfusion with age or reduced capillary density. To further investigate retinal vascular deficits, we utilized a FITC-dextran permeability assay to probe the integrity of the blood-retina barrier and found that NZO exhibited significantly higher instances of 70 kDa FITC-dextran leakage. We noted increased areas of nonperfusion in NZO relative to either B6 or WSB mice at 12 M of age (Fig. D-E). To determine if these areas of nonperfusion were associated with capillary loss, we isolated and stained retinal vascular networks from 12 M animals. We found there was significantly more acellular retinal capillaries in 12 M NZO retinas (41 acellular capillaries/mm 2 ) compared to B6 (21/mm 2 ) and WSB (18/mm 2 ) retinas (Fig. F-G). Altogether, these data suggest NZO mice exhibit DR-relevant microvascular dysfunction and capillary loss by 12 M of age, which may influence RGC loss and progression of neurodegeneration. To identify potential mechanisms driving increased susceptibility to RGC loss in NZO mice, we compared proteome differences between NZO mice and the other pigmented strains at 4 M. This approach revealed that 97 proteins were altered in NZO mice relative to the other strains, which were associated with antioxidant pathways such as “Glutathione metabolic process” including the protein Glutathione Synthetase (GSS), “Cellular detoxification” including the enzyme Glyoxalase 1 (GLO1), and changes to proteins associated with the NADH oxidoreductase complex (Supplemental Fig. A-B, Supplemental Table 4). These data indicate NZO retinas may be particularly susceptible to oxidative stress associated with metabolic dysfunction due to reduced protective capacities. Aging is an important risk factor for numerous vision-threatening diseases , yet, little is known about the interaction of age, sex, and genetic background. To better understand how aging across diverse genetic contexts may alter retinal function, we profiled the retinal transcriptome and proteome of nine genetically diverse mouse strains at the ages of 4, 12 and ≥ 18 M (Fig. A). Several strains had natural attrition , reducing the number of aged mice of particular strains available for tissue collection and data analysis (see Supplemental Table 1 for full details). Our initial analysis of the resulting transcriptomic and proteomic data included sex, strain, and age as covariates. After filtering the multi-omics data, we analyzed 22,928 genes and 4,854 proteins across these groups. Our transcriptomic analyses included many cell-type specific transcripts such as: Vwf (endothelial cells) , Rho (photoreceptors) , Gfap (astrocytes) , Glul (Müller glia) , Calb1 (horizontal cells) , Calb2 (amacrine cells) , Tmem119 (microglia) , and Rbpms (retinal ganglion cells). The detected proteins included uniquely expressed proteins in cell-types including rhodopsin (RHO, photoreceptors), glial fibrillary acidic protein (GFAP, astrocytes) and glutamine synthetase (GLUL, müller glia), Von Willebrand factor (VWF, endothelial cells), calbindin (CALB1, interneurons), and calretinin (CALB2, amacrine cells). First, we visualized transcriptomic and proteomic data using principal component analyses (PCA). This revealed that strain was the largest contributor to variation in the transcriptomic data (Fig. B) with more subtle patterns observed in the proteomic data (Fig. C). The less pronounced strain signature in proteomic data may be due to the more limited numbers of proteins analyzed but may also be due in part to buffering at the protein level . We next evaluated if the predictive strength of the transcriptome in determining protein abundance differed across strains and found that there were similar small (R = 0.19–0.28) but significant ( p < 0.0001) correlations of protein abundance and RNA expression regardless of strain or age (Supplemental Fig. ). As examples of the importance of strain variation in selected pathways regardless of age critical for retinal health, genes associated with the GO term “Regulation of complement”, including Cfh (Fig. D-E) or genes associated with the GO term “Detection of visible light”, including Crb1, displayed substantial strain-dependent variation (Fig. F-G). Proteins within the “Metabolic process” GO term, including acetyl-CoA acetyltransferase 1 (ACAT1), also displayed strain-dependent expression patterns (Fig. H-I). Thus, genetic context was a major source of variation in retinal molecular signatures. To define an overall retinal aging signature regardless of genetic context, we tested for differential gene and protein expression using age as the predictive variable across genetic backgrounds. We found that 4929 genes and 271 proteins were differentially expressed with aging across genetically diverse mice (Fig. A-B). Of these 271 differentially expressed proteins (DEPs), 131 were also significantly modulated at the gene level (Fig. C). Surprisingly, sex did not have a significant impact on the aging signatures across strains. Enrichment analyses revealed significant alteration of GO terms involving immune processes, photoreceptors, metabolism, and chromatin organization at both the gene and protein level (Fig. D-E). Differentially expressed photoreceptor genes and proteins decreased with age, while the differentially expressed genes (DEGs) and DEPs associated with immune processes largely increased with age (Supplemental Table 2). Next, we sought to examine how this aging signature was modified by genetic context. PCA visualization of the aging signature genes and proteins suggested differences in gene and protein expression between the strains of mice (Fig. A-B). For instance, genes associated with the GO term “Regulation of histone methylation”, including Mthfr (Fig. C-D), genes associated with the GO term “Antigen processing and peptide presentation”, including B2m (Fig. E-F) , and genes associated with the GO term “Retina homeostasis”, including Rho (Fig. G-H) all displayed substantial strain- and age-dependent variation. The GO pathway “Retina homeostasis” and Rho expression seemed to decrease with age in the albino strains and WSB mice more rapidly than the other 5 strains. Overall, these observed differences in the aging signature suggest genetic context dictates susceptibility to aging-associated retinal diseases. Common retinal neurodegenerative diseases involve RGC and/or photoreceptor loss . Single nucleus RNA-sequencing datasets have revealed several marker genes for specific retinal neurons . Therefore, we probed for changes in these neuronal populations by visualizing marker gene differences. We first profiled changes in photoreceptor-specific gene expression across strains using PCA (Fig. A). We found that WSB, NOD, and AJ mice showed age-associated shifts away from the other strains (Fig. A). In fact, these three strains exhibited reduced photoreceptor gene expression relative to the other strains at 18 M (Fig. B). In accordance with our transcriptomics data, AJ mice are known to develop age-associated loss of photoreceptors as shown by histological and functional analyses, while albino NOD mice can develop type 1 diabetes concomitant with neural and vascular damage by electron microscopy . In our analysis, WSB mice exhibited age-dependent differences in photoreceptor marker gene expression (Fig. C). WSB mice do not carry mutations in known retinal degeneration genes , and to date, there are no published reports of WSB photoreceptor loss. Thus, WSB mice may serve as a novel model of photoreceptor degeneration. In addition to photoreceptors, we profiled changes in RGC marker genes across strains. NZO, CAST, and BALBc mice exhibited a slight leftward shift in the PCA plot for RGC genes (Fig. D), and heatmap visualization of RGC marker genes revealed low expression of these genes in aged NZO, CAST and BALBc mice relative to other strains (Fig. E). A previous report suggested CAST mice develop significantly fewer RGCs (and thus have smaller optic nerves) than other strains, which may account for the reduced expression of RGC genes in this strain . Like many albino strains, BALBc mice have been shown to develop age-related retinal dysfunction exacerbated by vivarium light cycles by many measures including optical coherence tomography (OCT), electroretinography (ERG), histology and fundus imaging . However, there have been no previous reports of retinal phenotypes in NZO mice. To better understand how aging influences RGC gene expression in NZO mice, we visualized gene expression changes across age in NZO mice and found striking reductions in these genes (Fig. F). These data suggest NZO mice may be especially susceptible to RGC loss with age. Collectively, our multi-omics analysis suggested genetic context significantly contributes to retinal aging signatures. We created a user-friendly web application to increase data accessibility and foster hypothesis development and refinement, which can be accessed at: https://thejacksonlaboratory.shinyapps.io/Howell_AgingRetinaOmics/ . For example, one could search for an individual gene of interest (e.g., “ Sigmar1 ”) and visualize the RNA and/or protein expression (Supplemental Fig. A-B) across all strains. Alternatively, a set of genes/proteins can be visualized in a PCA style plot (Supplemental Fig. C). Our analyses suggested shared changes to photoreceptor markers across all nine strains, along with previously unreported age-dependent changes to retinal neurons in NZO and WSB mice. We therefore sought to verify these predictions using a battery of clinically-relevant in vivo ophthalmological exams at the ages of 4, 8, 12 and/or ≥ 18 M in NZO, WSB, and B6 mice of both sexes (Fig. A). Gross changes to the anterior segment of NZO, WSB, and B6 eyes were not detected using slit lamp imaging (Fig. B). In contrast, fundus imaging revealed age-associated increases of fundus abnormalities in WSB, NZO, and B6 mice (Fig. C-D). Fundus abnormalities in B6 mice included abnormal retinal spots (Fig. C-D). In NZO mice, these abnormalities included spots resembling cotton wool spots and exudates, which are commonly observed in human DR and other retinal vascular diseases . The presence of these phenotypes peaked by 12 M, however, what appeared to be an epiretinal membrane was uniquely observed in aged NZO eyes (7 of 38 NZO 18 M eyes; Supplemental Fig. ). WSB mice exhibited fundus abnormalities as early as 4 M, which became increasingly severe with age (Fig. C-D). These abnormalities included potential pigment disruption, retinal spots, and waxy optic disc pallor. Altogether, these data indicate age-related changes to DR-relevant fundus phenotypes in NZO mice and retinal degeneration-relevant phenotypes in WSB mice. To investigate whether fundus abnormalities corresponded with gross retinal degeneration, OCT was employed to measure retinal layer thicknesses across age in each strain (Fig. A). As total retinal thickness differed by strain in young animals (Fig. A-B), we determined the effect of age within each strain separately. Age reduced total retinal thickness across all three strains (Fig. B). There was a subtle but significant interaction effect between sex and age in NZO mice (Fig. B). The nerve fiber layer (NFL), ganglion cell layer (GCL), and inner plexiform layer (IPL) are largely composed of RGC axons, cell bodies, and dendritic processes/synapses, respectively. These layers were difficult to discern via OCT due to technical limitations, therefore, the NFL, GCL, and IPL were measured as one complex (NFL/IPL/GCL). For the NFL/IPL/GCL complex, there was a significant age effect in WSB mice (Fig. C) and an age-by-sex effect in NZO mice (Fig. C). We found significant age-associated reductions in the inner nuclear layer (INL) and outer nuclear layer (ONL) in NZO and B6 mice (Fig. D-E). WSB mice exhibited age- and sex-by-age driven reductions in the INL and ONL, respectively (Fig. D-E). In brief, 1) B6 mice lost 6% total retinal thickness by 18 M, which was due to a 19% loss of INL thickness and 12% loss of ONL thickness; 2) NZO mice also lost 6% total retinal thickness by 18 M, which was due to a 21% loss of INL thickness and 17% loss of ONL thickness; 3) WSB mice lost 22% total retinal thickness by 18 M, which was due to a 19% loss of INL thickness and 49% loss of ONL thickness. As sex was associated with such subtle effects, and no significant effects in the aging signatures within the multi-omics analyses were observed, we combined sexes for future analyses. Altogether, these data suggest that aging mice exhibit neural retinal thinning regardless of genetic context but that the specific age- and sex-affected disruptions are strongly influenced by genetics. Moreover, these perturbations were strongest in WSB mice, which may disproportionally lose photoreceptors within the ONL. Given the extensive thinning of the ONL in WSB mice, we next sought to evaluate photoreceptor function with scotopic and photopic ERGs (dark-adapted and light-adapted, respectively) at 4 and 18 M (Fig. ). As was predicted by our omics analyses, age reduced these amplitudes across strains (Fig. A-I). At a luminance that elicited maximal ERG amplitudes (10 cd.s/m 2 ), WSB mice exhibited a 52% decline in scotopic a-wave amplitudes, while B6 and NZO mice had 15% and 26% reductions, respectively (Fig. B-C). These data suggest drastic rod dysfunction in aged WSB mice. Scotopic b-waves were reduced by 34% in WSB mice, 17% in B6 mice, and 27% in NZO mice (Fig. D-E). Both NZO and WSB mice had an age-associated decrease in scotopic oscillatory potentials (OPs; 24% and 37%, respectively), which was not observed in B6 mice (Fig. F-G). Strikingly, WSB mice appeared to exhibit nearly absent OPs even at young ages. In contrast, photopic b-wave amplitudes were similar across strains in young mice, and WSB and NZO mice exhibited mild age-associated reductions in photopic b-wave amplitudes (20% and 25%, respectively; Fig. H-I). These data suggest relatively conserved cone function across strains with age. Given drastic ONL thinning and rod photoreceptor dysfunction in WSB mice, we next sought to quantify photoreceptor cell loss more precisely, as OCT images may not capture changes outside of the central retina due to technical limitations. Thus, we next employed routine hematoxylin and eosin (H&E) staining on retinal sections at 4, 12, and 18 M in B6, NZO, and WSB mice (Fig. A). In accordance with OCT measurements, the number of ONL cells decreased with age across strains (Fig. B). B6 retinas had a 12–14% decrease in ONL cells by 12 M, which declined by an additional 10–14% by 18 M; netting a 23–25% overall loss. NZO retinas had a 17–15% loss of ONL cells by 12 M, which did not decline further by 18 M. WSB retinas had an 18–20% loss of ONL cells by 12 M, but lost an additional 24–50% of ONL cells by 18 M; netting a 37–60% overall loss. This loss was most drastic in the center retina. In fact, we observed nearly complete loss of ONL cells in the central retina of 18 M WSB animals alongside instances of abnormal retinal structures (Supplemental Fig. ). It is important to note that in many species, the retina thins and elongates with age, which could be falsely interpreted as neural loss . To address this possibility, we measured retinal lengths across our histological preparations and found that there was indeed age-related retinal elongation in B6 (9%) and NZO (14%) mice by 18 M with no observed change in 4 vs 18 M WSB mice (Fig. C-D). To add stringency to our ONL measurements, we normalized each measurement to the change in retinal length in each strain, and the overall ANOVA results were nearly identical (Fig. E). Our ERG analysis indicated that it was likely that rod photoreceptors were specifically degenerating in WSB mice with age. To confirm this, we performed immunohistochemistry using markers for rods (RHO, Supplemental Fig. A) and a subset of cones (opsin 1, long-wave (OPN1LW), Supplemental Fig. B). First, the width of the IS/OS was significantly altered by location and age across all three strains (Supplemental Fig. C). B6 IS/OS thickness declined by 27% in the center retina between 12 and 18 M and declined 22% in the periphery between 4 and 12 M. NZO IS/OS thickness declined by 26% in the periphery between 4 and 12 M. WSB IS/OS thickness declined by 47% in the central retina and 30% in the middle retina by 18 M. This was strongly mirrored by overall retinal area positive for RHO (Supplemental Fig. D). B6 RHO + area declined by 27% in the center retina between 12 and 18 M. NZO RHO + area declined by 41% in the peripheral retina by 18 M. WSB RHO + area declined by 47% in the central retina and 42% in the middle retina by 18 M. However, we found that there were no changes in OPN1LW + cell numbers with age in B6 and NZO mice (Supplemental Fig. B,E). There was a slight location-by-age effect in WSB mice with no significant changes in OPN1LW + cones for any age comparison identified by post-hoc analysis (Supplemental Fig. B,E). These assays indicate that aging reduces rod photoreceptor function across all three mouse strains, and WSB mice exhibit early retinal dysfunction with severe rod dysfunction and drastic loss of central ONL cells with age. Altogether, these considerations strongly indicate WSB animals exhibit a profound loss of photoreceptors. To begin to understand the mechanisms driving the early and progressive photoreceptor dysfunction in WSB mice, we compared proteome differences between 4 M WSB with other pigmented strains. This approach revealed that 258 proteins were altered in WSB mice relative to the other strains and were associated in pathways involving essential mitochondrial metabolism, including “Generation of precursor metabolites and energy”, “Purine ribonucleotide metabolic process”, and reductions to proteins associated with the BBSome (Supplemental Fig. A-B, Supplemental Table 3). The BBSome is critical for photoreceptor outer segment function, and loss of BBSome proteins have been shown to cause photoreceptor degeneration and dysfunction . Our data show reduced expression of WSB BBSome proteins relative to other strains even at a young age. Furthermore, Retinol Dehydrogenase 8 (the major outer segment enzyme responsible for converting all- trans -retinal to all- trans -retinol and preventing accumulation of toxic by-products) was significantly lower in WSB retinas . This is supported by our IHC analyses indicating that WSB mice exhibit much thinner IS/OS layers relative to B6 or NZO mice even at 4 M (Supplemental Fig. C). Thus, early WSB photoreceptor dysfunction may be due in part to cell-intrinsic deficits in outer segment functions which culminate in photoreceptor loss. Our transcriptomic data suggested NZO mice lose RGCs with age (Fig. D-F). Thus, we utilized PERG to probe for gross changes in RGC potentials . At 4 M, NZO mice exhibited blunted PERG amplitudes (Fig. A-B), which declined by 41% to near-noise levels by 18 M—indicating profound RGC dysfunction. Interestingly, B6 mice displayed a 28% reduction in PERG amplitudes by 18 M, while WSB mice were resilient to age-related PERG decline. To determine whether RGCs are lost (versus less functional), we performed whole-mount immunohistochemistry for the commonly utilized marker of RGCs: RNA Binding Protein, MRNA Processing Factor (RBPMS) to quantify this neuronal population across the entire retina. We found that 12 and 18 M NZO mice had a 35–56% decrease in surviving RBPMS + RGCs relative to 4 M NZO animals, compared to 5–13% and 25–26% for B6 and WSB, respectively (Fig. C-D). Collectively, these data indicate NZO mice display RGC dysfunction and loss. In addition to apparent RGC loss, aged NZO mice have been shown to exhibit most features of metabolic syndrome . It is well established that retinal microvascular dysfunction coincides with metabolic stress—including diabetes and hypertension . Therefore, we next evaluated retinal vascular changes using fluorescein angiography in vivo in B6, NZO, and WSB mice. We found differences in the mean intensity of fluorescein dye across strains (Fig. A-C). Interestingly, age significantly reduced the mean intensity across all three strains, with a dramatic effect observed in NZO mice (Fig. A-C). B6 and WSB mice had a 70% and 60% reduction in mean fluorescence intensity of fluorescein, respectively, while NZO mice had an 86% reduction in intensity. We also observed several instances of microaneurysms and hemorrhages in NZO animals by fundus and angiography (Supplemental Fig. A-E). Furthermore, we detected a few instances of Prussian Blue + deposits, indicating subtle intraretinal vascular leakage in aging NZO mice (Supplemental Fig. F). Collectively, these data suggested that NZO animals exhibit reduced retinal perfusion with age or reduced capillary density. To further investigate retinal vascular deficits, we utilized a FITC-dextran permeability assay to probe the integrity of the blood-retina barrier and found that NZO exhibited significantly higher instances of 70 kDa FITC-dextran leakage. We noted increased areas of nonperfusion in NZO relative to either B6 or WSB mice at 12 M of age (Fig. D-E). To determine if these areas of nonperfusion were associated with capillary loss, we isolated and stained retinal vascular networks from 12 M animals. We found there was significantly more acellular retinal capillaries in 12 M NZO retinas (41 acellular capillaries/mm 2 ) compared to B6 (21/mm 2 ) and WSB (18/mm 2 ) retinas (Fig. F-G). Altogether, these data suggest NZO mice exhibit DR-relevant microvascular dysfunction and capillary loss by 12 M of age, which may influence RGC loss and progression of neurodegeneration. To identify potential mechanisms driving increased susceptibility to RGC loss in NZO mice, we compared proteome differences between NZO mice and the other pigmented strains at 4 M. This approach revealed that 97 proteins were altered in NZO mice relative to the other strains, which were associated with antioxidant pathways such as “Glutathione metabolic process” including the protein Glutathione Synthetase (GSS), “Cellular detoxification” including the enzyme Glyoxalase 1 (GLO1), and changes to proteins associated with the NADH oxidoreductase complex (Supplemental Fig. A-B, Supplemental Table 4). These data indicate NZO retinas may be particularly susceptible to oxidative stress associated with metabolic dysfunction due to reduced protective capacities. Vision-threatening retinal degenerative disorders are becoming increasingly prominent as the aging population increases . Age is a critical risk factor in the development of retinal neurodegeneration. Studies in both humans and mice have demonstrated that genetics influences susceptibility to age-related retinal degeneration. However, few studies have investigated molecular changes occurring in the retina with age, or how these changes may be modified by genetic context. Elucidating these molecular changes will be critical in improving our understanding of retinal aging and disease, potentially informing novel models and therapeutic targets. Furthermore, these data can identify mechanisms that are critical for neurodegenerative diseases that impact the brain and are an essential foundation to determine if retinal changes can be used as a biomarker for brain neurodegeneration. In the present study, we investigate retinal transcriptomic and proteomic changes with age across nine genetically diverse mouse strains. This diverse population, with the inclusion of both sexes across young (4 M), middle-aged (12 M), and aged (18 M +) mice, is intended to improve the relevance of these data to the heterogeneous human population. Not unlike the human population, the nine strains of mice utilized have varying levels of natural attrition due to differences in lifespan . This did result in our samples ranging from 18–25 M for the 18 M + category with variance between and within strains. However, it is important to note that the underlying predicted phenotypes in the omics data were reevaluated in separate cohorts for the subsequent investigations. Our multi-omic investigation revealed common aging signatures across strains, which included reduced photoreceptor and visual processing functions concomitant with activation of immune signaling pathways at both the gene and protein level. In concordance with this, we found subtle but significant loss of ONL cells and a decline in scotopic ERG a- and b-waves across strains measured, including the widely utilized B6 mouse. These data indicate subtle age-related loss of rod photoreceptors and INL cells, respectively. Our results align with the reduced ERG function and photoreceptor loss observed in human patients with age . These data suggest that even in the absence of frank neurodegeneration, there is increasing neural dysfunction with age in the murine retina. It is unclear whether the activation of immune response pathways is in response to neural dysfunction, or if immune activation is directly responsible for the observed dysfunction. Proper balance between oxidative stress and antioxidant systems are thought to be critical for retinal homeostasis . Pathological aging is thought to disrupt the balance of these systems, yet our analyses of these data did not reveal a common oxidative stress response with aging in either the proteomic or transcriptional analysis, however, it is possible that increasing oxidative stress may alter protein post-translational modifications thus altering functions and would not have been detected by our assays. Future studies will be required to probe these possibilities specifically. Strain was the strongest contributor to variation within the transcriptomic and proteomic data. In our analyses, we highlight that genetic context dictated the expression patterns of genes and proteins associated with a conserved aging signature. For example, genes associated with conserved aging-induced “Antigen processing and peptide presentation” pathway displayed substantial strain-dependent differences, with PWK mice exhibiting less age-dependent changes in these genes. The identified common aging signature across mice suggested altered visual function with age. However, some strains exhibited greater changes in these pathways, including those associated with photoreceptors. While previous work has linked aging to dysregulation of retinal neurons, overt loss of photoreceptors and RGCs is intimately associated with many human diseases . To infer changes in photoreceptor and RGC populations with age, we visualized the expression of cell-type specific marker genes across ages. It is important to note that as we used whole retina tissues for our multi-omic analyses, the most common cell-types will be responsible for much of the detected RNA and proteins – e.g., photoreceptors. As such, many of the cell type-specific markers were not detected in the proteomics analysis of 4885 proteins. This is almost certainly due to a large dynamic range of protein abundances in the retina which would result in low abundance proteins being difficult to measure in the presence of very high abundance proteins . However, we were still able to detect some cell-type specific proteins in the proteomics dataset including ones that denote glia including GFAP and GLUL, endothelial cells such as VWF, and for non-photoreceptor neurons (examples proteins: CALB1 and CALB2, and choline acetyltransferase). In future work, one could deplete photoreceptors before proceeding with proteomics analyses to enhance discovery of less abundant retinal proteins. Importantly, we were able to detect significant differences in many marker genes by RNA-seq even for cell-types like RGCs, various glia, and endothelial cells (example marker genes: Rbpms , Gfap, Glul, Tmem119, Aif1, Vwf, Pecam1 ) which compromise ~ 2–3% of total retinal cells suggesting the depth of sequencing per sample was robust enough to overcome these limitations. Our RNA-seq analyses of these marker genes support previous reports that AJ, and BALBc mice exhibit age-related neuronal loss . Our transcriptomic analyses identified age-related photoreceptor loss in WSB mice. Indeed, WSB mice exhibited hyperpigmented clumps on fundus exam, which was strikingly reminiscent of human RP phenotypes . Consistent with human RP, WSB mice had drastic photoreceptor loss with age. This loss appeared to be largely rod-specific, as scotopic ERG a-waves exhibited significant decline with age, and consequential inner retinal signaling was substantially blunted (as indicated by lower b-wave and nearly non-existent OPs). Patients with RP can exhibit severely reduced or absent OPs . Cone photoreceptors appeared to be subtly lost with age (as evident by a more subtle loss of photopic ERG amplitudes), and RGCs were relatively spared from this degeneration (as indicated by comparable RBPMS + cell loss to B6 mice along with intact PERG responses). Interestingly, WSB photoreceptor loss was most severe at the central retina. This contrasts with RP disease progression—where peripheral rods are lost first—and is more consistent with patterns of photoreceptor loss in observed in human AMD. The central retina of the mouse shares some similarities to the human macula, including a higher density of photoreceptors, a thinner Bruch’s membrane, and fewer retinal pigment epithelium cells per photoreceptor . Thus, WSB mice model phenotypes relevant to multiple retinal degenerative diseases. To investigate mechanisms relevant to photoreceptor degeneration in WSB mice, we compared the retinal proteome of 4 M WSB mice to other pigmented strains. These results indicated that WSB mice exhibit DEPs involved in essential photoreceptor functions—including mitochondrial and retinol metabolism, and ciliary proteins within the BBSome . These analyses predict the mechanisms that initiate the RP-like phenotype are intrinsic to photoreceptors. Importantly, WSB mice are not known to carry any rd mutations . Thus, WSB mice are likely to harbor novel mutations driving photoreceptor loss, which could be uncovered using genetic mapping studies. Altogether, these data suggest that WSB mice may be a unique model with which one could study mechanisms of age-related retinal degeneration relevant to human AMD and RP. Our transcriptomic analyses also predicted an age-related loss of RGCs in NZO mice. Fundus examination of NZO mice revealed age-associated incidences of spots akin to cotton wool spots, exudates, and what appeared to be an epiretinal membrane in aged mice. Cotton wool spots and exudates are commonly found in many retinal vascular diseases, including DR . These abnormalities are thought to arise from microvascular dysfunction and may indicate areas of local ischemia and blood-retina barrier breakdown . Future work will be required to better understand whether the epiretinal membrane is similar to that observed in human patients. The presence of these abnormalities has not been previously reported in spontaneous mouse models of DR but cotton wool spots have been observed in a primate model of DR after chronic type 2 diabetes . In fact, widely used mouse models of DR fail to recapitulate the normal progression of disease pathology, and most models are limited to phenotypes associated with early stages of DR . In DR, it is thought that initial hyperglycemia and metabolic dysfunction drives microvascular damage, promoting vascular compromise, eventually resulting in an increasingly hypoxic environment, promoting deleterious neovascularization . The precise timing of neuronal loss amidst the microvascular dysfunction is not entirely clear in human patients, however, there are reports that there is loss of INL cells and RGCs even prior to overt DR . NZO mice are a polygenic model of metabolic syndrome exhibiting profound obesity, hypertension, hyperlipidemia, and insulin resistance, with approximately 50% of male mice becoming overtly hyperglycemic with age . We observed no clear sex-dependent differences in the progression of NZO-associated retinal phenotypes, suggesting these changes may be due to the milieu of metabolic impairments NZO animals develop beyond hyperglycemia alone. To probe the molecular causes of retinopathy in NZO mice, we investigated early retinal proteomic differences in NZO retinas. The retinal proteome of 4 M NZO showed reduced expression of glutathione metabolic process and cellular detoxification pathways relative to other strains. In fact, the GSS protein (which is essential to produce glutathione) was significantly downregulated in NZO retinas relative to other strains. Glutathione levels are significantly reduced in many retinal diseases . Furthermore, levels of the glutathione-dependent enzyme GLO1 were significantly lower in 4 M NZO retinas. GLO1 is a major detoxification enzyme for methylglyoxal, a precursor of advanced glycation end products (AGEs) . Methylglyoxal, AGEs, and GLO1 have been linked to DR progression in patients and in rodent models . As these analyses utilized whole retinal tissue it does not likely reflect changes in RGCs specifically, but rather reflects the retinal environment at large. Altogether, reductions in protective pathways likely increase susceptibility to oxidative stress-induced damage, which may be mediated by AGEs resulting from systemic metabolic dysfunction . Given not all cell-types were well represented in the proteomics dataset (e.g., microglia) single cell-based studies are underway to understand both RGC intrinsic changes in aging NZO mice and critically how immune cells influence the progression of retinopathy. In accordance with the progression of human DR and retinal vascular disease, NZO animals displayed significant microvascular dysfunction relative to either WSB or B6 mice at 12 M by multiple measures, including fluorescein angiography, FITC-Dextran leakage, and presence of acellular capillaries. NZO mice also exhibited age-associated loss of RBPMS + RGCs and INL thickness, which manifested as reduced visual function as measured by both PERG and ERG. Collectively, these data suggest NZO mice better model the progression of human retinal vascular disease and DR than existing models, which includes the stepwise development of vascular insufficiencies and neuronal dysfunction. In totality, our efforts: 1) generated a publicly available multi-omics database for retinal aging research across diverse genetic contexts to facilitate hypothesis development and refinement, 2) identified a common retinal aging signature across mice suggestive of reduced photoreceptor function, 3) determined that genetic context is a major driver of the molecular changes associated with retinal aging, 4) characterized WSB mice as a model for overt age-related and regionalized rod-specific photoreceptor loss, and 5) identified NZO mice as a novel mouse model of retinal vascular disease and RGC loss with striking similarities to human disease. Our study provides a valuable multi-omics resource for those investigating the aging retina, mechanisms of neurodegeneration, or the retina as a biomarker for changes within the brain. Our inclusion of the eight founder strains of the CC and DO diverse mouse populations provides important context for other phenotype- and expression-based analyses conducted with these populations. From these data, we identified and validated two novel models of age-related retinal neurodegeneration. Supplementary Material 1. Supplementary Material 2. Supplementary Material 3. Supplementary Material 4. Supplementary Material 5. Supplementary Material 6. Supplementary Material 7. Supplementary Material 8. Supplementary Material 9. Supplementary Material 10. Supplementary Material 11. Supplementary Material 12. Supplementary Material 13. |
Artificial intelligence and machine learning in ophthalmology: A review | c49cc4f0-a7ca-4b79-afad-afa9f33e7bf7 | 10155540 | Ophthalmology[mh] | Diabetic retinopathy Screening for diabetic retinopathy (DR) is essential as it facilitates early detection and treatment, thereby preventing vision loss. This is relevant in Canada as 3.7 million people have diabetic retinopathy; the incidence of DR is reported to be as high as 40% in at-risk populations, and a significant proportion of patients are not screened. DR is an optimally suited area for AI, which can help overcome screening barriers, improving access and preventing vision loss. Early studies of AI and DR focused on lesion detection and have evolved classifying DR with a predominant focus on standard color fundus photography. In 2016, both Abràmoff et al . and Gulshan et al . reported algorithms using convolutional neural networks (CNN) that were able to detect referrable diabetic retinopathy (area under the curve [AUC] of 0.980 and 0.991, respectively). Subsequent studies used larger data sets demonstrated good detection of referrable diabetic retinopathy with AUCs of 0.97 and 0.94, respectively. Further studies have prospectively evaluated the performance of AI in detecting referrable DR. Heydon et al . reported that EyeArt v2.1 had a 95.7% sensitivity for referrable DR. In addition to standard fundus photography, AI detection of DR has been studied using optical coherence tomography (OCT) images, ultra-widefield (UWF) imaging, and even smartphone-captured retinal images. Intraretinal fluid identified by OCT can be identified accurately by CNN; for instance Lee et al . used manually segmented macular OCT images to develop a CNN capable of detecting macular edema (with a cross-validation Dice coefficient of 0.911). UWF imaging allows visualization of up to 200° of the fundus, potentially catching additional diabetic-related peripheral disease. Nagasawa et al . found high sensitivity (94.7%) and specificity (97.2%) of a CNN in detecting proliferative DR on UWF images. Similarly, Wang et al . found high sensitivity (91.7%), though limited specificity (50.0%), for referrable DR using UWF images. Access and availability of imaging tools are challenges in effective DR screening. Natarajan et al . reported a smartphone-based, offline AI system that had a high sensitivity for detecting referrable diabetic retinopathy. There are a number of commercially available AI-developed DR screening platforms including IDx-DR (Iowa), which holds FDA approval, and EyeArt (California), which is designated as a European Union Class IIa medical device. Age-related macular degeneration Age-related macular degeneration (AMD) is a common cause of vision loss, with an estimated 196 million patients impacted globally. Early detection and treatment of wet AMD can minimize vision loss. Given the burden of disease, AI could assist in mass screening of OCT and retinal photographs without in-person evaluations. The research in this field started from ML with databases of under 1000 images to now over 490,000 images with high sensitivity and specificity rates. Burlina et al . used a database of over 130,000 images from 4613 patients to develop a DL algorithm for automated detection of AMD. Their DL system reported a 92% accuracy in identifying individuals with moderate and advanced AMD. Similarly, a study by Vaghefi et al . demonstrated that combining DL modalities in AMD—specifically fundus photographs, OCT, and OCT angiography scans—increased accuracy from 91% to 96% in detecting AMD compared to OCT alone. Keenan et al . recently published a paper on an AI algorithm that could accurately quantify volume of fluid in neovascular AMD patients. This has potential in monitoring response to treatment. Deep learning has also been used to quantify other key features associated with AMD including intraretinal fluid (IRF), subretinal fluid (SRF), pigment epithelial detachment (PED), ellipsoid zone loss, drusen, fibrosis, and subretinal hyperreflective material. Similarly, Moraes et al . published a paper on automated quantification of key features in AMD while Fu et al . demonstrated that automatically captured quantitative parameters could predict visual change following treatment. Additional applications in the retina Moving beyond diagnosis of individual disease entities, De Fauw et al . reported a deep learning architecture that identified referrable retinal disease via OCT images, achieving a performance comparable to retina subspecialists (AUC = 99.21). This system was able to identify neovascular AMD, geographic atrophy, drusen, macular edema, macular holes, central serous retinopathy, vitreomacular traction, and epiretinal membrane. Deep learning is able to predict retinal function on microperimetry based on structural assessment of OCT in patients with Stargardt disease. This may assist in assessing patients with inherited retinal disease while monitoring progression or treatment effect in clinical trials. Other AI systems are able to identify central serous retinopathy, pachychoroid vasculopathy, sickle cell disease, and macular telangiectasia. Aside from ocular diagnosis, DL can also predict demographics including age, gender, and cardiovascular risk factors such as systolic blood pressure, smoking status, and major adverse cardiac events.
Screening for diabetic retinopathy (DR) is essential as it facilitates early detection and treatment, thereby preventing vision loss. This is relevant in Canada as 3.7 million people have diabetic retinopathy; the incidence of DR is reported to be as high as 40% in at-risk populations, and a significant proportion of patients are not screened. DR is an optimally suited area for AI, which can help overcome screening barriers, improving access and preventing vision loss. Early studies of AI and DR focused on lesion detection and have evolved classifying DR with a predominant focus on standard color fundus photography. In 2016, both Abràmoff et al . and Gulshan et al . reported algorithms using convolutional neural networks (CNN) that were able to detect referrable diabetic retinopathy (area under the curve [AUC] of 0.980 and 0.991, respectively). Subsequent studies used larger data sets demonstrated good detection of referrable diabetic retinopathy with AUCs of 0.97 and 0.94, respectively. Further studies have prospectively evaluated the performance of AI in detecting referrable DR. Heydon et al . reported that EyeArt v2.1 had a 95.7% sensitivity for referrable DR. In addition to standard fundus photography, AI detection of DR has been studied using optical coherence tomography (OCT) images, ultra-widefield (UWF) imaging, and even smartphone-captured retinal images. Intraretinal fluid identified by OCT can be identified accurately by CNN; for instance Lee et al . used manually segmented macular OCT images to develop a CNN capable of detecting macular edema (with a cross-validation Dice coefficient of 0.911). UWF imaging allows visualization of up to 200° of the fundus, potentially catching additional diabetic-related peripheral disease. Nagasawa et al . found high sensitivity (94.7%) and specificity (97.2%) of a CNN in detecting proliferative DR on UWF images. Similarly, Wang et al . found high sensitivity (91.7%), though limited specificity (50.0%), for referrable DR using UWF images. Access and availability of imaging tools are challenges in effective DR screening. Natarajan et al . reported a smartphone-based, offline AI system that had a high sensitivity for detecting referrable diabetic retinopathy. There are a number of commercially available AI-developed DR screening platforms including IDx-DR (Iowa), which holds FDA approval, and EyeArt (California), which is designated as a European Union Class IIa medical device.
Age-related macular degeneration (AMD) is a common cause of vision loss, with an estimated 196 million patients impacted globally. Early detection and treatment of wet AMD can minimize vision loss. Given the burden of disease, AI could assist in mass screening of OCT and retinal photographs without in-person evaluations. The research in this field started from ML with databases of under 1000 images to now over 490,000 images with high sensitivity and specificity rates. Burlina et al . used a database of over 130,000 images from 4613 patients to develop a DL algorithm for automated detection of AMD. Their DL system reported a 92% accuracy in identifying individuals with moderate and advanced AMD. Similarly, a study by Vaghefi et al . demonstrated that combining DL modalities in AMD—specifically fundus photographs, OCT, and OCT angiography scans—increased accuracy from 91% to 96% in detecting AMD compared to OCT alone. Keenan et al . recently published a paper on an AI algorithm that could accurately quantify volume of fluid in neovascular AMD patients. This has potential in monitoring response to treatment. Deep learning has also been used to quantify other key features associated with AMD including intraretinal fluid (IRF), subretinal fluid (SRF), pigment epithelial detachment (PED), ellipsoid zone loss, drusen, fibrosis, and subretinal hyperreflective material. Similarly, Moraes et al . published a paper on automated quantification of key features in AMD while Fu et al . demonstrated that automatically captured quantitative parameters could predict visual change following treatment.
Moving beyond diagnosis of individual disease entities, De Fauw et al . reported a deep learning architecture that identified referrable retinal disease via OCT images, achieving a performance comparable to retina subspecialists (AUC = 99.21). This system was able to identify neovascular AMD, geographic atrophy, drusen, macular edema, macular holes, central serous retinopathy, vitreomacular traction, and epiretinal membrane. Deep learning is able to predict retinal function on microperimetry based on structural assessment of OCT in patients with Stargardt disease. This may assist in assessing patients with inherited retinal disease while monitoring progression or treatment effect in clinical trials. Other AI systems are able to identify central serous retinopathy, pachychoroid vasculopathy, sickle cell disease, and macular telangiectasia. Aside from ocular diagnosis, DL can also predict demographics including age, gender, and cardiovascular risk factors such as systolic blood pressure, smoking status, and major adverse cardiac events.
While AI has been heavily researched in the posterior segment, the application of AI in anterior segment disease and diagnostic research is now coming to the forefront of ophthalmology literature. Conjunctivitis Using the Japan Ocular Allergy Society (JOAS) classification, Hiroki Masumoto trained a neural network to grade conjunctival hyperemia. The system graded the severity of the hyperemia with a high degree of accuracy. Trachoma is a blinding disease secondary to infection by ocular strains of Chlamydia trachomatis . Using eyelid images from a database of two clinical trials—the Niger arm of the Partnership for Rapid Elimination of Trachoma trial (PRET) and the Trachoma Amelioration in Northern Amhara (TANA) trial—machine learning was used to accurately classify trachomatis changes. Lacrimal apparatus Lacrimal scintigraphy (LS) is an objective and reliable method of studying the lacrimal drainage system and tear flow. Park et al . developed machine and deep learning algorithms using LS images to classify lacrimal duct pathology in patients with epiphora. The system showed accuracy comparable to a trained oculoplastic specialist. Dry eye Meibomian glands (MGs) are believed to play a critical role in ocular surface health. Dysfunction of MGs is the most frequent cause of dry eyes. Meibography, or photo documentation of MGs of the eyelids with transillumination or infrared light, is a common test for the diagnosis, treatment, and management of MG dysfunction (MGD). Wang et al . developed a DL approach to digitally segment MG atrophy and computing percent atrophy in meibography images, providing quantitative information of gland atrophy. The algorithm achieved a 95.6% meiboscore grading accuracy, outperforming the lead clinical investigator by 16.0% and the clinical team by 40.6%. The algorithm also achieved a 97.6% and 95.4% accuracy for eyelid and atrophy segmentations, respectively. Stegman et al . developed a ML segmentation algorithm to measure tear meniscus thickness via OCT to measure tear film quantity. The system showed reproducible results although the sample size was small. Keratoconus Keratoconus is a non-inflammatory corneal disorder characterized by stromal thinning and astigmatism. Kuo et al . retrospectively collected corneal topographic results over time to develop a DL algorithm to detect keratoconus. The model had fair accuracy for keratoconus screening, and furthermore, it predicted subclinical keratoconus. The sensitivity and specificity of all CNN models were over 0.90, and the AUC reached 0.995 in one of the three tested models. Dos Santos et al . designed and trained a neural network (CorneaNet) to segment cornea OCT images. The algorithm measured the thickness of the three main layers, namely the epithelium, Bowman’s layer, and the middle stroma, in patients with keratoconus and those with healthy eyes. All models revealed very similar performances when identifying keratoconus and had a validation accuracy ranging from 99.45% to 99.57%. Lavric et al . devised the KeratoDetect, a neural network that achieved a high level of performance in detecting keratoconus from cornea topographies. With an accuracy of 99.33%, the author claimed that it could assist ophthalmologists in rapid screening of patients. Similarly, Kamiya et al . evaluated the diagnostic accuracy of six colored anterior segment OCT maps: anterior elevation, anterior curvature, posterior elevation, posterior curvature, total refractive power, and pachymetry map. DL was able to identify and classify keratoconus eyes and stage of disease. Shi et al . developed an automated classification system using ML and combining Scheimpflug and ultra-high resolution OCT. The system showed excellent performance (AUC = 0.93) in discriminating subclinical keratoconus from normal corneas. The author found that epithelial features assessed by OCT were the most important features when identifying keratoconus. Abdelmotaal et al . was able to identify keratoconus and subclinical keratoconus via ML using color-coded corneal maps obtained by a Scheimpflug camera.
Using the Japan Ocular Allergy Society (JOAS) classification, Hiroki Masumoto trained a neural network to grade conjunctival hyperemia. The system graded the severity of the hyperemia with a high degree of accuracy. Trachoma is a blinding disease secondary to infection by ocular strains of Chlamydia trachomatis . Using eyelid images from a database of two clinical trials—the Niger arm of the Partnership for Rapid Elimination of Trachoma trial (PRET) and the Trachoma Amelioration in Northern Amhara (TANA) trial—machine learning was used to accurately classify trachomatis changes.
Lacrimal scintigraphy (LS) is an objective and reliable method of studying the lacrimal drainage system and tear flow. Park et al . developed machine and deep learning algorithms using LS images to classify lacrimal duct pathology in patients with epiphora. The system showed accuracy comparable to a trained oculoplastic specialist.
Meibomian glands (MGs) are believed to play a critical role in ocular surface health. Dysfunction of MGs is the most frequent cause of dry eyes. Meibography, or photo documentation of MGs of the eyelids with transillumination or infrared light, is a common test for the diagnosis, treatment, and management of MG dysfunction (MGD). Wang et al . developed a DL approach to digitally segment MG atrophy and computing percent atrophy in meibography images, providing quantitative information of gland atrophy. The algorithm achieved a 95.6% meiboscore grading accuracy, outperforming the lead clinical investigator by 16.0% and the clinical team by 40.6%. The algorithm also achieved a 97.6% and 95.4% accuracy for eyelid and atrophy segmentations, respectively. Stegman et al . developed a ML segmentation algorithm to measure tear meniscus thickness via OCT to measure tear film quantity. The system showed reproducible results although the sample size was small.
Keratoconus is a non-inflammatory corneal disorder characterized by stromal thinning and astigmatism. Kuo et al . retrospectively collected corneal topographic results over time to develop a DL algorithm to detect keratoconus. The model had fair accuracy for keratoconus screening, and furthermore, it predicted subclinical keratoconus. The sensitivity and specificity of all CNN models were over 0.90, and the AUC reached 0.995 in one of the three tested models. Dos Santos et al . designed and trained a neural network (CorneaNet) to segment cornea OCT images. The algorithm measured the thickness of the three main layers, namely the epithelium, Bowman’s layer, and the middle stroma, in patients with keratoconus and those with healthy eyes. All models revealed very similar performances when identifying keratoconus and had a validation accuracy ranging from 99.45% to 99.57%. Lavric et al . devised the KeratoDetect, a neural network that achieved a high level of performance in detecting keratoconus from cornea topographies. With an accuracy of 99.33%, the author claimed that it could assist ophthalmologists in rapid screening of patients. Similarly, Kamiya et al . evaluated the diagnostic accuracy of six colored anterior segment OCT maps: anterior elevation, anterior curvature, posterior elevation, posterior curvature, total refractive power, and pachymetry map. DL was able to identify and classify keratoconus eyes and stage of disease. Shi et al . developed an automated classification system using ML and combining Scheimpflug and ultra-high resolution OCT. The system showed excellent performance (AUC = 0.93) in discriminating subclinical keratoconus from normal corneas. The author found that epithelial features assessed by OCT were the most important features when identifying keratoconus. Abdelmotaal et al . was able to identify keratoconus and subclinical keratoconus via ML using color-coded corneal maps obtained by a Scheimpflug camera.
Glaucoma is the second most common cause of irreversible blindness worldwide. Early detection of glaucoma has been shown to reduce vision loss. Digital photography of the optic nerve is a common method to screen for glaucoma and is used effectively as part of many teleglaucoma programs. Comprehensive evaluation of a glaucoma suspect might include spectral domain optical coherence tomography (SD-OCT), perimetry, tonometry, pachymetry, and gonioscopy. AI algorithms have been developed to identify optic nerve changes via optic disc photographs and SD-OCT and thereby predict glaucomatous field changes. Very little work has been done to date with DL for tonometry, pachymetry, and gonioscopy. Glaucoma diagnosis and screening Color fundus photography of the optic nerve is an inexpensive and available method to screen for glaucoma. ML has been utilized to improve identification of early glaucomatous changes to the optic nerve captured by photography. Computer segmentation of the optic nerve into disc, cup, and vessels to create a glaucoma score showed an AUC of 98.2% when compared to standard cup-to-disc ratio score (AUC 91.4%) on three glaucoma-related public datasets. Once incorporated into teleglaucoma screening programs, such algorithms will automate detection of early glaucoma. Simultaneous or near-simultaneous capture of optic nerve OCT scans at the time of color fundus photography has enabled the rapid development of highly accurate DL algorithms for identification of glaucomatous nerve damage. This method of training DL systems—first published by Medeiros et al . in 2018 using high resolution digital images captured in the research setting—has removed the errors and biases associated with human grading of nerve damage. They showed that once trained with OCT data as truth, the DL system could discriminate glaucomatous optic nerves from healthy eyes with the area under the ROC curves of 0.944 (95% confidence interval [CI], 0.912–0.966) and 0.940 (95% CI, 0.902–0.966), respectfully ( P = 0.724). This seminal work has been replicated and confirmed through publications by various groups around the world utilizing other fundus image repositories. Machine-to-machine (M2M) deep learning algorithms trained with SD-OCT to assess monoscopic optic nerve photographs are able to identify glaucomatous optic nerve damage more accurately than glaucoma specialists. This increase in accuracy suggests that DL systems will replace human review of disc photographs for glaucoma screening programs of the future. It is possible to detect progression of glaucomatous nerve damage by fundus photography utilizing DL algorithms confirmed by OCT. Medeiros et al . assessed temporally disparate disc photographs of 5529 patients over time. They utilized a DL CNN trained with OCT data from the same patients. The ROC curve area was 0.86 (95% CI, 0.83–0.88) when differentiating between progressors and non-progressors. Agitha et al . showed a similar benefit of a DL model used on 1113 fundus images to achieve an accuracy of 94%, sensitivity of 85%, and specificity of 100% in the automatic diagnosis of glaucoma. In the setting of glaucoma detection, OCT has been utilized primarily to provide an objective truth reference for glaucoma-related DL CNN training. Recent work has shown that OCT DL algorithms are able to identify glaucomatous damage reliably by utilizing various datasets and various artificial intelligence algorithms. DL algorithms are also able to predict glaucomatous visual field with OCT nerve topography. The sensitivity and specificity of ML classifiers to diagnose glaucoma can be improved by combining standard automated perimetry and OCT data when compared to OCT alone. Standard automated perimetry (SAP) is perhaps the most exciting area to assess with ML. The availability of long-term visual field test results, often over decades, in patients with and without glaucoma has provided extensive datasets for artificial intelligence researchers. Artificial intelligence is able to identify glaucoma four years in advance of diagnosis using original visual field data with good reliability. Asaoka et al . retrospectively assessed visual field data over 15 years in 51 patients with open-angle glaucoma and 87 healthy participants. Their deep feedforward neural network (FNN) showed an AUC of 92.6% (95% CI, 89.8%–95.4%) when identifying pre-perimetric glaucoma. Unsupervised ML classifiers showed a sensitivity of 82.8% and specificity of 93.1% in the identification of glaucomatous patterns by frequency doubling technology (FDT). This suggests that machine learning could become an important adjunct where visual field testing is performed as part of any glaucoma screening program. Glaucoma progression Machine classifier algorithms are able to identify glaucoma progression by visual fields. Progression of patterns (POP) is a variational machine learning classifier that was able to identify more eyes with progression of glaucomatous optic neuropathy in glaucoma suspects and glaucoma than were identified by guided progression analysis (GPA). Deep learning is also able to forecast future Humphrey visual fields in patients with glaucoma. Wen et al . was able to predict the development of future visual field changes up to five years in the future using deep learning networks.
Color fundus photography of the optic nerve is an inexpensive and available method to screen for glaucoma. ML has been utilized to improve identification of early glaucomatous changes to the optic nerve captured by photography. Computer segmentation of the optic nerve into disc, cup, and vessels to create a glaucoma score showed an AUC of 98.2% when compared to standard cup-to-disc ratio score (AUC 91.4%) on three glaucoma-related public datasets. Once incorporated into teleglaucoma screening programs, such algorithms will automate detection of early glaucoma. Simultaneous or near-simultaneous capture of optic nerve OCT scans at the time of color fundus photography has enabled the rapid development of highly accurate DL algorithms for identification of glaucomatous nerve damage. This method of training DL systems—first published by Medeiros et al . in 2018 using high resolution digital images captured in the research setting—has removed the errors and biases associated with human grading of nerve damage. They showed that once trained with OCT data as truth, the DL system could discriminate glaucomatous optic nerves from healthy eyes with the area under the ROC curves of 0.944 (95% confidence interval [CI], 0.912–0.966) and 0.940 (95% CI, 0.902–0.966), respectfully ( P = 0.724). This seminal work has been replicated and confirmed through publications by various groups around the world utilizing other fundus image repositories. Machine-to-machine (M2M) deep learning algorithms trained with SD-OCT to assess monoscopic optic nerve photographs are able to identify glaucomatous optic nerve damage more accurately than glaucoma specialists. This increase in accuracy suggests that DL systems will replace human review of disc photographs for glaucoma screening programs of the future. It is possible to detect progression of glaucomatous nerve damage by fundus photography utilizing DL algorithms confirmed by OCT. Medeiros et al . assessed temporally disparate disc photographs of 5529 patients over time. They utilized a DL CNN trained with OCT data from the same patients. The ROC curve area was 0.86 (95% CI, 0.83–0.88) when differentiating between progressors and non-progressors. Agitha et al . showed a similar benefit of a DL model used on 1113 fundus images to achieve an accuracy of 94%, sensitivity of 85%, and specificity of 100% in the automatic diagnosis of glaucoma. In the setting of glaucoma detection, OCT has been utilized primarily to provide an objective truth reference for glaucoma-related DL CNN training. Recent work has shown that OCT DL algorithms are able to identify glaucomatous damage reliably by utilizing various datasets and various artificial intelligence algorithms. DL algorithms are also able to predict glaucomatous visual field with OCT nerve topography. The sensitivity and specificity of ML classifiers to diagnose glaucoma can be improved by combining standard automated perimetry and OCT data when compared to OCT alone. Standard automated perimetry (SAP) is perhaps the most exciting area to assess with ML. The availability of long-term visual field test results, often over decades, in patients with and without glaucoma has provided extensive datasets for artificial intelligence researchers. Artificial intelligence is able to identify glaucoma four years in advance of diagnosis using original visual field data with good reliability. Asaoka et al . retrospectively assessed visual field data over 15 years in 51 patients with open-angle glaucoma and 87 healthy participants. Their deep feedforward neural network (FNN) showed an AUC of 92.6% (95% CI, 89.8%–95.4%) when identifying pre-perimetric glaucoma. Unsupervised ML classifiers showed a sensitivity of 82.8% and specificity of 93.1% in the identification of glaucomatous patterns by frequency doubling technology (FDT). This suggests that machine learning could become an important adjunct where visual field testing is performed as part of any glaucoma screening program.
Machine classifier algorithms are able to identify glaucoma progression by visual fields. Progression of patterns (POP) is a variational machine learning classifier that was able to identify more eyes with progression of glaucomatous optic neuropathy in glaucoma suspects and glaucoma than were identified by guided progression analysis (GPA). Deep learning is also able to forecast future Humphrey visual fields in patients with glaucoma. Wen et al . was able to predict the development of future visual field changes up to five years in the future using deep learning networks.
Retinopathy of prematurity Retinopathy of prematurity (ROP) is a vasoproliferative retinal disease that is a leading cause of childhood blindness. The Early Treatment for Retinopathy of Prematurity (ETROP) study has shown that screening and early intervention is critical for improving visual outcomes. Improved survival of extremely premature infants has increased the prevalence of ROP, particularly in developing nations. AI can play a vital role in assisting with ROP diagnosis, thereby improving treatment outcomes. In a study by Brown et al. , researchers showed that a DL algorithm trained with wide-field retinal photographs outperformed 6 out of 8 ROP experts on an independent data set of 100 images in diagnosing ROP. The algorithm was trained with a database of 5511 fundus images and demonstrated a 93% sensitivity and 94% specificity in determining ROP severity. Tong et al . developed a neural network that was trained for ROP identification with 36,000 fundus images. This system achieved an accuracy of 0.903 for ROP severity classification and demonstrated comparable to or better diagnostic ability when compared to retina subspecialist. Other research have shown similar success with ROP severity classification and deep learning. Integration of AI into an ROP screening program will likely occur in the near future. Congenital cataracts Pediatric cataract is one of the leading causes of juvenile blindness, with an estimated prevalence of 4.24 per 10,000 live births. Congenital cataract guardian (CC-Guardian) is an AI agent that incorporates individualized prediction and scheduling, and intelligent telehealth follow-up computing for congenital cataracts. The system exhibits high sensitivity and specificity and has been integrated to a web-based smartphone app. The intelligent agent consists of three functional modules: (i) a prediction module that identifies potential high-risk congenital cataract patients who are likely to suffer complications, (ii) a dispatching module that schedules individual follow-up based on the prediction results, and (iii) a telehealth module that makes intervention decisions in each follow-up examination. All the records were derived from routine examinations at the Childhood Cataract Program of the Chinese Ministry of Health. Amblyopia In Korea, Chun et al . assessed a DL system to predict the range of refractive error in children using a smartphone photorefraction image to screen for amblyopia and compared it to a cycloplegic refraction. The DL tool showed an accuracy of 81.6%.
Retinopathy of prematurity (ROP) is a vasoproliferative retinal disease that is a leading cause of childhood blindness. The Early Treatment for Retinopathy of Prematurity (ETROP) study has shown that screening and early intervention is critical for improving visual outcomes. Improved survival of extremely premature infants has increased the prevalence of ROP, particularly in developing nations. AI can play a vital role in assisting with ROP diagnosis, thereby improving treatment outcomes. In a study by Brown et al. , researchers showed that a DL algorithm trained with wide-field retinal photographs outperformed 6 out of 8 ROP experts on an independent data set of 100 images in diagnosing ROP. The algorithm was trained with a database of 5511 fundus images and demonstrated a 93% sensitivity and 94% specificity in determining ROP severity. Tong et al . developed a neural network that was trained for ROP identification with 36,000 fundus images. This system achieved an accuracy of 0.903 for ROP severity classification and demonstrated comparable to or better diagnostic ability when compared to retina subspecialist. Other research have shown similar success with ROP severity classification and deep learning. Integration of AI into an ROP screening program will likely occur in the near future.
Pediatric cataract is one of the leading causes of juvenile blindness, with an estimated prevalence of 4.24 per 10,000 live births. Congenital cataract guardian (CC-Guardian) is an AI agent that incorporates individualized prediction and scheduling, and intelligent telehealth follow-up computing for congenital cataracts. The system exhibits high sensitivity and specificity and has been integrated to a web-based smartphone app. The intelligent agent consists of three functional modules: (i) a prediction module that identifies potential high-risk congenital cataract patients who are likely to suffer complications, (ii) a dispatching module that schedules individual follow-up based on the prediction results, and (iii) a telehealth module that makes intervention decisions in each follow-up examination. All the records were derived from routine examinations at the Childhood Cataract Program of the Chinese Ministry of Health.
In Korea, Chun et al . assessed a DL system to predict the range of refractive error in children using a smartphone photorefraction image to screen for amblyopia and compared it to a cycloplegic refraction. The DL tool showed an accuracy of 81.6%.
The development and modern usage of AI in research has become a breakthrough for optimization and efficiency. With the growth of electronic medical records, healthcare providers and hospitals are able to accumulate a wealth of patient information. A common barrier to sifting through this information is the time required to appropriately review each individual item. With the advent of AI, after developing a computer-generated algorithm or suitably training an automated system to batch patient information, data collection can be completed in a fraction of the time that it would take to be done manually. Ophthalmology is a medical specialty that is conducive to retrieving these large amounts of data due to its rapid access of ophthalmic imaging and objective markers (e.g., visual acuity, intraocular pressure [IOP], retinal thickness, etc.). The Intelligent Research in Sight (IRIS) Registry is one of the largest clinical datasets that includes data about demographics, disease conditions, and visit rates in ophthalmology. The Smart Eye Database stores electronic medical records of ophthalmology patients which are stratified by eye conditions. Datasets such as IRIS and the Smart Eye Database allow us to appreciate subtle correlations, conduct multicenter studies, incorporate multimodal analyses, identify novel imaging patterns, and increase the power in studies, all of which may not be possible with smaller sets of data. As described by Joshi et al. , this large collection of medical information, or “big data,” serves as a perfect substrate for AI, ML, and DL to develop and run algorithms at a scale that would never have been possible before.
Ophthalmology is a specialty well-suited for AI integration. The extensive use of multi-modal digital imaging and diagnostic tests captured over time in all ophthalmology subspecialties provide a treasure trove of opportunities for machine learning that are now being realized. Artificial intelligence and machine learning solutions have begun the evolution from research setting to a clinical tool that will be invaluable for ophthalmologists in all clinical settings. Financial support and sponsorship Nil. Conflicts of interest There are no conflicts of interest.
Nil.
There are no conflicts of interest.
|
Characterization of sequence variations in the extended flanking regions using massively parallel sequencing in 21 A-STRs and 21 Y-STRs | 1e51bdf5-0905-4c21-a0ed-3990e6019b00 | 11380771 | Forensic Medicine[mh] | Massively parallel sequencing (MPS), also referred to as next-generation sequencing (NGS), has recently gained significant attention for its immense potential in the field of forensic genetics . Short tandem repeat (STR), a forensic genetic marker, has gained widespread recognition in criminal investigations and court trials over the past two decades . The utilization of MPS for STR analysis has demonstrated numerous advantages in several previous studies . Compared to conventional capillary electrophoresis (CE), MPS can analyze more STR loci simultaneously without limitation by restrictions in size-based separation or fluorescence dye detection . Besides, the sequence variations within the core repeat and flanking regions of STRs could be identified by sequencing and significantly improved the identification ability of STRs . Currently, the mainstream MPS platforms used in forensic were usually the long-reads sequencing platforms, including Novaseq™ (PE250, Illumina, San Diego, CA, USA), MiSeq™ System (PE300, Illumina), MiSeq FGx™ System (PE300, Verogen, San Diego, CA, USA), Ion torrent PGM (SE400, Thermo Fisher, Foster City, CA, USA), Ion torrent S5 (SE400, Thermo Fisher), and MGISEQ-2000 platform (SE400, MGI Tech, Shenzhen, Guangdong, China) . Due to the current technical limitations, the quality of the MPS sequencing read at its end tends to be comparatively lower so that there are restrictions on the length of reads in the flanking regions of STRs. Despite the mention of flanking region polymorphisms in several previous studies , the majority of research has primarily focused on the core repeat regions of STRs or variations in the adjacent flanking regions. Consequently, the current state of research lacks systematic investigations into the variations in flanking regions, particularly those occurring at more distal positions from the core repeat regions. The sequence variants could also improve the STR performance in forensic application and were likely associated with bioinformatics analysis , even caused discordances between MPS and CE methods in some cases . Thus, the flanking region variations in commonly used forensic STRs need to be further explored. For current commercial MPS STR panels, such as ForenSeq™ DNA Signature Prep Kit (Verogen) and Precision ID GlobalFiler™ NGS STR Panel v2 (Thermo Fisher), the coverage range in the flanking regions of these panels was limited. Thus, we developed a MPS system to detect not only the STR core repeat regions, but also variations located in distal positions in the flanking regions. For the system, DNA samples were amplified using two sets of multiplex PCR assay, respectively. The amplification of each STR locus was performed using two pairs of primers, with each primer pair encompassing a wider flanking region on one side (ensuring coverage of the core repeat region). Consequently, the flanking region variations of each STR loci were maximally detectable within the length range compatible with current MPS technology reads. Hence, the alleles observed in the present study covered more flanking regions. The concordance of the system was confirmed by comparing with CE method. The allele frequencies and forensic parameters were calculated in Chinese Han population. Moreover, we statistically analyzed the sequence variations in the flanking regions and evaluated their impacts on forensic application. Sample collection and DNA extraction This study was approved by the Ethical Committee of Fudan University (2020016). Peripheral blood samples were collected using FTA cards from 350 unrelated Chinese Han male individuals. All individuals provided their written informed consent for the collection of blood samples. Genomic DNA was extracted using the BioRobotEZ1 Advanced XL and EZ1 DNA Investigator kits (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. DNA was then quantified using a Qubit 2.0 Fluorometer and a Qubit dsDNA HS Assay Kit (Thermo Fisher). Selection of STR loci In this study, twenty-one autosomal STRs (A-STR) were selected (CSF1PO, FGA, TH01, TPOX, vWA, D1S1656, D2S1338, D2S441, D3S1358, D5S818, D6S1043, D7S820, D8S1179, D10S1248, D12S391, D13S317, D16S539, D18S51, D19S433, D21S11 and D22S1045), which were FBI CODIS core STR loci or frequently used in forensic. Besides, twenty-one Y-chromosomal STRs (Y-STR) were selected (DYS19, DYS385a-b, DYS389I-II, DYS390, DYS391, DYS392, DYS437, DYS438, DYS439, DYS444, DYS447, DYS448, DYS456, DYS533, DYS549, DYS570, DYS576, DYS635 and GATA-H4), which were commonly used for the construction of Chinese Y-STR database. Construction of multiplex PCR systems In order to enhance the detection of flanking regions surrounding STRs within MPS reads, two pairs of primers were meticulously designed for each STR, ensuring comprehensive coverage of the core region containing the longest known allele. For each pair of primers, one primer binding site was designed in close proximity to the core repeat region of STR, while the other was extended towards the flanking region to encompass the maximum possible length range suitable for MPS reads (considering primer length, sequencing joint, and barcode; typically reaching approximately 200 bp). It is advisable to avoid utilizing the identified SNP within the adjacent region of STR for primer binding. The schematic diagram of primers design was shown in Fig. . The primer pair for each locus was designed using Oligo Primer Analysis Software Version 7 (Molecular Biology Insights, Cascade, USA) ( https://oligo.net/ ), and subsequently assessed for potential non-specific hybridizations through the NCBI Basic Local Alignment Search Tool (BLAST) ( http://blast.ncbi.nlm.nih.gov/Blast.cgi ). The positions and sizes of the amplicons are presented in Table A and B, while the primer sequences can be found in Supplementary Table . The designed primers were divided into two sets of multiplex PCR systems (pool1 and pool2) for amplification to avoid interaction. The KAPA 2G Fast Multiplex PCR kit (Roche Diagnostics, Rotkreuz, Switzerland) was employed for multiplex amplification. The amplification reaction comprised of 10 µl of KAPA 2G Fast Multiplex PCR Mix, 5 ng of DNA templates, and 1 µl of primer Mix. Amplification was conducted on a GeneAmp™ 9700 PCR System (Thermo Fisher) with the following parameters: initial denaturation at 96℃ for 3 min; followed by 20 cycles of denaturation at 96℃ for 30 s, annealing at 60℃ for 4 min; final extension step at 72℃ for 20 min and hold at 4℃. Amplicons were purified using the Agencourt AMPure XP PCR Purification System (Beckman Coulte, Fullerton, CA, USA) as recommended by the manufacturers. Library preparation, normalized and sequencing The purified amplicons were ligated to adapters with barcodes using the KAPA 2G Fast Multiplex PCR kit for each library. The ligation reaction included 10 µl of KAPA 2G Fast Multiplex PCR Mix, 1 µl of 4 μm TRA-70xR, 1 µl of 4 μm TRA-50xF, and 8 ng of purified amplicons. The enrichment reaction was performed under the following conditions: initial denaturation at 96℃ for 3 min; followed by a total of ten cycles consisting of denaturation at 96℃ for 30 s, annealing at 58℃ for 30 s, extension at 72℃ for 30s; final extension step at 72℃ for 2 min; and holding temperature set to 4℃. Library purification was carried out using the Agencourt AMPure XP PCR Purification System according to the manufacturer’s recommendations. Subsequently, pool1 and pool2 from the same sample were combined. The quantity of each library was determined using the Qubit 2.0 Fluorometer with the Qubit dsDNA HS Assay Kit. The concentration unit ng/µl was converted to nM using the formula nM = ng/µl / (660 g/mol ×300 bp) ×10 6 ( https://www.science.smith.edu/wp-content/uploads/sites/36/2015/09/Converting-ng-to-nM.pdf ). Each library was normalized and pooled to an equimolar concentration of 10 nM, as recommended by the manufacturer. Pooled libraries were sequenced on the MiSeq FGx™ System (Verogen) with Miseq V3 reagent (PE300 strategy), following standard protocol. A total of five sequencing runs were performed, each consisting of 70 samples along with positive and negative controls. Data analysis The raw reads were filtered by removing the uncompleted ends and sequencing errors. The reads were then screened and only the ones contain the primer sequences for STRs were used for STR genotyping. A custom Perl script was used to annotate the STRs by directly comparing the sequence in each read with core repeat region for each STR allele. The structure of the STR core repeat region was compiled based on data from the STR Sequencing Project or previous studies’ reports . Reads which contain the primer sequences for STRs but can’t annotate to any collected STR allele were used for variation calling. These reads were firstly mapped to human reference genome (GRCh38) using bowtie (version 2-2.2.5) with the “--very-sensitive-local” model . Sequences that differ by up to 1 bp substitution or 1–4 bp insertion/deletion were tolerated in order to get more information of variations. Variations with support of both forward and reverse reads were recorded. The selected reads were split into 5’ flanking, 3’ flanking and core repeat sequences depending on the STR nomenclature recommended by the DNA commission of the International Society for Forensic Genetics (ISFG) . A custom Perl script was used to annotate the sequence variation in each read. Alleles with a minimum of 30× coverage were used for further analysis. The balance threshold of alleles within the locus was set at 60%, and the stutter threshold was uniformly set at 30%. The core repeat sequences from one sample should be consistent between pool1 (5’ flanking) and pool2 (3’ flanking). Thus, alleles from two pools were merged according to core repeat sequences manually. All sequence data were exported and reviewed manually. All alleles identified in the sequencing analysis were then integrated for further analysis. Capillary electrophoresis (CE) assessing concordance Ninety-two samples were randomly chosen for the purpose of assessing concordance (The samples with allele dropout due to quality control were excluded). The overlap between STR genotypes generated using MPS and CE methods was assessed by employing two commercially available CE-based STR kits. The GlobalFiler™ PCR Amplification kit (Thermo Fisher) was employed for typing A-STRs, while the Yfiler™ Platinum PCR Amplification kit (Thermo Fisher) was used for typing Y-STRs. These two kits encompassed all 42 STRs except for D6S1043. The ABI 3500XL Genetic Analyzer (Thermo Fisher) was utilized to separate and detect the PCR products in accordance with the manufacturer’s guidelines. The electrophoretic results were subsequently analyzed using GeneMapper ® ID-X software v1.4 (Thermo Fisher). To assess any discrepancies in the CE-based typing, we utilized the binary sequence alignment (BAM) file and examined it with the Integrative Genomics Viewer (IGV) . Identification of sequence variations All merged alleles were integrated following the ISFG recommendations, enabling comparison of these data with the STR Sequencing Project records and previous studies . Sanger sequencing was conducted to verify sequence variations not present in the STR Sequencing Project or previous studies, utilizing a BigDye1 Terminator v3.1 Cycle Sequencing Kit (Thermo Fisher). Forensic parameters and Y-haplotype frequencies The length and sequence variants per locus were tallied and categorized based on length (length-based, LB), sequence without flanking region (core repeat regions sequence-based, RSB), and sequence with flanking region (core repeat and flanking regions sequence-based, FSB), respectively. For A-STRs, the forensic parameters were calculated using STRAF , which encompassed allele count (Nall), genotype count (N), genetic diversity (GD), expected heterozygosity (Hexp), random match probability (RMP), discrimination power (DP), polymorphism information content (PIC), power of exclusion (PE), typical paternity index (TPI) and observed heterozygosity (Hobs). Allele frequencies for LB, RSB, and FSB were calculated in Excel by determining the ratio of the allele count to the total count. HWE was tested using STRAF with a Bonferroni correction applied for multiple comparisons. For Y-STRs, the frequencies of Y-haplotypes were determined using a direct counting method that incorporated LB, RSB, and FSB. This study was approved by the Ethical Committee of Fudan University (2020016). Peripheral blood samples were collected using FTA cards from 350 unrelated Chinese Han male individuals. All individuals provided their written informed consent for the collection of blood samples. Genomic DNA was extracted using the BioRobotEZ1 Advanced XL and EZ1 DNA Investigator kits (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. DNA was then quantified using a Qubit 2.0 Fluorometer and a Qubit dsDNA HS Assay Kit (Thermo Fisher). In this study, twenty-one autosomal STRs (A-STR) were selected (CSF1PO, FGA, TH01, TPOX, vWA, D1S1656, D2S1338, D2S441, D3S1358, D5S818, D6S1043, D7S820, D8S1179, D10S1248, D12S391, D13S317, D16S539, D18S51, D19S433, D21S11 and D22S1045), which were FBI CODIS core STR loci or frequently used in forensic. Besides, twenty-one Y-chromosomal STRs (Y-STR) were selected (DYS19, DYS385a-b, DYS389I-II, DYS390, DYS391, DYS392, DYS437, DYS438, DYS439, DYS444, DYS447, DYS448, DYS456, DYS533, DYS549, DYS570, DYS576, DYS635 and GATA-H4), which were commonly used for the construction of Chinese Y-STR database. In order to enhance the detection of flanking regions surrounding STRs within MPS reads, two pairs of primers were meticulously designed for each STR, ensuring comprehensive coverage of the core region containing the longest known allele. For each pair of primers, one primer binding site was designed in close proximity to the core repeat region of STR, while the other was extended towards the flanking region to encompass the maximum possible length range suitable for MPS reads (considering primer length, sequencing joint, and barcode; typically reaching approximately 200 bp). It is advisable to avoid utilizing the identified SNP within the adjacent region of STR for primer binding. The schematic diagram of primers design was shown in Fig. . The primer pair for each locus was designed using Oligo Primer Analysis Software Version 7 (Molecular Biology Insights, Cascade, USA) ( https://oligo.net/ ), and subsequently assessed for potential non-specific hybridizations through the NCBI Basic Local Alignment Search Tool (BLAST) ( http://blast.ncbi.nlm.nih.gov/Blast.cgi ). The positions and sizes of the amplicons are presented in Table A and B, while the primer sequences can be found in Supplementary Table . The designed primers were divided into two sets of multiplex PCR systems (pool1 and pool2) for amplification to avoid interaction. The KAPA 2G Fast Multiplex PCR kit (Roche Diagnostics, Rotkreuz, Switzerland) was employed for multiplex amplification. The amplification reaction comprised of 10 µl of KAPA 2G Fast Multiplex PCR Mix, 5 ng of DNA templates, and 1 µl of primer Mix. Amplification was conducted on a GeneAmp™ 9700 PCR System (Thermo Fisher) with the following parameters: initial denaturation at 96℃ for 3 min; followed by 20 cycles of denaturation at 96℃ for 30 s, annealing at 60℃ for 4 min; final extension step at 72℃ for 20 min and hold at 4℃. Amplicons were purified using the Agencourt AMPure XP PCR Purification System (Beckman Coulte, Fullerton, CA, USA) as recommended by the manufacturers. The purified amplicons were ligated to adapters with barcodes using the KAPA 2G Fast Multiplex PCR kit for each library. The ligation reaction included 10 µl of KAPA 2G Fast Multiplex PCR Mix, 1 µl of 4 μm TRA-70xR, 1 µl of 4 μm TRA-50xF, and 8 ng of purified amplicons. The enrichment reaction was performed under the following conditions: initial denaturation at 96℃ for 3 min; followed by a total of ten cycles consisting of denaturation at 96℃ for 30 s, annealing at 58℃ for 30 s, extension at 72℃ for 30s; final extension step at 72℃ for 2 min; and holding temperature set to 4℃. Library purification was carried out using the Agencourt AMPure XP PCR Purification System according to the manufacturer’s recommendations. Subsequently, pool1 and pool2 from the same sample were combined. The quantity of each library was determined using the Qubit 2.0 Fluorometer with the Qubit dsDNA HS Assay Kit. The concentration unit ng/µl was converted to nM using the formula nM = ng/µl / (660 g/mol ×300 bp) ×10 6 ( https://www.science.smith.edu/wp-content/uploads/sites/36/2015/09/Converting-ng-to-nM.pdf ). Each library was normalized and pooled to an equimolar concentration of 10 nM, as recommended by the manufacturer. Pooled libraries were sequenced on the MiSeq FGx™ System (Verogen) with Miseq V3 reagent (PE300 strategy), following standard protocol. A total of five sequencing runs were performed, each consisting of 70 samples along with positive and negative controls. The raw reads were filtered by removing the uncompleted ends and sequencing errors. The reads were then screened and only the ones contain the primer sequences for STRs were used for STR genotyping. A custom Perl script was used to annotate the STRs by directly comparing the sequence in each read with core repeat region for each STR allele. The structure of the STR core repeat region was compiled based on data from the STR Sequencing Project or previous studies’ reports . Reads which contain the primer sequences for STRs but can’t annotate to any collected STR allele were used for variation calling. These reads were firstly mapped to human reference genome (GRCh38) using bowtie (version 2-2.2.5) with the “--very-sensitive-local” model . Sequences that differ by up to 1 bp substitution or 1–4 bp insertion/deletion were tolerated in order to get more information of variations. Variations with support of both forward and reverse reads were recorded. The selected reads were split into 5’ flanking, 3’ flanking and core repeat sequences depending on the STR nomenclature recommended by the DNA commission of the International Society for Forensic Genetics (ISFG) . A custom Perl script was used to annotate the sequence variation in each read. Alleles with a minimum of 30× coverage were used for further analysis. The balance threshold of alleles within the locus was set at 60%, and the stutter threshold was uniformly set at 30%. The core repeat sequences from one sample should be consistent between pool1 (5’ flanking) and pool2 (3’ flanking). Thus, alleles from two pools were merged according to core repeat sequences manually. All sequence data were exported and reviewed manually. All alleles identified in the sequencing analysis were then integrated for further analysis. Ninety-two samples were randomly chosen for the purpose of assessing concordance (The samples with allele dropout due to quality control were excluded). The overlap between STR genotypes generated using MPS and CE methods was assessed by employing two commercially available CE-based STR kits. The GlobalFiler™ PCR Amplification kit (Thermo Fisher) was employed for typing A-STRs, while the Yfiler™ Platinum PCR Amplification kit (Thermo Fisher) was used for typing Y-STRs. These two kits encompassed all 42 STRs except for D6S1043. The ABI 3500XL Genetic Analyzer (Thermo Fisher) was utilized to separate and detect the PCR products in accordance with the manufacturer’s guidelines. The electrophoretic results were subsequently analyzed using GeneMapper ® ID-X software v1.4 (Thermo Fisher). To assess any discrepancies in the CE-based typing, we utilized the binary sequence alignment (BAM) file and examined it with the Integrative Genomics Viewer (IGV) . All merged alleles were integrated following the ISFG recommendations, enabling comparison of these data with the STR Sequencing Project records and previous studies . Sanger sequencing was conducted to verify sequence variations not present in the STR Sequencing Project or previous studies, utilizing a BigDye1 Terminator v3.1 Cycle Sequencing Kit (Thermo Fisher). The length and sequence variants per locus were tallied and categorized based on length (length-based, LB), sequence without flanking region (core repeat regions sequence-based, RSB), and sequence with flanking region (core repeat and flanking regions sequence-based, FSB), respectively. For A-STRs, the forensic parameters were calculated using STRAF , which encompassed allele count (Nall), genotype count (N), genetic diversity (GD), expected heterozygosity (Hexp), random match probability (RMP), discrimination power (DP), polymorphism information content (PIC), power of exclusion (PE), typical paternity index (TPI) and observed heterozygosity (Hobs). Allele frequencies for LB, RSB, and FSB were calculated in Excel by determining the ratio of the allele count to the total count. HWE was tested using STRAF with a Bonferroni correction applied for multiple comparisons. For Y-STRs, the frequencies of Y-haplotypes were determined using a direct counting method that incorporated LB, RSB, and FSB. Sequencing performance and quality control The analysis of 350 samples involved a total of five sequencing runs. A total of 93,185,964 reads were obtained. There was a good balance among the samples, and at least 50,000 reads were obtained for each sample, indicating that the library was pooled well. After removing invalid data (reads with a coverage of less than 30× and reads of stutters) and primer dimers, the average depth of coverage (Doc) of each STR locus was 330×. The average Doc of A-STR was 456×, which was significantly higher than that of Y-STR (191×). The average Doc of pool1 was 313×, and 347× for pool2. For A-STRs, the highest average Doc locus was D5S818 pool1 (1891×) and the lowest was D21S11 pool2 (175×). For Y-STRs, the highest average Doc was DYS391 pool2 (592×), and the lowest was DYS447 pool2 (67×). The Doc of each STR locus was shown in Supplementary Table . Sequence merging and concordance analysis The sequence merging was performed manually. For A-STRs, all profiles of 350 samples were obtained. The core repeat sequences of each A-STR from pool1 and pool2 of the same sample were compared and confirmed to be consistent. The minimum sequencing depth was higher than 30×. The Y-STRs exhibited a lower depth of coverage compared to the A-STRs. A limited number of Y-STR alleles (DYS19 (6/350), DYS389I-II (2/350), DYS437 (8/350), DYS447 (11/350), DYS448 (16/350), DYS533 (2/350), and GATA-H4 (2/350)) either experienced dropout or had sequencing depths below 30× in single or double pools, rendering the MPS-STR typing for these loci invalid and excluding them from the statistical analysis. Ninety-two random samples were genotyped using two commercial CE-based STR kits and concordance between two methods was confirmed (alleles which were dropout due to quality control were not taken into consideration). Variations in the flanking regions A total of 58 variations in the flanking regions (53 SNPs/SNVs and 5 InDels) were observed at 25 loci in 350 samples. The majority of these variations (48/58) were distributed in A-STRs. With the exception of CSF1PO, D3S1358, FGA, and D12S391, variations were observed in the flanking regions. Among these loci, D7S820 exhibited the highest number of variations (seven), while four variations each were observed in the flanking regions of D1S1656, TH01, and TPOX. Other variations (10/58) were distributed in 8 Y-STRs, of which two variations were observed in DYS437 and DYS438, respectively. Among 58 variations in the flanking regions, 7 were not recorded in the Single-Nucleotide Polymorphism database (dbSNP, https://www.ncbi.nlm.nih.gov/snp/ ). All variations were verified using Sanger sequencing (data not provided) and were listed in Supplementary Table . Diversity of observed alleles For all 350 samples, all A-STR alleles (14,700) and a total of 7,301 Y-STR alleles were identified (49 Y-STR alleles were excluded due to quality control measures or dropout). The implementation of MPS analysis resulted in a significant enhancement in allele diversity. The magnitude of this enhancement varied across different loci in terms of the number of unique alleles observed. For the A-STRs, a total of 228 unique LB alleles, 348 unique RSB alleles, and 480 unique FSB alleles were observed. The application of MPS resulted in an increase in the number of identified unique alleles for all A-STRs compared to CE. Notably, twelve loci exhibited increased allele diversity due to variations in both core repeat and flanking region sequences (D1S1656, D2S1338, D2S441, D5S818, D6S1043, D7S820, D8S1179, D13S317, D18S51, D19S433, D21S11, and vWA), while four loci showed increased allele diversity solely due to variations in the core repeat regions (D3S1358, D12S391, FGA, and CSF1PO) and five loci displayed an increase in allele diversity only due to variations in the flanking regions (D10S1248, D16S539, D22S1045, TH01, and TPOX). It was found that D2S1338 and D21S11exhibited the highest levels of genetic diversity with 49 unique alleles each, whereas CSF1PO showed relatively low levels with only eight unique alleles observed. D2S1338 demonstrated the highest growth rate (276.92%), whereas D18S51 exhibited the lowest growth rate (5.88%). The comparison of unique alleles per A-STR locus among LB, RSB, and FSB was conducted, as depicted in Table A and Fig. A. For the Y-STRs, a total of 148 unique LB alleles, 258 unique RSB alleles, and 271 unique FSB alleles were observed. The increase in diversity for Y-STRs was comparatively smaller than that observed for the A-STRs. Approximately 61.9% of Y-STRs (13/21) exhibited an augmented allelic diversity. Among these loci, seven Y-STRs demonstrated enhanced allele diversity due to variations in both core repeat and flanking region sequences (DYS385ab, DYS390, DYS437, DYS438, DYS447, and GATA-H4). Four Y-STRs (DYS389II, DYS392, DYS448, and DYS635) displayed an increase in allele diversity exclusively owing to differences in core repeat region sequences, whereas two Y-STRs (DYS391 and DYS570) showed enhanced diversity solely due to variations in the flanking regions. For the remaining eight Y-STR loci (DYS19, DYS389I, DYS439, DYS444, DYS456, DYS533, DYS549, and DYS576), no variation was detected either in the core repeat or flanking regions indicating a lack of increase in allele diversity. Notably, among all Y-STRs, DYS447 exhibited not only the highest level of diversity with 53 unique alleles but also showcased a remarkable growth rate of 562.50%. The comparison of unique alleles per Y-STR locus among LB, RSB, and FSB was conducted, as depicted in Table B and Figure B. Allele frequencies and forensic parameters The frequencies of LB, RSB, and FSB alleles were determined for all STR loci in 350 individuals using the counting method. The results are summarized in Supplementary Tables and . Following Bonferroni correction (α = 0.05/21), all A-STRs allele data demonstrated adherence to HWE expectations, as indicated in Supplementary Table . Therefore, the population data obtained from this study can be considered representative. The A-STR loci were analyzed using STRAF software, and the forensic parameters (as described in the section) are presented in Supplementary Table . The average GD/Hexp values for A-STRs, when analyzed based on length, were 0.7911; however, when assessed based on the repeat region sequence and accounting for sequence variation in the flanking regions, they increased to 0.8106 and 0.8346, respectively. The combined RMP for the 21 A-STRs were calculated as 6.42 × 10 –25 , 4.45 × 10 –27 , and 2.48 × 10 –29 for LB, RSB, and FSB, respectively. By solely incorporating sequence variations in the repeat regions, the combined RMP decreased significantly by more than a hundredfold compared to length-based alleles alone. Furthermore, inclusion of sequence variations in both repeat and flanking regions enabled MPS analysis to reduce the combined RMP drastically by over 2.5887 × 10 4 times compared to CE method implementation. The observed trends were consistent across other forensic parameters resulting from a significant increase in the availability of unique alleles. A total of 340 distinct Y-STR haplotypes were observed in 350 individuals. Despite an increase in allele diversity, sequence variations did not contribute to the augmentation of unique haplotypes. Out of the 340 unique Y-STR haplotypes, one-time occurrence was noted for 333 haplotypes, twice for 5 haplotypes, thrice for 1 haplotype, and four times for another single haplotype. The random match probability (RMP) was calculated as 0.0031, while the discrimination power (DP) and heterozygosity coefficient (HD) were determined as 0.9969 and 0.9998 respectively. The analysis of 350 samples involved a total of five sequencing runs. A total of 93,185,964 reads were obtained. There was a good balance among the samples, and at least 50,000 reads were obtained for each sample, indicating that the library was pooled well. After removing invalid data (reads with a coverage of less than 30× and reads of stutters) and primer dimers, the average depth of coverage (Doc) of each STR locus was 330×. The average Doc of A-STR was 456×, which was significantly higher than that of Y-STR (191×). The average Doc of pool1 was 313×, and 347× for pool2. For A-STRs, the highest average Doc locus was D5S818 pool1 (1891×) and the lowest was D21S11 pool2 (175×). For Y-STRs, the highest average Doc was DYS391 pool2 (592×), and the lowest was DYS447 pool2 (67×). The Doc of each STR locus was shown in Supplementary Table . The sequence merging was performed manually. For A-STRs, all profiles of 350 samples were obtained. The core repeat sequences of each A-STR from pool1 and pool2 of the same sample were compared and confirmed to be consistent. The minimum sequencing depth was higher than 30×. The Y-STRs exhibited a lower depth of coverage compared to the A-STRs. A limited number of Y-STR alleles (DYS19 (6/350), DYS389I-II (2/350), DYS437 (8/350), DYS447 (11/350), DYS448 (16/350), DYS533 (2/350), and GATA-H4 (2/350)) either experienced dropout or had sequencing depths below 30× in single or double pools, rendering the MPS-STR typing for these loci invalid and excluding them from the statistical analysis. Ninety-two random samples were genotyped using two commercial CE-based STR kits and concordance between two methods was confirmed (alleles which were dropout due to quality control were not taken into consideration). A total of 58 variations in the flanking regions (53 SNPs/SNVs and 5 InDels) were observed at 25 loci in 350 samples. The majority of these variations (48/58) were distributed in A-STRs. With the exception of CSF1PO, D3S1358, FGA, and D12S391, variations were observed in the flanking regions. Among these loci, D7S820 exhibited the highest number of variations (seven), while four variations each were observed in the flanking regions of D1S1656, TH01, and TPOX. Other variations (10/58) were distributed in 8 Y-STRs, of which two variations were observed in DYS437 and DYS438, respectively. Among 58 variations in the flanking regions, 7 were not recorded in the Single-Nucleotide Polymorphism database (dbSNP, https://www.ncbi.nlm.nih.gov/snp/ ). All variations were verified using Sanger sequencing (data not provided) and were listed in Supplementary Table . For all 350 samples, all A-STR alleles (14,700) and a total of 7,301 Y-STR alleles were identified (49 Y-STR alleles were excluded due to quality control measures or dropout). The implementation of MPS analysis resulted in a significant enhancement in allele diversity. The magnitude of this enhancement varied across different loci in terms of the number of unique alleles observed. For the A-STRs, a total of 228 unique LB alleles, 348 unique RSB alleles, and 480 unique FSB alleles were observed. The application of MPS resulted in an increase in the number of identified unique alleles for all A-STRs compared to CE. Notably, twelve loci exhibited increased allele diversity due to variations in both core repeat and flanking region sequences (D1S1656, D2S1338, D2S441, D5S818, D6S1043, D7S820, D8S1179, D13S317, D18S51, D19S433, D21S11, and vWA), while four loci showed increased allele diversity solely due to variations in the core repeat regions (D3S1358, D12S391, FGA, and CSF1PO) and five loci displayed an increase in allele diversity only due to variations in the flanking regions (D10S1248, D16S539, D22S1045, TH01, and TPOX). It was found that D2S1338 and D21S11exhibited the highest levels of genetic diversity with 49 unique alleles each, whereas CSF1PO showed relatively low levels with only eight unique alleles observed. D2S1338 demonstrated the highest growth rate (276.92%), whereas D18S51 exhibited the lowest growth rate (5.88%). The comparison of unique alleles per A-STR locus among LB, RSB, and FSB was conducted, as depicted in Table A and Fig. A. For the Y-STRs, a total of 148 unique LB alleles, 258 unique RSB alleles, and 271 unique FSB alleles were observed. The increase in diversity for Y-STRs was comparatively smaller than that observed for the A-STRs. Approximately 61.9% of Y-STRs (13/21) exhibited an augmented allelic diversity. Among these loci, seven Y-STRs demonstrated enhanced allele diversity due to variations in both core repeat and flanking region sequences (DYS385ab, DYS390, DYS437, DYS438, DYS447, and GATA-H4). Four Y-STRs (DYS389II, DYS392, DYS448, and DYS635) displayed an increase in allele diversity exclusively owing to differences in core repeat region sequences, whereas two Y-STRs (DYS391 and DYS570) showed enhanced diversity solely due to variations in the flanking regions. For the remaining eight Y-STR loci (DYS19, DYS389I, DYS439, DYS444, DYS456, DYS533, DYS549, and DYS576), no variation was detected either in the core repeat or flanking regions indicating a lack of increase in allele diversity. Notably, among all Y-STRs, DYS447 exhibited not only the highest level of diversity with 53 unique alleles but also showcased a remarkable growth rate of 562.50%. The comparison of unique alleles per Y-STR locus among LB, RSB, and FSB was conducted, as depicted in Table B and Figure B. The frequencies of LB, RSB, and FSB alleles were determined for all STR loci in 350 individuals using the counting method. The results are summarized in Supplementary Tables and . Following Bonferroni correction (α = 0.05/21), all A-STRs allele data demonstrated adherence to HWE expectations, as indicated in Supplementary Table . Therefore, the population data obtained from this study can be considered representative. The A-STR loci were analyzed using STRAF software, and the forensic parameters (as described in the section) are presented in Supplementary Table . The average GD/Hexp values for A-STRs, when analyzed based on length, were 0.7911; however, when assessed based on the repeat region sequence and accounting for sequence variation in the flanking regions, they increased to 0.8106 and 0.8346, respectively. The combined RMP for the 21 A-STRs were calculated as 6.42 × 10 –25 , 4.45 × 10 –27 , and 2.48 × 10 –29 for LB, RSB, and FSB, respectively. By solely incorporating sequence variations in the repeat regions, the combined RMP decreased significantly by more than a hundredfold compared to length-based alleles alone. Furthermore, inclusion of sequence variations in both repeat and flanking regions enabled MPS analysis to reduce the combined RMP drastically by over 2.5887 × 10 4 times compared to CE method implementation. The observed trends were consistent across other forensic parameters resulting from a significant increase in the availability of unique alleles. A total of 340 distinct Y-STR haplotypes were observed in 350 individuals. Despite an increase in allele diversity, sequence variations did not contribute to the augmentation of unique haplotypes. Out of the 340 unique Y-STR haplotypes, one-time occurrence was noted for 333 haplotypes, twice for 5 haplotypes, thrice for 1 haplotype, and four times for another single haplotype. The random match probability (RMP) was calculated as 0.0031, while the discrimination power (DP) and heterozygosity coefficient (HD) were determined as 0.9969 and 0.9998 respectively. Over the past decade, there has been significant advancement in sequencing technology. Various techniques, including clonal nucleic acid amplification followed by sequencing by synthesis (SBS) , nanopore sequencing , and real-time single-molecule sequencing , have laid the groundwork for novel sequencing approaches. Among these, the advantages of MPS over conventional CE methods have positioned it as a focal point in forensic genetics . MPS-based STR genotyping research has been performed in a large number of previous studies, in which the flanking region variations have been mentioned several times. Silva et al. analyzed MPS-STR data of 59 individuals from the Brazilian population by using the PowerSeq™ Auto System and 17 flanking region variations were observed, of which 16 were located in the flanking regions of 9 A-STRs and another was in Y-STR. Kim et al. reported 15 variations in the flanking regions of 8 A-STR in a Korean population involving 250 individuals. Gettings et al. used the PowerSeq™ Auto System to perform MPS-STR studies on 183 individuals from different ethnic groups and summarized all polymorphisms within 500 bp upstream and downstream of commonly used STR in forensic genetics in the 1000 Genomes Phase 3 database. It was demonstrated that a large number of population genetic data involving STR flanking region variations was needed to support MPS forensic application. Herein, we have developed this MPS system to systematically investigate the flanking polymorphisms of common STRs. Currently, the majority of MPS-STR studies have utilized commercially available MPS-STR panels, with the ForenSeq™ DNA Signature Prep Kit being widely acknowledged as the most frequently employed panel in MPS-STR analysis. It has consistently demonstrated remarkable efficacy in numerous studies. With the utilization of this commercial MPS kit, Delest et al. conducted a comprehensive MPS-STR investigation in a French population consisting of 169 individuals. Similarly, Khubrani et al. performed an extensive MPS-STR analysis on the Saudi Arabian population, while Wendt et al. carried out an insightful MPS-STR study on Yavapai Native Americans. Precision ID GlobalFiler™ NGS STR Panel v2 is another widely used commercial MPS-STR panel. Dash et al. employed this panel to present a comprehensive investigation on sequence variations, flanking region variations, and allele frequencies at 31 STRs within the Indian population. Similarly, Wang et al. conducted an investigation encompassing Chinese Tibetan and Han populations with Precision ID GlobalFiler™ NGS STR Panel v2, while Barrio et al. presented a study focused on MPS data of 31 STRs from 496 Spanish individuals. In terms of loci overlap, ForenSeq™ DNA Signature Prep Kit encompasses 41 STRs (except for DYS447) investigated in this study, whereas Precision ID GlobalFiler™ NGS STR Panel v2 includes all A-STRs examined. Compared with previous studies, the emphasis of our study was primarily on the detection of variations in the flanking regions. With the premise of ensuring that the core region was detected, two primer pairs per STR were designed to maximize the detectability of flanking region variations within the length range compatible with current MPS technology reads. Among the 58 variations observed in the flanking regions in our study, 37 were located sufficiently distal from the core repeat region to remain undetected by using ForenSeq™ DNA Signature Prep Kit, despite successful detection of STR core repeat region and proximal flanking variations. Although the detected range of Precision ID GlobalFiler™ NGS STR Panel v2 is greater than ForenSeq™ DNA Signature Prep Kit (SE400 > PE300), it should be noted that among the 48 observed variations in the flanking regions of A-STRs, 23 were located outside of the detectable range within the overlapped loci. The detail was shown in Supplementary Table . By contrast, all reported variations in the flanking regions of overlapped loci from previous studies above fell within the range of our developed MPS system (listed in Table ), although certain variants were not detected in this study due to population disparities or their low frequency. Utilizing MPS to analyze STR can not only distinguish different alleles with the same length, but also identify the variations in the core repeat regions and flanking regions. Integration of the core repeat regions with adjacent flanking regions variations enables the formation of alleles that serve as compound markers, encompassing STR-SNP/SNV/InDel variations. The MPS-STR alleles can significantly enhance the performance of STR in forensic applications by increasing allelic diversity, aiding kinship interpretation, and facilitating mixed sample separation. In this study, the diversity of observed alleles based on LB, RSB, and FSB indicated that the sequence variants had a greater impact on A-STRs and could clearly increase MPS-STR genotyping’s ability to identify individuals. The findings were in line with the reports of previous studies . However, our study revealed a higher diversity of observed alleles based on FSB compared to previous studies, primarily due to enhanced detection of variations in the flanking regions. In contrast, it appears that the impact of sequence variants on Y-STR haplotype typing was comparatively diminished in this study. Comparing the MPS and CE method, there was no increase in the number of unique Y-STR haplotypes despite of the increase Y-STR allelic diversity. Except for the SNP in the flanking region of GATA-H4 (no rs, position: ChrY: 16631756, frequency 68/350), the variations of the detected Y-STR flanking regions were generally low in the population of this study, which did not improve the ability to identify Y-STR haplotypes. The CE method currently represents the gold standard in forensic genetics . Prior to implementing MPS in forensic casework, it is imperative to ensure the compatibility of MPS data with existing CE-based forensic databases. Therefore, it is crucial to evaluate the concordance between MPS and CE. The presence of variations in the flanking regions not only poses challenges to STR genotype analysis but also contributes to occasional discordances. Although we have confirmed the concordance between our MPS system and CE method through a sample set of ninety-two randomly selected individuals in this study, sporadic discrepancies have been previously reported . In some loci, the variations in the adjacent flanking regions may cause the flanking sequence being structurally the same as the core repeat motifs. For example, in D13S317, the core repeat motif is TATC and the reference sequence (RefSeq) of 3’ flanking region adjacent to the core repeat region is AATCAATC. Two SNP (rs9546005 and rs202043589) are frequently observed so that 3’ flanking region sequence is T ATCAATC (rs9546005 was underlined) or TATC T ATC (rs202043589 was underlined). The one or two additional repeats are not involved by length-based CE typing, but may be counted by MPS if the bioinformatic algorithms only focus on the core repeat motif. In the East Asia (EAS) population from the 1000 Genomes Phase 3 dataset, the frequency of the rs9546005 T allele is 48.5%, and the frequency of the rs202043589 T allele is 5.9%, whereas the frequencies of rs9546005 and rs202043589 were 50.9% (356/700) and 6.4% (45/700), respectively, in this study. Analogously, another example in D5S818, the 5’ flanking region contains CTCT adjacent to the core repeat motif (ATCT). The variant A allele of rs73801920 can construct an additional ATCT repeat which may be counted by sequence-based typing, whereas would not be counted by CE typing. The 1000 Genomes does not include rs73801920 frequency, and in this study, the frequency of the A allele was 17.6% (123/700). Such potential discordances have been resolved by improvements in bioinformatics analysis. In addition, there were other three scenarios that could cause discordance between MPS and CE as Gettings et al. summarized: ① InDels in the flanking region: a discordance would occur if an InDel was located inside the CE amplified region but outside the MPS bioinformatic recognition sites. ② Bioinformatic null allele I: one allele would drop out in the bioinformatic pipeline due to variants in the same region as a bioinformatic recognition site, which led to the incorrect appearance of a homozygote. ③ Bioinformatic null allele II: there was not a bin in the bioinformatic configuration file that matched the observed allele. The challenges presented in Scenarios ② and ③ can be addressed through advancements in bioinformatics analysis, as discussed earlier. However, Scenario ① remains unresolved due to the current limitations of MPS read length. In this study, Among the 58 variations in the flanking regions observed, five were flanking region InDels. Especially in D18S51 and D19S433, the variations observed were all InDels. Because these InDels were both within the detection range of MPS and CE, no discordance was observed. In summary, it is very helpful for forensic researchers to understand the flanking region polymorphisms in STR analysis, since discordances between MPS and CE are mainly caused by the variations in the flanking regions. The detectable lengths of the flanking regions within the range of MPS reads primarily depend on the lengths of the core repeat regions, given that the overall length detected by MPS remains constant. The lengths of the core repeat regions of STRs vary from locus to locus. For locus with relatively short core repeat regions, such as TPOX and TH01, the range of the flanking regions that can be detected is large. The sequence variation in TPOX has rarely been reported in previous studies. In a large population study, Novroski et al. using the MiSeq FGx analyzed 777 unrelated individuals from four major population groups (US Caucasian n = 210, African American n = 200, Hispanic n = 198, and East Asian n = 169). In 27 A-STRs, Only TPOX did not have any sequence variation to be observed. It is indicated that few sequence variations are in the core repeat region or adjacent flanking regions of TPOX. In our study, a total of four variations were observed in the distal flanking region of TPOX (rs149212737, rs115970091, rs1449872726, rs13413321), in which the frequency of rs13413321 was more than half (368/700). Similarly, four variations (rs535300047, rs1472955972, rs1279211197, and rs369097987) were observed in the distal flanking region of TH01 which was rarely reported with sequence variations as well. By contrast, for locus with relatively long core repeat regions, such as FGA and D21S11, the range of the flanking regions that can be detected is limited. In our study, no variation was found in the flanking region of FGA, and only one variation (rs1051967683) was observed in D21S11. The 58 flanking region variations observed in this study could be divided into three categories: ① The variations with high polymorphism, such as rs9546005 and rs202043589 of D13S317, rs58390469 and rs79534691 of D2S441, rs1728369 and rs11642858 of D16S539. This category of variations has a great impact on MPS-STR typing and can significantly improve the identification ability of STRs, which are reflected in the forensic parameters (Supplementary Table ). ② The variations with low frequency (< 1%), such as rs1336239361 of D2S1338 and rs1030964212 of D8S1179. The increased number of alleles with this kind of flanking region variations has a negligible effect on the observed increase in heterozygosity. Although this category of variations is of limited help in improving the discriminative power of MPS-STR typing, it may play an important role in some cases such as separating mixture samples. ③ The variations in linkage disequilibrium, such as rs75219269, rs11063969, and rs11063971 of vWA. The three SNPs exhibit linkage with each other, despite the relatively high frequency of the mutant type (130/700). Furthermore, these SNPs were consistently observed within the same allele, vWA [CE14]-GRCh38-Chr12-5983868-5984169 [TAGA]3[TGGA][TAGA]3[CAGA]4[TAGA][CAGA][TAGA] 5,983,970-G; 5,984,116-T; 5,984,121-T; 5,984,134-C, and the sequence variations were also detected in the core repeat regions (Supplementary Table ). This category of variations is essentially equivalent to only one variation. The effectiveness of MPS as a supplementary tool to the conventional CE method in forensic routine work has been consistently demonstrated by numerous studies. However, there are still a few pending tasks that need to be accomplished prior to its complete implementation. In this study, we presented a multiplex MPS system designed to selectively amplify the flanking regions of STRs, aiming to investigate the potential variations in these regions. The concordance and characterization of MPS were conducted on 21 A-STRs and 21 Y-STRs in a Chinese Han population. The compatibility between MPS and CE methods was reaffirmed. A total of 58 variations, including 53 SNPs/SNVs and 5 InDels, were observed in the flanking regions. Comparisons with previous studies primarily focus on the detection range of our developed system in the overlapped STR loci. Furthermore, the analysis and comparison of allele diversity, allele frequencies, and forensic parameters per locus by LB, RSB, and FSB effectively highlighted the advantages of MPS as well as the significance of polymorphisms in flanking regions. In conclusion, revealing additional sequence variations in the core repeat and flanking regions of STRs can yield numerous advantages beyond a mere enhancement in discrimination power when utilizing these markers for direct matching or relationship calculations. The identification of variations in the flanking regions can play important rule in both MPS and CE method. For CE, it is advantageous to avoid the presence of polymorphic primer binding sites during assay design, to elucidate the underlying reasons for discordance observed among different primer sets amplifying identical STR loci, and to comprehend the origin of alleles exhibiting atypical sizes. In the case of MPS, it proves beneficial in enhancing allelic diversity, facilitating kinship interpretation, and separating mixture samples, as well as refining both amplification primer design and bioinformatic algorithm development. This study delved deeper into the genetic information of forensic commonly used STR and advanced the application of massively parallel sequencing in forensic genetics. Below is the link to the electronic supplementary material. Supplementary Material 1 |
Clinical Pathology in the Adult Sick Horse | 1ccbf221-7bd8-499d-9f1f-11f0787715c3 | 7127838 | Pathology[mh] | • Horses with acute inflammatory intestinal conditions, for example, enteritis and colitis, often present with clinical and hematologic evidence of endotoxemia, plasma volume contraction, acid-base disturbances, and electrolyte derangements. • Horses with chronic enteropathies frequently display evidence of malabsorption and protein loss, including weight loss despite good appetite, hypoproteinemia characterized predominantly by hypoalbuminemia, and blunted glucose-absorption curves. • Liver disease is common in horses but liver failure is uncommon. • Liver-specific enzymes sorbitol dehydrogenase and glutamate dehydrogenase reflect hepatocellular injury, whereas γ-glutamyltransferase indicates biliary disease. Other enzymes, such as aspartate aminotransferase, lactic dehydrogenase (hepatocellular), and alkaline phosphatase (biliary), may support the diagnosis of hepatopathy, but these enzymes are not liver-specific. • Liver function tests include conjugated and unconjugated bilirubin, ammonia, bile acids, and coagulation tests (prothrombin/partial thromboplastin times).
The gastrointestinal tract and liver comprise key components of the equine digestive system and together have important functions in metabolism, digestion, detoxification, and synthesis. Disorders of the gastrointestinal tract and liver are common in clinical practice, whereas failure of either organ system is less common. Hematologic and biochemical analysis can be helpful for identifying organ dysfunction, narrowing down the differential diagnostic list, and, in many cases, monitoring progress and response to treatment. This article details hematologic and biochemical tests that are important in the evaluation of intestinal and hepatic diseases and reviews bloodwork trends frequently observed in adult horses affected by enteropathy or hepatopathy.
Horses with acute inflammatory intestinal conditions, for example, proximal enteritis and colitis, often present with hematologic and biochemical findings suggestive of endotoxemia (leukopenia characterized by neutropenia), plasma volume contraction (increased hematocrit, high urine-specific gravity [USG], and prerenal azotemia), and electrolyte derangements (hyponatremia, hypochloremia, and hypomagnesemia). These derangements result from intestinal inflammation and mucosal barrier disruption, leading to fluid, electrolyte, and protein loss as well as endotoxin and bacterial translocation into the blood stream. Neutropenia reflects neutrophil margination and sequestration in the intestinal tract, with left shift and toxic changes commonly observed. Strong ion acidosis characterized by hyponatremia and hyperlactatemia also is common, although a hypoproteinemic alkalosis occasionally may occur. Hemoconcentration and plasma volume contraction occurs secondarily to fluid sequestration and loss via the intestines. Lower than expected total protein (especially albumin) concentration, considering the relative erythrocytosis and estimated degree of dehydration, occurs frequently in horses with acute colitis and indicates protein loss from the diseased bowel. Hypocalcemia often is observed in horses with hypoalbuminemia and reflects the high proportion of protein-bound calcium in circulation. Ionized (unbound) calcium, which better reflects physiologic calcium homeostasis, usually is normal. Clinicopathologic derangements can vary in severity between cases, depending on the degree of intestinal damage and, in 1 study, severity of electrolyte loss, hemoconcentration, and prerenal azotemia all were predictors of survival. Measurement of l -lactate concentrations in blood and/or peritoneal fluid has become an increasingly popular diagnostic and prognostic indicator in horses presented for colic and other acute intestinal disorders. Lactate is produced by mammalian cells under anaerobic conditions during global or local tissue ischemia/hypoxia. In horses presented for colic, peritoneal lactate concentrations are higher than blood lactate concentrations (sampled at the same time) in horses with surgical lesions necessitating intestinal resection and anastamosis. In practice, peritoneal fluid lactate concentrations that are twice that of blood are highly suggestive a strangulating surgical lesion. This diagnostic test can be particularly helpful in identifying strangulating lesions early in the course of disease (hours), during which horses may display signs of severe abdominal pain but still have normal hematologic and biochemical profiles. Blood lactate concentrations alone may also be of prognostic value in horses with acute intestinal disease. In one prospective study evaluating horses presented for surgical colic, higher blood lactate levels at admission and at 24 hours and 72 hours postoperatively was associated with non-survival. Markedly increased blood lactate concentrations at admission are associated with poorer outcomes in horses with large colon volvulus, and horses presenting for colitis with blood lactate concentrations that remain increased in the face of fluid therapy are less likely to survive to discharge. , In the latter cases, monitoring changes in blood lactate concentration after fluid resuscitation generally is of greater prognostic value than a single measurement. It has been suggested that hyperlactatemia in horses should be categorized by the physiologic mechanism of excessive lactate production as a means to potentially increase its utility as a diagnostic and prognostic test. Type A hyperlactatemia occurs in response to inadequate tissue perfusion and oxygenation and is observed in horses with dehydration, hypovolemia, and hypoxemia. Type B hyperlactatemia is produced by inflamed and/or ischemic tissues and is observed in horses with inflammatory or strangulating intestinal lesions. Although type A hyperlactatemia generally responds rapidly to restoration of tissue perfusion (often through volume replacement and fluid therapy), type B hyperlactatemia persists until the underlying inflammatory or ischemic condition is corrected. Many horses with inflammatory or ischemic intestinal lesions also are dehydrated and volume-contracted, and increased lactate concentrations in these patients likely represented simultaneous type A and type B hyperlactatemia. Applied clinically, these concepts support serial measurement of blood lactate as a means to identify the source of hyperlactatemia and provide useful information regarding severity and prognosis for survival. This was corroborated by a prospective observational study of horses presenting for gastrointestinal disease, in which a rapid reduction in blood lactate concentration in response to correction of dehydration and restoration of perfusion (type A) was associated with increased survival, whereas persistently increased blood lactate or lactate concentrations that increased in the face of supportive therapy (type B) were associated with more severe intestinal lesions and poorer survival outcomes. This also is true for peritoneal fluid, in which lactate concentrations that remain increased in the face of medical therapy is suggestive of a strangulating or severely inflamed lesion in horses presented for colic. Lactate concentrations may be measured in blood or peritoneal fluid using either benchtop or portable handheld analyzers, although variable results may be observed with handheld analyzers. The same instrument should be used when comparing blood and peritoneal fluid lactate concentrations in a single patient. Compared with horses, ponies presenting for gastrointestinal disease have higher blood lactate concentrations (median 2.8 mmol/L vs 1.6 mmol/L in 1 study). This difference may be explained by carbohydrate metabolism via the Cori cycle, in which blood glucose (which also was found to be higher in the study ponies) leads to the generation of lactate. Increased liver enzyme activities occasionally are observed in horses presented for acute gastrointestinal disorders and likely reflect anatomic proximity of the 2 organ systems and direct communication via the biliary system and portal circulation. A retrospective study examining horses with colic observed that increased g-glutamyltransferase (GGT) activity was observed in 49% of horses with right dorsal displacement but only 2% of horses with left dorsal displacement of the large colon, a finding attributed to extrahepatic biliary obstruction from bile duct compression by the displaced colon. Increased liver enzyme activities (GGT, alkaline phosphatase [ALP], and aspartate aminotransferase [AST]) also have been reported in horses with proximal enteritis, which is thought to be due to hepatic injury secondary to ascending enteric bacteria from the common bile duct, absorption of endotoxin from the portal circulation, and/or hepatic hypoxia from systemic inflammation. Hyperammonemia with clinical signs of encephalopathy occasionally is observed in horses with acute gastrointestinal disease in the absence of concurrent liver disease. These horses present most frequently for diarrhea, colic, and neurologic signs (dullness, blindness, aimless wandering, and obtundation). Blood ammonia concentrations can range from slightly above normal to more than 1000 μmol/L (case 1, discussed later), and higher concentrations at admission were associated with nonsurvival in 1 retrospective report. Hyperammonemia of gastrointestinal origin also has been reported as a cause of neurologic signs and high fatality rates in horses infected with equine coronavirus. , Ammonia concentrations in blood samples rapidly increase with storage after collection and, therefore, special handling is required for accurate measurement. Blood should be drawn into EDTA or heparin anticoagulant tubes and centrifuged and plasma separated from red blood cells immediately. If within 1 hour of the laboratory, plasma may be chilled on ice until analysis. If analysis is greater than 1 hour from collection or the sample must be shipped, the separated plasma should be immediately frozen and shipped overnight on ice. It is imperative that the sample remain frozen until analysis, because thawed samples quickly accumulate ammonia.
Chronic enteropathies are uncommon in horses and often present clinically as weight loss, diarrhea, and/or recurrent colic. Weight loss is a consistent presenting complaint, reported in 78% of horses in 1 retrospective study. Differential diagnoses for chronic enteropathy include inflammatory bowel disease and alimentary lymphoma, both of which can affect the small intestine, large colon, or both, as well as parasitism, salmonellosis, sand enteropathy, and right dorsal colitis (RDC), which primarily affect the large colon. Although intestinal biopsies (duodenal and/or rectal) can be helpful for determining the nature and extent of intestinal involvement in some horses, , clinical signs and serum biochemistry results also can provide clues as to the nature and severity of disease. Horses with chronic colonic disease often have impaired water resorption and present with chronic diarrhea. In 1 retrospective study, the clinicopathologic abnormalities detected most frequently in horses with chronic diarrhea included neutrophilia, hypoalbuminemia, hyperglobulinemia, and increased ALP activity. Clinical signs of dehydration and endotoxemia typically observed in acute colitis cases are less common, and many horses with chronic enteropathies are able to compensate for excessive fecal water loss and maintain normal or nearly normal clinicopathologic profiles. Hypoproteinemia predominantly characterized by hypoalbuminemia is a common finding in horses affected by chronic enteropathy. The severity of hypoproteinemia and hypoalbuminemia may be of prognostic value and was found positively correlated with nonsurvival in a retrospective study examining horses with weight loss despite good appetite. RDC, a complication of nonsteroidal anti-inflammatory drug (NSAID) treatment, is associated with particularly marked protein loss, often resulting in plasma protein concentrations less than 5.0 g/dL and albumin concentrations less than 1.5 g/dL. Although any NSAID is thought to be capable of causing RDC, phenylbutazone, a nonselective cyclooxygenase inhibitor, frequently is implicated. Prolonged administration of phenylbutazone, at 8.8 mg/kg, orally every 24 hours, to 12 healthy adult horses resulted in consistent hypoalbuminemia in 1 clinical trial. Neutropenia also was observed (likely due to marginalization and sequestration in inflamed sections of bowel), and 2 horses developed clinical colitis. The combination of phenylbutazone and flunixin increased the risk for ulcerative damage to the intestinal tract and created severe gastric ulceration and fatal colitis in 1 prospective study. Serial monitoring of albumin concentrations is advisable in horses receiving NSAID administration and a diagnosis of RDC should be strongly considered in horses that develop hypoproteinemia and hypoalbuminemia during treatment. This diagnosis is supported further by observing localized right dorsal colon wall thickening on transabdominal ultrasound . In addition to routine hematologic and biochemical testing, the oral glucose absorption test can support a diagnosis of chronic protein-losing enteropathy. This test is simple to perform, requires no specialized equipment, and can be performed stall-side. In 2 retrospective studies, abnormal glucose absorption was demonstrated in 70% of horses with inflammatory bowel disease and in 57% of horses with chronic diarrhea. To perform an oral glucose absorption test 1. Fast horse for 12 hours to 18 hours. 2. Measure blood glucose. 3. Administer glucose at 1-g/kg body weight as a 20% solution to the unsedated horse via nasogastric tube. 4. Measure blood glucose every 30 minutes for 2 hours, then every hour for 4 hours. Accurate serial blood glucose measurements can be obtained easily stall-side using point-of-care glucometers calibrated specifically for horses. In normal horses, blood glucose concentrations should rise to higher than 185% of baseline by 120 minutes post–glucose administration and should return to normal by 6 hours. Horses with malabsorptive enteropathy display a delayed rise in blood glucose and diminished peak concentrations compared with normal horses.
Liver disease may result from toxic, infectious, hypoxic, neoplastic, vascular, or metabolic causes. Liver disease is detected most commonly by measuring activity of liver-specific enzymes in serum or plasma. Increased hepatic enzyme activity often is a result of secondary liver disease from toxemia, hypoxia, and so forth, and hepatic function remains normal in most horses with these disorders. Primary liver disease most commonly occurs from toxic, infectious, or metabolic causes and may progress to loss of function and clinical signs of hepatic failure. Liver (hepatobiliary) failure occurs when this system has lost some or all of its functionality. Failure generally occurs when greater than 70% of hepatic function is lost and this can be determined by clinical evidence (eg, jaundice, photosensitization, and central nervous system signs) of liver failure, along with abnormal liver function tests, such as bile acids, ammonia, and so forth. It is critical to remember that liver enzyme activities are not indicators of hepatic function! Interpretation of biochemical results is always best made in combination with anamnesis and a thorough clinical examination. Clinical examination, biochemical test results, ultrasonographic findings, and, if indicated, liver biopsy are best used in combination to determine the importance of the liver disease, possible causes, proper treatments, and prognosis.
Biochemical testing is imperative when attempting to diagnose liver disease or liver failure. From a clinical perspective, biochemical results can be helpful in narrowing the differential diagnoses for liver analyte changes and, when evaluated over time, help predict prognosis. Biochemical enzyme testing, especially GGT activity, also can be used to identify subclinical hepatotoxin exposure, such as during outbreaks of pyrrolizidine alkaloid toxicity. Enzyme testing can be useful in determining treatment duration, for example, serial GGT measurements to determine duration of antimicrobial treatment of bacterial cholangiohepatitis. Equine liver-specific enzymes include sorbitol dehydrogenase (SDH), glutamate dehydrogenase (GLDH), and GGT, which reflect hepatocellular injury (SDH and GLDH) and cholestasis, biliary necrosis, or hyperplasia (GGT), respectively. AST, lactic dehydrogenase (LDH), and ALP also reflect hepatocellular (AST and LDH) and biliary (ALP) disease, but these enzymes are not liver-specific , , . Increased activities of SDH, GLDH, and AST occur with even mild acute hepatocellular injury, and the magnitude of the enzyme increase may not correspond to the functional status of the liver. SDH is released from the cytosol of the hepatocyte and has a short half-life (approximately 12 hours). Thus, repeated SDH measurements can be helpful in determining resolution or progression of acute hepatocellular disease. The clinical use of SDH measurements for detection of liver disease is affected by its instability in shipped or nonfrozen stored samples. Samples that are refrigerated may be relatively stable for up to 24 hours. GLDH is located in the mitochondria of hepatocytes, and activities are abnormally high in many horses with acute hepatocellular disease. The calculated sensitivities of increased GLDH activity for the detection of hepatic necrosis and of hepatic lipidosis were 78% and 86%, respectively, in 1 study. GLDH is more stable and has a slightly longer half-life than SDH (see ). The improved stability of GLDH makes it a recommended test for detecting acute hepatocellular disease when sample shipping is required. Horses with severe chronic fibrosis (cirrhosis) occasionally may have SDH and GLDH activities within normal reference intervals. GGT is an excellent screening test for hepatic disease in the horse. In the authors’ experience, it is rare that a horse with moderate to severe liver disease does not have increased GGT activity. Increases in GGT activity also are highly specific for liver disease because diseases in other body organs (kidney and pancreas) that contain GGT do not result in abnormal serum or plasma GGT activity. GGT activity may continue to increase for several days after an acute hepatic insult has resolved, presumably due to biliary hyperplasia. Although the greatest increase in GGT activity is seen with biliary disease, small amounts of GGT can be released after hepatocellular injury. If multiple horses stabled together have increased GGT activity, toxic causes should be considered. In horses with hepatic disease, relative increases in hepatocellular versus biliary enzyme activities can be helpful when formulating a causative differential diagnostic list. For example, if GGT activity is markedly increased and SDH, GLDH, or AST activities are increased only modestly, diseases that predominantly affect the biliary system, for example, cholangiohepatitis, should be considered most likely. Conversely, if hepatocellular-derived enzyme activity is very high and GGT activity is increased only mildly, then diseases that predominantly affect hepatocytes, for example, serum hepatitis, are more likely. Although somewhat dependent on the duration of disease, many causes of severe liver disease may result in a similar increase in hepatocellular and biliary enzyme activities, for example, pyrrolizidine alkaloid toxicity and hepatic lipidosis. The magnitude of increase in hepatocellular-derived enzymes may not correspond to hepatic function, and enzyme results should be viewed as a measure of disease and not a measure of function. In addition, the magnitude of changes in hepatocellular enzymes does not determine prognosis. For example, during a 2-year farm investigation of a forage-associated hepatopathy in Europe, more than 70 weanlings, yearlings, and adults had increased GGT (up to 1000 IU/L) and GLDH (up to 1200 IU/L) activities, yet total bilirubin and bile acid concentrations remained within the reference interval in almost all of the horses and no horses demonstrated signs of hepatic failure (Divers TJ 2016, personal observation). Instead, the prognosis for horses with liver disease is best determined by function test abnormalities (discussed later), etiology, fibrosis on liver biopsy, and presence or absence of hepatic encephalopathy.
Hepatocellular enzyme activities often are increased with many systemic disorders. This likely reflects inflammatory, vascular, hypoxic, and toxic insults to the liver secondary to the primary disorder and, in these cases, diagnostic and therapeutic attention should focus on the primary disease. Bile acid concentrations, which generally are considered a liver function test, can be increased in some horses with intestinal disorders, such as colic, enteritis, and equine dysautonomia. Moderate to markedly increased bile acid concentrations in horses with colic are associated with a guarded prognosis. Horses with displacement of the left colon to the right occasionally have increases in GGT activity along with increased concentrations of direct (conjugated) bilirubin and bile acids, resulting from obstruction of bile flow. These horses have an excellent prognosis after correction of the displacement. A small number of racehorses may have moderate increases in GGT (50–140 IU/L) activity with either no or only mild increases in other liver-derived enzyme activity, including ALP. The serum/plasma GGT activity generally remains in the 50-IU/L to 140-IU/L range for weeks in these horses if kept in work. Studies have demonstrated that GGT activity is correlated to cumulative training load and racing frequency and considered a maladaptation to training. , , , Oxidative stress has been hypothesized as a cause. The incidence of this increased GGT activity in racehorses in 1 study was 18%, but in some stables it may be higher. This abnormality has not been proved to affect performance, although many trainers believe there is a correlation between the high GGT syndrome and reduced performance.
Liver function tests become abnormal only after 70% or more of hepatic function is lost. Liver function test include increased direct (conjugated) and indirect (unconjugated) bilirubin concentrations, ammonia and bile acid concentrations, and coagulation tests, such as the prothrombin and activated partial thromboplastin times. , , In the authors’ experience, an increase in conjugated bilirubin above the normal upper limit of the reference interval is a common finding in horses with liver failure. When the abnormally high conjugated bilirubin concentration comprises 25% or more of the total bilirubin concentration, this is suggestive of a predominant biliary and obstructive disease. Increases in conjugated bilirubin (which is water-soluble) result in bilirubinuria, which may be detected by urine test strips or observing green-colored bubbles after shaking the urine. Rarely, a horse without liver disease has a positive bilirubin reading on the urine test strip. Increased unconjugated bilirubin concentration is a moderately sensitive test for liver failure but lacks specificity because increases also may occur with anorexia and hemolysis or, on rare occasions, may be seen in a healthy horse. The latter condition may be caused by a congenital deficiency in glucuronyl transferase, and affected horses can maintain total bilirubin concentrations of 9 mg/dL or greater. Bile acid concentrations above 20 μmol/L are a good predictor of liver failure. , Milder increases (up to 20 μmol/L) may occur in a few horses without hepatic disease that are anorexic for 2 or more days. Horses with chronic liver disease and persistently increased bile acid concentrations greater than 20 μmol/L have a guarded to poor prognosis. , Bile acid concentrations should not be used as a predictor of prognosis in horses with acute liver disease. In states of negative energy balance, triglyceride concentrations frequently are increased in horses but hepatic lipidosis resulting in liver failure rarely occurs unless visible lipemia is noted. Therefore, high triglyceride concentrations alone should not be used to diagnose hepatic lipidosis and liver failure. Albumin concentrations rarely are low in horses with acute (6%) or chronic (18%) liver failure, and hypoalbuminemia is neither a sensitive nor specific test for liver failure in the horse. Conversely, globulins are increased in 48% of horses with liver failure. The albumin-to-globulin ratio is more likely to be low in horses with chronic versus acute liver disease and failure. Although an inconsistent finding, urea nitrogen concentrations may be low with liver failure, presumably due to decreased synthesis in the urea cycle. Clotting times often are increased in horses with liver failure due to insufficient hepatic synthesis of clotting factors II, V, VII, IX, X, XI, and XII. Coagulation abnormalities may not be detected in some horses with liver failure, even in those with obstructive biliary disease and failure. This is somewhat surprising considering the importance of bile acids in the absorption of vitamin K and the importance of vitamin K in synthesis of activated coagulation factors II, VII, IX, and X, along with the inhibitors proteins C and S. Regardless, clinical bleeding is uncommon and liver biopsies can be performed safely in most cases. One explanation for the safety of liver biopsy in horses with fulminant hepatic disease could be that platelet counts often remain within reference intervals in most horses with liver failure. Fibrinogen, an acute-phase protein made in the liver, usually is normal or mildly decreased in horses with acute or chronic liver failure, except in horses with cholangiohepatitis, where it may be high secondary to inflammation.
Lactate concentrations frequently are high and bicarbonate concentrations are usually low in horses with fulminant hepatic failure. The high lactate concentration likely is due to a combination of decreased hepatic clearance and increased production from hemodynamic alterations found with hepatic failure and likely responsible for the low bicarbonate concentration. Glucose concentrations often are surprisingly normal in most adult horses with hepatic failure but, in some cases, glucose may be very low. , Hematocrit, iron concentrations, and percentage iron saturation occasionally are high in horses with severe liver disease, in particular those with acute necrosis. The erythrocytosis can persist despite adequate rehydration. These clinicopathologic findings should not be interpreted as iron toxicity because that diagnosis can be confirmed only by histologic evidence of hemochromatosis.
A 19-year-old Appaloosa gelding was examined because of an acute onset of diarrhea and fever. Heart rate was 56 beats per minute, mucous membranes were abnormally red, and capillary refill time was 5 seconds. Blood samples were submitted for hemogram, biochemical profile, lactate concentration, and blood polymerase chain reaction (PCR) testing for Neorickettsia risticii, along with fecal testing for other common enteric infectious agents. Initial treatment included intravenously administered crystalloids and oxytetracycline. Supportive treatment with misoprostol and di-tri-octahedral smectite (Bio-Sponge Platinum Performance, Buellton Calif. USA) administered orally and flunixin meglumine administered intravenously, and distal limb cryotherapy occurred within 1 hour of hospital admission. The horse had a good clinical response to treatment over the first 18 hours but on day 2 developed acute neurologic signs, which included circling, head pressing, and ataxia. The ammonia concentration was markedly increased and treatment with orally administered lactulose and intravenously administered mannitol was initiated. Commercial equine plasma and a synthetic colloid also were administered intravenously on days 3 and 4, respectively. The blood PCR for N risticii was positive. The horse made a full recovery and was discharged from the hospital after 5 days. Case Discussion Hyperammonemia may develop in a small number of horses with acute gastrointestinal disease. The magnitude of the hyperammonemia is somewhat unique to the horse and may be related to microbiome changes in the gut (increased amounts of ammonia-producing bacteria) and/or increased intestinal permeability. In these cases, neurologic signs develop quickly and may lead to death in less than 24 hours, although some horses may have a rapid (<48 hours) decrease in ammonia concentrations and complete recovery if the primary intestinal disease resolves. The authors are not aware of a horse with ammonia concentrations this high that survived. Interpretation of clinical pathologic data • HCT of 66%, due to hypovolemia from gastrointestinal fluid losses (dehydration) • Inflammatory leukogram: the most common leukogram findings in acute severe colitis is leukopenia, due to a neutropenia with a left shift and concurrent toxic change in neutrophils. Not all horses have neutropenia, as observed in this case. A mild monocytosis is a commonly observed feature of N risticii infection in horses. • Lower than expected total protein (especially albumin) concentration considering the relative erythrocytosis and estimated degree of dehydration. This combination occurs frequently in horses with acute colitis and indicates protein loss from the diseased bowel. The marked decrease in total protein and albumin concentrations between days 1 and 3 also is common in horses with colitis due to ongoing protein-losing enteropathy. The resultant decrease in colloid osmotic pressure can make crystalloid therapy less effective in maintaining intravascular volume because the administered crystalloid fluids tend to shift more rapidly out of the intravascular space. • Severe prerenal azotemia, which largely resolved with appropriate fluid therapy. • Hyponatremia and hypochloremia are both common findings with acute colitis in horses. In this horse, the measured decrease in the negatively charged ions, chloride (change of −31 mEq/L) and albumin (−0.6 g/dL or −2.0 mEq/L) , were greater than the decrease in the positively charged sodium (change of −23 mEq/L) , indicating that other negatively charged ions are likely increased. In this horse, the bicarbonate concentration was also very low (change of −18 mEq/L) and l -lactate concentration was very high, indicating a metabolic acidosis due to l -lactate. Other unmeasured anions, such as d -lactate or acids accumulating from the severe prerenal azotemia, also may have been present to help explain both the strong ion difference and the metabolic acidosis. • Hyperlactemia often is present in horses with acute severe colitis as a result of hypovolemia and endotoxin/cytokine effects on global tissue perfusion (type A) with additional lactate production from the local damage to the bowel wall (type B). This horse had an excellent initial response to treatment and lactate concentrations decreased quickly following fluid therapy. Horses that do not have substantial decreases in lactate concentrations after fluid resuscitation have a more guarded prognosis. ,
Hyperammonemia may develop in a small number of horses with acute gastrointestinal disease. The magnitude of the hyperammonemia is somewhat unique to the horse and may be related to microbiome changes in the gut (increased amounts of ammonia-producing bacteria) and/or increased intestinal permeability. In these cases, neurologic signs develop quickly and may lead to death in less than 24 hours, although some horses may have a rapid (<48 hours) decrease in ammonia concentrations and complete recovery if the primary intestinal disease resolves. The authors are not aware of a horse with ammonia concentrations this high that survived. Interpretation of clinical pathologic data • HCT of 66%, due to hypovolemia from gastrointestinal fluid losses (dehydration) • Inflammatory leukogram: the most common leukogram findings in acute severe colitis is leukopenia, due to a neutropenia with a left shift and concurrent toxic change in neutrophils. Not all horses have neutropenia, as observed in this case. A mild monocytosis is a commonly observed feature of N risticii infection in horses. • Lower than expected total protein (especially albumin) concentration considering the relative erythrocytosis and estimated degree of dehydration. This combination occurs frequently in horses with acute colitis and indicates protein loss from the diseased bowel. The marked decrease in total protein and albumin concentrations between days 1 and 3 also is common in horses with colitis due to ongoing protein-losing enteropathy. The resultant decrease in colloid osmotic pressure can make crystalloid therapy less effective in maintaining intravascular volume because the administered crystalloid fluids tend to shift more rapidly out of the intravascular space. • Severe prerenal azotemia, which largely resolved with appropriate fluid therapy. • Hyponatremia and hypochloremia are both common findings with acute colitis in horses. In this horse, the measured decrease in the negatively charged ions, chloride (change of −31 mEq/L) and albumin (−0.6 g/dL or −2.0 mEq/L) , were greater than the decrease in the positively charged sodium (change of −23 mEq/L) , indicating that other negatively charged ions are likely increased. In this horse, the bicarbonate concentration was also very low (change of −18 mEq/L) and l -lactate concentration was very high, indicating a metabolic acidosis due to l -lactate. Other unmeasured anions, such as d -lactate or acids accumulating from the severe prerenal azotemia, also may have been present to help explain both the strong ion difference and the metabolic acidosis. • Hyperlactemia often is present in horses with acute severe colitis as a result of hypovolemia and endotoxin/cytokine effects on global tissue perfusion (type A) with additional lactate production from the local damage to the bowel wall (type B). This horse had an excellent initial response to treatment and lactate concentrations decreased quickly following fluid therapy. Horses that do not have substantial decreases in lactate concentrations after fluid resuscitation have a more guarded prognosis. ,
• HCT of 66%, due to hypovolemia from gastrointestinal fluid losses (dehydration) • Inflammatory leukogram: the most common leukogram findings in acute severe colitis is leukopenia, due to a neutropenia with a left shift and concurrent toxic change in neutrophils. Not all horses have neutropenia, as observed in this case. A mild monocytosis is a commonly observed feature of N risticii infection in horses. • Lower than expected total protein (especially albumin) concentration considering the relative erythrocytosis and estimated degree of dehydration. This combination occurs frequently in horses with acute colitis and indicates protein loss from the diseased bowel. The marked decrease in total protein and albumin concentrations between days 1 and 3 also is common in horses with colitis due to ongoing protein-losing enteropathy. The resultant decrease in colloid osmotic pressure can make crystalloid therapy less effective in maintaining intravascular volume because the administered crystalloid fluids tend to shift more rapidly out of the intravascular space. • Severe prerenal azotemia, which largely resolved with appropriate fluid therapy. • Hyponatremia and hypochloremia are both common findings with acute colitis in horses. In this horse, the measured decrease in the negatively charged ions, chloride (change of −31 mEq/L) and albumin (−0.6 g/dL or −2.0 mEq/L) , were greater than the decrease in the positively charged sodium (change of −23 mEq/L) , indicating that other negatively charged ions are likely increased. In this horse, the bicarbonate concentration was also very low (change of −18 mEq/L) and l -lactate concentration was very high, indicating a metabolic acidosis due to l -lactate. Other unmeasured anions, such as d -lactate or acids accumulating from the severe prerenal azotemia, also may have been present to help explain both the strong ion difference and the metabolic acidosis. • Hyperlactemia often is present in horses with acute severe colitis as a result of hypovolemia and endotoxin/cytokine effects on global tissue perfusion (type A) with additional lactate production from the local damage to the bowel wall (type B). This horse had an excellent initial response to treatment and lactate concentrations decreased quickly following fluid therapy. Horses that do not have substantial decreases in lactate concentrations after fluid resuscitation have a more guarded prognosis. ,
A 5-year-old previously healthy miniature horse mare presented with acute depression, icterus, anorexia, and inability to open the jaw. The mare was diagnosed with selenium-deficient masseter myopathy with secondary negative energy balance and hepatic lipidosis. Blood analysis included hemogram, biochemical profile, and lactate concentrations . A free-catch urine sample was dark brown, with a USG of 1.025, and a urine dipstick test revealed bilirubinuria and positive heme (blood) reaction. The mare was treated with intramuscular and oral selenium and vitamin E and supported with partial parenteral nutrition and made a full recovery. The prognosis for hepatic lipidosis can be excellent regardless of the triglyceride concentration if the triggering disease is resolved promptly and proper nutritional support is provided. Interpretation of Laboratory Findings • Liver disease and failure: this mare has evidence of liver disease (increased hepatocellular and biliary enzyme activities) in addition to muscle disease (increased creatine kinase [CK] activity). The increase in total and direct bilirubin concentration along with the clinical signs and other biochemical findings support a diagnosis of liver dysfunction (failure). The increased AST activity was a result of both muscle and liver disease. ○ The marked increase in SDH (>30 times the upper reference limit) and milder increase in GGT (slightly >3 times the upper reference limit) with 11% of the total bilirubin being direct bilirubin suggest that hepatocellular injury is more severe than cholestasis. ○ The normal SDH activity on day 5 reflects both the rapid improvement in the disease and the short half-life of SDH. GGT activity is still increased on day 5 due to the longer half-life of GGT and likely from some continued biliary proliferation. • Rhabdomyolysis: increased muscle enzyme activities (CK and AST) and positive heme reaction on urine dipstick due to myoglobin, all of which improved during hospitalization. The greater decrease in CK activity during 5 days of hospitalization is due to the shorter half-life (hours) compared with AST (days). • Negative energy balance with hypertriglyceridemia: miniature horses are at increased risk for developing hypertriglyceridemia, hyperlipemia, and hepatic lipidosis in response to anorexia. The increase in circulating lipids reflects increased mobilization of fat stores as well as decreased clearance/metabolism of lipids by the liver. Treatment with intravenous dextrose, parenteral nutrition, and/or enteral nutrition often results in rapid reduction in triglyceride concentrations and resolution of hepatic lipidosis. • The acidemia with a metabolic acidosis (low venous pH and low bicarbonate), mildly increased creatinine concentration (likely prerenal azotemia), and abnormally high l -lactate concentration are likely a result of dehydration and diminished tissue perfusion, although some of the increase in l -lactate may have occurred because of decreased hepatic dysfunction/metabolism. Dehydration is further supported by the USG of 1.025. Venous pH and creatinine and lactate concentrations all normalized rapidly in response to intravenous crystalloid fluid therapy.
• Liver disease and failure: this mare has evidence of liver disease (increased hepatocellular and biliary enzyme activities) in addition to muscle disease (increased creatine kinase [CK] activity). The increase in total and direct bilirubin concentration along with the clinical signs and other biochemical findings support a diagnosis of liver dysfunction (failure). The increased AST activity was a result of both muscle and liver disease. ○ The marked increase in SDH (>30 times the upper reference limit) and milder increase in GGT (slightly >3 times the upper reference limit) with 11% of the total bilirubin being direct bilirubin suggest that hepatocellular injury is more severe than cholestasis. ○ The normal SDH activity on day 5 reflects both the rapid improvement in the disease and the short half-life of SDH. GGT activity is still increased on day 5 due to the longer half-life of GGT and likely from some continued biliary proliferation. • Rhabdomyolysis: increased muscle enzyme activities (CK and AST) and positive heme reaction on urine dipstick due to myoglobin, all of which improved during hospitalization. The greater decrease in CK activity during 5 days of hospitalization is due to the shorter half-life (hours) compared with AST (days). • Negative energy balance with hypertriglyceridemia: miniature horses are at increased risk for developing hypertriglyceridemia, hyperlipemia, and hepatic lipidosis in response to anorexia. The increase in circulating lipids reflects increased mobilization of fat stores as well as decreased clearance/metabolism of lipids by the liver. Treatment with intravenous dextrose, parenteral nutrition, and/or enteral nutrition often results in rapid reduction in triglyceride concentrations and resolution of hepatic lipidosis. • The acidemia with a metabolic acidosis (low venous pH and low bicarbonate), mildly increased creatinine concentration (likely prerenal azotemia), and abnormally high l -lactate concentration are likely a result of dehydration and diminished tissue perfusion, although some of the increase in l -lactate may have occurred because of decreased hepatic dysfunction/metabolism. Dehydration is further supported by the USG of 1.025. Venous pH and creatinine and lactate concentrations all normalized rapidly in response to intravenous crystalloid fluid therapy.
|
I primi congressi di igiene industriale a Milano | 58286989-fc8c-404c-93ca-8d0704714178 | 7809938 | Preventive Medicine[mh] | Quando ci si avvicini alla determinazione organizzativa dell’igiene industriale, almeno in Italia e nel XX secolo, ci si riferisce a realtà associative organizzatesi soprattutto negli anni Sessanta del secolo scorso (con riferimento all’Associazione Italiana Degli Igienisti Industriali) . Ciò non significa, naturalmente, che il tema dell’igiene industriale non fosse stato affrontato, in linea teorica e nelle sue estrinsecazioni pratiche, anche precedentemente. Se il tema dell’igiene industriale poteva comparire episodicamente nella letteratura scientifica nazionale ottocentesca, è verso la fine del secolo XIX, che esso si sostanzia in una dimensione inizialmente legata alle problematiche antinfortunistiche. A cavallo dei secoli XIX e XX si svilupparono in Italia nuove idee e prassi antinfortunistiche e nella realtà milanese questo problema fu affrontato da molti punti di vista. Si assistette ad una convergenza di determinazioni teoriche, applicazioni pratiche, costituzione di strutture scientifico-assistenziali, proposizione di regolamentazioni e normative, che mutarono il volto della realtà antinfortunistica: in questo contesto anche l’igiene industriale giocò un ruolo da protagonista. Si possono riconoscere alcuni passaggi significativi, che interessarono la città di Milano e indirizzano la nostra attenzione agli ultimi anni dell’Ottocento, almeno come un preciso punto di partenza. Il primo avvenimento può essere agevolmente riconosciuto nel IIIeme Congrès international des Accidents du Travail et des Assurances Sociales svoltosi a Milano dal 1° al 6 ottobre 1894 . Pochi mesi dopo, possiamo registrare l’interesse delle organizzazioni operaie per questo tema, con l’organizzazione di una specifica riunione congressuale promossa dalla Camera del Lavoro . Nel 1896 si registra la costituzione di una struttura medica d’assistenza, l’Associazione per l’Assistenza Medica negli Infortuni del Lavoro : questa istituzione sarà interessata agli sviluppi dell’igiene industriale nel periodo precedente la Prima Guerra Mondiale. Ricordati taluni interessanti apporti proposti nell’ambito degli Ingegneri e degli Architetti all’esordio del XX secolo, non dobbiamo dimenticare lo svolgimento dell’EXPO del 1906 ed il suo impulso alla tenuta di riunioni congressuali . Negli anni immediatamente prossimi alla costruzione ed attivazione della Clinica del Lavoro milanese si assistette ad una differenziazione del tema antinfortunistico sia in senso medico , sia in quello delle singole lavorazioni . Si giunse così allo svolgimento del Ier Congrès Technique International de Prévention des accidents du travail et d’Hygiène industrielle, che ebbe luogo a Milano dal 27 al 31 maggio 1912 : l’igiene industriale si dotava di un’autonomia disciplinare compiuta. Le sessioni dei Congressi Internazionali sugli Infortuni e le Assicurazioni sociali, da un lato proponevano un’organizzazione internazionale che si andava collaudando, basata sull’attività di un Bureau centrale (spesso dotato di un proprio strumento di collegamento e diffusione delle proprie elaborazioni), lo svolgimento di congressi a scadenza triennale, la disseminazione delle elaborazioni scientifiche, tecniche e assistenziali nelle varie, differenti realtà. Le sessioni milanesi del III Congresso (che si svolsero dal 1° al 6 ottobre 1894) si dimostrarono particolarmente interessanti, per il nostro assunto, perché si iniziò a delineare la necessità di un approccio tecnico, di igiene industriale: esso si fece via via più specifico ed autonomo . La dimensione internazionale del Congresso fu espressa dalla partecipazione di 747 delegati dai seguenti paesi: Italia 367; Francia 167; Impero Germanico 98; Svizzera 27; Impero Austro Ungarico 23; Belgio 20; Paesi Bassi 20; Impero Russo 8; Regno Unito 7; Stati Uniti d’America 3; Spagna 2; Brasile 1; Danimarca 1; Norvegia 1; Portogallo 1; Svezia 1. Dei quattro temi principali (le assicurazioni contro gli infortuni e le assicurazioni sociali; la prevenzione degli infortuni del lavoro; il lavoro delle donne e dei bambini; la legislazione) a noi interessa sottolineare l’analisi dei mezzi generali ed individuali di prevenzione applicabili nell’industria estrattiva, gli esempi specifici di applicazioni a grandi stabilimenti tessili come quello diretto da Ernesto De Angeli (1849-1907) a Milano ovvero l’analisi specifica di singoli mezzi di protezione individuale, quali occhiali o maschere . Se questo Congresso poteva essere definito come non particolarmente differente dal mainstream d’epoca, si deve ricordare che il mondo operaio milanese si confrontò in breve tempo con taluni temi congressuali e in forma autonoma. Nel 1895 la Camera del Lavoro milanese organizzò un congresso dedicato allo studio degl’infortuni del lavoro in rapporto all’igiene, al lavoro delle donne e dei fanciulli ed all’istruzione obbligatoria, che si tenne dal 17 al 19 marzo . Furono presenti 170 associazioni di categoria e singoli delegati, provenienti principalmente dalla Lombardia e dall’Emilia Romagna. Si trattò anche dell’occasione, per molte attiviste femminili, di venire in contatto con il tema dei rischi del lavoro e degli esiti degli infortuni: si pensi alla partecipazione di Ersilia Bronzini Majno (1859-1933), la fondatrice dell’Unione Femminile, Annetta Ferla che fu fra le fondatrici del Partito dei Lavoratori Italiani (poi Partito Socialista); Giuditta Brambilla (della Federazione femminile) della Camera del Lavoro milanese; Modesta Calcagni Rossi (dell’Unione generale femminile). Anna Kuliscioff (1855-1925) aveva preso parte anche al Congresso del 1894 . Dalle riunioni milanesi promanò anche una struttura assistenziale medica, decentrata sul territorio (posizionata in vicinanza di importanti insediamenti industriali nella parte nord-occidentale della città di Milano) : essa svolse azione antinfortunistica e anche azione di presidio chirurgico territoriale. Potremmo ora chiederci, appuntando la nostra attenzione sugli anni immediatamente scavalcanti la fine del XIX e l’inizio del XX secolo, se si possano ritrovare esempi di determinazioni igienistiche industriali in altri ambiti. Ci si imbatte negli insediamenti industriali, caratterizzati da una predominanza dell’industria tessile, grazie alla presenza del fiume Olona. La valle dell’Olona era entrata nella storia della medicina del lavoro, seppur con la denominazione storpiata di Orlanatal (Valle dell’Orlana, invece che Olonatal Valle dell’Olona) in ragione della diffusione europea degli studi sulla osteomalacia puerperale condotti da Gaetano Casati (1838-1897) nel 1871 . Entriamo allora in contatto con una figura di ingegnere che fu attivo anche nella costituzione del nuovo ospedale di Legnano: si tratta di Leonardo Sconfietti. Egli, al X Congresso degli Ingegneri ed Architetti del 1902, tenutosi a Cagliari, propose una relazione relativa alle misure per assicurare l’igiene dei lavoratori per quanto riguarda la temperatura, lo stato igrometrico e la purezza dell’aria, nei locali adibiti ad uso industriale, pur rispettando le esigenze delle varie industrie . L’interesse di Sconfietti andava, naturalmente, in maniera primaria alle industrie tessili: uno dei problemi da lui affrontati era quello dello svolgersi, in talune condizioni climatiche e stagionali di una nebbia artificiale (la cosiddetta fumana) in ragione di talune lavorazioni tessili (soprattutto quelle di candeggio, tintoria e stamperia). Un altro tema da lui affrontato, e connesso al precedente, era quello della regolamentazione dell’umidità necessaria per talune lavorazioni, in maniera non nocevole alla salute dei lavoratori. Si trattava solo di proposte teoriche? Sconfietti realizzò proprio a Legnano le sue proposte, consistenti in una particolare struttura e forma delle coperture dei reparti, atte a minimizzare la formazione di nebbie, associate ad un sistema centralizzato di riscaldamento e climatizzazione . Esse furono applicate con successo nel Cotonificio Cantoni, anche nella sede di Castellanza: si trattava dell’industria tessile promossa da Costanzo Cantoni (1800-1876) e dal figlio Eugenio (1824-1888), che all’epoca era diretta da Costanzo Cantoni jr . Il lavoro di Sconfietti si incentra sulla regolazione delle caratteristiche dell’aria, in rapporto alla sua respirabilità in particolari condizioni delle lavorazioni tessili: in linea teorica egli sostiene che l’umidità relativa al 98,5% (nebbia, fumana) consenta la respirabilità ad una temperatura inferiore ai 23°C, tuttavia sarebbe la saturazione relativa a tale condizione, espressa in g. d’acqua per m3 di aria (g. 20,383) il riferimento da tenere presente. Il problema per talune lavorazioni tessili (filande di sete, tintorie) era rappresentato dalla formazione della nebbia, mentre per altre lavorazioni (filatura e tessitura di cotone) una certa umidità era necessaria per ottenere un buon prodotto. Si può per inciso ricordare che queste caratteristiche delle lavorazioni tessili sarebbero state ancora ritenute ineludibili quasi sessant’anni più tardi ed influivano sugli interventi per il condizionamento dell’aria . Tornando al lavoro di Sconfietti, egli si proponeva di ricostituire e rendere costanti le condizioni termo-igrometriche nelle quali la nebbia non si formava, cioè quelle della stagione estiva. La proposta dell’autore era quella di predisporre un opportuno riscaldamento dell’aria: si trattava di rivoluzionare la costruzione della copertura dei reparti, in modo da rendere presente una camera d’aria isolante. Per il riscaldamento si proponeva il riciclo degli scarti delle caldaie a vapore; per la ventilazione l’uso di appositi ventilatori (nella stagione estiva essi avrebbero potuto introdurre aria fresca e secca). Per impedire l’ingresso di aria fredda nelle stagioni invernali/primaverili, si prevedeva l’ottenimento di una lieve pressione positiva all’interno dei reparti. Nel 1902 i lavori al Cotonificio Cantoni non erano ancora conclusi, ma erano in uno stato avanzato. Sconfietti analizzò anche le esperienze straniere: dal Cotton Cloth Factories Act inglese del 1889, ritenuto per alcuni versi ancora poco protettivo della salute dei lavoratori, alle ricerche condotte dalla Societé Industrielle de Moulhouse, che ritenevano insufficienti tutti i mezzi fino ad allora (1899) proposti. Sconfietti propose un sistema di umidificazione e raffrescamento basato sull’esperienza dei conduttori delle caldaie a vapore: si tratta di un sistema molto simile a quello degli attuali ventilatori ad acqua. Dobbiamo fare un passo indietro nel tempo, e riferirci ancora al 1894, anno di assoluto rilievo per la nostra trattazione. Con la fondazione dell’Associazione degli Industriali d’Italia per Prevenire gli Infortuni del Lavoro, si misero in moto differenti iniziative, a talune delle quali già si è fatto cenno. Attraverso la sua organizzazione sanitaria si elaborarono anche sussidi tecnici di prevenzione e di intervento: si pensi, per esempio, alle cassette antinfortunistiche ed ai posti di soccorso negli stabilimenti . Sotto l’egida dell’Associazione fu in seguito decisa la realizzazione di un Museo di igiene industriale: la proposta e la realizzazione avvennero in più fasi, riferibili al biennio 1904-1905. Il piccolo museo del 1904 era divenuta l’esposizione permanente del 1905: dal concetto di museo di prevenzione si era passati al più vasto ambito della sicurezza ed igiene del lavoro . L’EXPO del 1906 confermò la necessità di tale struttura educativa e ne determinò sviluppo e stabilizzazione. Purtroppo il Museo è andato perduto, ma ne possiamo ricostruire la struttura e le scelte espositive . Nel periodo compreso fra il 1906 e il 1912 alcuni avvenimenti di enorme rilevanza condizionarono l’evoluzione dei temi che ci interessano. In primo luogo deve essere ricordato l’inizio dell’attività, prevalentemente in forma ambulatoriale (1907) e la costruzione ed attivazione della Clinica del Lavoro milanese, inaugurata nel 1910, nell’edificio che tuttora la ospita . Inoltre, il fenomeno di autonomizzazione dell’igiene industriale si appalesò attraverso diverse riunioni congressuali, portatrici di caratteristiche peculiari. A questo proposito si possono ricordare il II Congresso Medico Internazionale per gl’Infortuni del Lavoro, svoltosi a Roma dal 23 al 27 maggio 1909 o quello per le Mutue Infortuni sul Lavoro in agricoltura svoltosi a Milano il 19 febbraio 1911 . In questo periodo, a proposito della pubblicistica tecnica nella quale spiccano i volumi pubblicati da Ulrich (Ulrico) Hoepli (1847-1935) , troviamo un testo di Giovanni Allevi (1870-1932), dedicato a Le malattie dei lavoratori e l’igiene industriale . Allevi era un medico di idee socialiste e fu impegnato come Consigliere comunale ed Assessore nelle giunte socialiste milanesi di inizio Novecento. Il volume di Allevi ripercorre le tematiche classiche (anche della medicina sociale) del tempo: la fatica, l’ambiente professionale, le condizioni materiali dei lavoratori, l’igiene industriale, le neuropatie e le psicopatie professionali, le tossicosi (in maniera analitica), le principali lavorazioni (anch’esse in forma analitica), il tabacco, le patologie infettive (in forma analitica), la legislazione sociale nei paesi moderni, i problemi del lavoro, le organizzazioni di classe e difesa sociale. Fra il 27 e il 31 maggio 1912 si tenne a Milano il Ier Congrès Technique International de Prévention des accidents du travail et d’Hygiène industrielle . Rispetto alle riunioni della fine del secolo XIX si erano definitivamente separati dalla discussione congressuale i temi legati alle assicurazioni sociali. Stava nell’aggettivo technique il passaggio fondamentale, che lo rendeva indissolubilmente legato alla determinazione di un ambito disciplinare autonomo per l’igiene industriale . La riunione congressuale era stata indetta dall’Associazione degli Industriali d’Italia per Prevenire gli Infortuni del Lavoro e quindi noi ritroviamo fra i partecipanti molti fra gli esponenti che fin dal 1894 si erano impegnati nello sviluppo di un moderno approccio ai temi del lavoro. Scorrendo l’elenco dei membri del Comitato Organizzatore, ritroviamo gli ingegneri Luigi Pontiggia e Francesco Massarelli, ma anche i rappresentanti di associazioni che avevano fatto e facevano la storia come l’Association Normande pour prévenir les accidents du travail, di grandi imprese industriali come la Brown-Boveri, di Musei industriali (Museum of Safety Devices and Industrial Hygiene di New York; Museum van Voorwerpenter Voor koming van Ongelukken in Fabricken di Amsterdam). L’esperienza dei congressi dedicati agli infortuni e alle assicurazioni sociali era ormai in via di superamento, proprio perché i problemi tecnici richiedevano una trattazione propria, in un contesto nel quale la presenza dell’igiene industriale appariva ineludibile. Non ci stupiamo quindi che il Congresso di Milano si dovesse occupare unicamente di questioni tecniche, tralasciando le regolamentazioni, le assicurazioni sociali, la dimensione medica, proprie di altri congressi e riunioni disciplinari. Lo schema, tuttavia, ripercorreva le positive esperienze congressuali del passato: la dimensione internazionale, la proposizione di relazioni generali, la possibilità di comunicazioni integrative. I temi generali dell’igiene industriale (costruzione delle officine, ventilazione, riscaldamento, illuminazione, refettori, latrine, lavatoi, docce, spogliatoi, servizio sanitario) e degli infortuni furono affrontati da Giovanni Offredi, ingegnere, Vice Direttore delle Acciaierie di Terni; molti temi particolari furono affrontati, quali le polveri . Fra i temi di analisi tecnica antinfortunistica si sottolinea anche quello relativo ai mezzi di segnalazione del pericolo. Esso è storicamente interessante non solo dal punto di vista interno (della tecnica antinfortunistica), ma anche perché risente dei grandi dibattiti scientifici d’epoca, a proposito della teoria della Gestalt e dei colori, che correlavano l’ambito medico a quello psicologico. Le proposte (uso del contrasto di fasce bianche-nere diagonali come indicatori di pericolo) furono applicate non solo all’ambito industriale, ma anche (con modifiche coloristiche) alle segnalazioni ferroviarie ad ala . Per quanto concerne il tema di nostro interesse, si devono ricordare le escursioni a Castellanza e Legnano, nelle quali fu visitato anche il Cotonificio Cantoni (con l’applicazione dei sistemi di umidificazione e condizionamento dell’aria citati) . Si segnala anche la visita allo stabilimento milanese della Società Italiana Ernesto De Angeli per l’Industria dei Tessuti Stampati (con la costituzione dei posti di Soccorso allestiti secondo i modelli dell’Associazione per l’Assistenza Medica negli Infortuni del Lavoro). Milano si dimostrava, nei primi anni del XX secolo, luogo di grande attività per lo sviluppo di tutte le discipline collegate al lavoro: eravamo ancora in quell’aureo periodo che avrebbe preceduto l’immane primo conflitto mondiale. La fiducia nel progresso, nella scienza, nel bene fare prometteva risultati ed infondeva speranze che da lì a poco tempo sarebbero state spazzate via. Per la storia dell’igiene industriale, quegli anni furono fondamentali per la piena maturazione disciplinare autonoma. l'autore non ha dichiarato alcun potenziale conflitto di interesse in relazione alle materie trattate nell’articolo |
Determination of drug-related problems in the hematology service: a prospective interventional study | 616c2d36-f038-4463-8e09-797be9432646 | 11067252 | Internal Medicine[mh] | Hematological malignancies include a variety of diseases such as Hodgkin lymphoma, non-Hodgkin lymphoma, leukemias, and multiple myeloma . New treatment strategies were developed for all these diseases and the survival time of patients was increased . Hematological cancer patients require combination therapy using a variety of antineoplastic agents and supportive care medications . Polypharmacy is the use of multiple medications and is common in this patient group . Polypharmacy increases the risk of drug-related problems (DRPs) . DRPs are defined as an event or situation involving medication that interferes with desired health outcomes. DRPs include inappropriate dosage and method of administration, drug-drug interactions, drug omissions and monitoring deficiencies, and adverse drug reactions . This may fail to achieve drug therapy goals or harm the patient . It also causes prolonged hospital stay, readmission, and increased mortality . Within a multidisciplinary team, clinical pharmacists can detect and prevent DRPs early through comprehensive medication review . Clinical pharmacy services are pretty new in Turkey. Although there have been postgraduate programs (master’s degree, doctorate) related to clinical pharmacy for years, there has been a clinical pharmacy specialty program since 2018 . Only graduates of the clinical specialty program can work in public hospitals . Therefore, the number of clinical pharmacists actively working in hospitals is relatively low. The contributions of clinical pharmacists in identifying and preventing DRPs have been demonstrated in many clinical departments . However, studies on determining DRPs in patients with hematological malignancy are limited . In a study conducted in an onco-hematology and bone marrow transplant unit in Brazil , the frequency of DRPs was found to be 135 (9%). 135 interventions were performed by the pharmacist and 90% were accepted. In a study conducted in France , 552 (12.6%) DRPs were found. Medication problems were mostly related to anti-infective agents, and oncologists’ acceptance of interventions was found to be high (96%). In a study conducted in Korea , a total of 1187 DRPs were identified in 438 (23.9%) of 1836 hospitalized patients with hematological malignancy. Pharmacists’ intervention was accepted by 88.3%. In a study examining the clinical and economic impact of pharmacist interventions in an outpatient hematology-oncology department in France , a total of 1970 pharmacist interventions were performed, corresponding to an average of 3.5 pharmacist interventions/patient, and the total cost savings was €175,563. The clinical pharmacist’s cost-benefit ratio was found to be €3.7 for every €1 invested. As far as it is known, no study shows that DRPs are determined by the clinician in the hematology service in Turkey. Therefore, this study aims to determine drug-related problems by a clinical pharmacist within the multidisciplinary team in patients with a diagnosis of hematological malignancy hospitalized in the hematology services of a university hospital in Turkey. Study design This study was conducted prospectively between December 2022 and May 2023 in the hematology service of Suleyman Demirel University Research and Application Hospital in Isparta, Turkey. All patients over the age of 18 who were hospitalized in the hematology service for more than 24 h were included in the study. Only the first hospitalization of each patient was evaluated. Informed consent was obtained from all participants before they participated in the study. Ethics Committee approval was obtained from Suleyman Demirel University Faculty of Medicine Clinical Research Ethics Committee (Approval No:274, Date:28.09.2022). Setting The service where the research was conducted had 15 beds and two physicians and assistant physicians were working. There was no stem cell transplant unit in the hospital. Isparta was a small city with a population of 449,777 . The hospital and patient population where the study was conducted were smaller than the hospitals in Turkey’s metropolitan cities. Sample size The sample size was calculated based on the approximate number of patients admitted to the hematology service during the previous 6 months. With the Raosoft sample size calculator, the sample size was found to be minimum 123 with a population size of 180, 5% margin of error, 95% confidence interval and 50% distribution rate . Data collection The clinical pharmacist in the study was an academic, did not routinely work in this hospital, and was present at the hospital for this study. The clinical pharmacist performed comprehensive medication reviews of patients and provided interventions. The patients’ socio-demographic characteristics, history, diagnosis, comorbidities, medications used, laboratory test results, and interventions were recorded in the data collection form by the clinical pharmacist. The patients’ data were obtained from the hospital database, patient files, and patients. In general, interventions were made through verbal communication. UpToDate® and Sanford Guide to Antimicrobial Therapy Mobile® software were used for the interventions . The Lexicomp Drug Interactions® tool, accessed via UpToDate®, was used to identify drug-drug interactions . According to Lexicomp Drug Interactions®, drug interactions consist of five categories. A -no known interaction, B- no action required, C -monitor therapy, D- consider changing therapy, X- avoid combination. The presence of at least one of the risk levels C, D, and X was defined as a potential drug-drug interactions because it was clinically significant . Polypharmacy was defined as the use of 5 or more medications . DRPs were determined using the Pharmaceutical Care Network Europe (PCNE) 9.1 Turkish version. PCNE 9.1 has 3 primary fields for problems, 9 primary fields for causes, 5 primary fields for planned interventions, 3 primary fields for acceptance level (of interventions), and 4 primary fields for status of the problem. Problems include treatment effectiveness and safety, while reasons include drug selection, drug form dose selection, and treatment duration . Statistical analysis Statistical analysis was performed using SPSS 20. Continuous variables were expressed as median-interquartile range, and categorical variables were expressed as percentage and frequency. The normality of the data was analysed with the Kolmogorov-Smirnov test. The Mann-Whitney U test was used to compare continuous independent variables, and the Chi-Square test was used for categorical variables. The Pearson Chi-Square (> 25), the Continuity Correction (5–25), and the Fisher’s Exact test (< 5) were used according to the number of cases. Multiple logistic regression analysis was performed to determine the best predictor(s) which effect on the presence of DRP. Any variable whose univariable test had a p value < 0.10 was accepted as a candidate for the multivariable model along with all variables of known clinical importance. Odds ratios, 95% confidence intervals and Wald statistics for each independent variable were also calculated. A p-value smaller than 0.05 was considered statistically significant. This study was conducted prospectively between December 2022 and May 2023 in the hematology service of Suleyman Demirel University Research and Application Hospital in Isparta, Turkey. All patients over the age of 18 who were hospitalized in the hematology service for more than 24 h were included in the study. Only the first hospitalization of each patient was evaluated. Informed consent was obtained from all participants before they participated in the study. Ethics Committee approval was obtained from Suleyman Demirel University Faculty of Medicine Clinical Research Ethics Committee (Approval No:274, Date:28.09.2022). The service where the research was conducted had 15 beds and two physicians and assistant physicians were working. There was no stem cell transplant unit in the hospital. Isparta was a small city with a population of 449,777 . The hospital and patient population where the study was conducted were smaller than the hospitals in Turkey’s metropolitan cities. The sample size was calculated based on the approximate number of patients admitted to the hematology service during the previous 6 months. With the Raosoft sample size calculator, the sample size was found to be minimum 123 with a population size of 180, 5% margin of error, 95% confidence interval and 50% distribution rate . The clinical pharmacist in the study was an academic, did not routinely work in this hospital, and was present at the hospital for this study. The clinical pharmacist performed comprehensive medication reviews of patients and provided interventions. The patients’ socio-demographic characteristics, history, diagnosis, comorbidities, medications used, laboratory test results, and interventions were recorded in the data collection form by the clinical pharmacist. The patients’ data were obtained from the hospital database, patient files, and patients. In general, interventions were made through verbal communication. UpToDate® and Sanford Guide to Antimicrobial Therapy Mobile® software were used for the interventions . The Lexicomp Drug Interactions® tool, accessed via UpToDate®, was used to identify drug-drug interactions . According to Lexicomp Drug Interactions®, drug interactions consist of five categories. A -no known interaction, B- no action required, C -monitor therapy, D- consider changing therapy, X- avoid combination. The presence of at least one of the risk levels C, D, and X was defined as a potential drug-drug interactions because it was clinically significant . Polypharmacy was defined as the use of 5 or more medications . DRPs were determined using the Pharmaceutical Care Network Europe (PCNE) 9.1 Turkish version. PCNE 9.1 has 3 primary fields for problems, 9 primary fields for causes, 5 primary fields for planned interventions, 3 primary fields for acceptance level (of interventions), and 4 primary fields for status of the problem. Problems include treatment effectiveness and safety, while reasons include drug selection, drug form dose selection, and treatment duration . Statistical analysis was performed using SPSS 20. Continuous variables were expressed as median-interquartile range, and categorical variables were expressed as percentage and frequency. The normality of the data was analysed with the Kolmogorov-Smirnov test. The Mann-Whitney U test was used to compare continuous independent variables, and the Chi-Square test was used for categorical variables. The Pearson Chi-Square (> 25), the Continuity Correction (5–25), and the Fisher’s Exact test (< 5) were used according to the number of cases. Multiple logistic regression analysis was performed to determine the best predictor(s) which effect on the presence of DRP. Any variable whose univariable test had a p value < 0.10 was accepted as a candidate for the multivariable model along with all variables of known clinical importance. Odds ratios, 95% confidence intervals and Wald statistics for each independent variable were also calculated. A p-value smaller than 0.05 was considered statistically significant. This study included 140 patients. Almost half (55%) of the patients were male and the median age was 65 (55–74) years. The median length of hospital stay was 8 (5–14) days. The median number of medications used by the patients was 6 (4–7). Polypharmacy was present in 67% of the patients. Older age, longer hospital stay, presence of acute lymphoblastic leukemia, presence of comorbidities, higher number of medications used, and polypharmacy rate were statistically significantly higher in the DRP group than in the non-DRP group ( p < 0.05). Table shows the socio-demographic and clinical characteristics of the patients. At least one DRP was detected in 69 (49.3%) patients and the total number of DRPs was 152. Possible or actual adverse drug events (96.7%) were the most common DRPs. The most important cause of DRPs were drug choice (94.7%), and the highest frequency within its subcategories was the combination of inappropriate drugs (93.4%). Potential drug-drug interactions were detected in at least one C risk in 43 (30.7%) patients, at least one D risk in 11 (7.9%) patients, and at least one X risk in 6 patients (4.3%). The clinical pharmacist performed 104 (68.4%) interventions on the prescriber, of which 100 (96.15%) were accepted and fully implemented. All 120 DRPs (78.9%) were resolved, and 28 DRPs (18.4%) were not possible or necessary to be resolved. Table shows the classification of DRPs. Table shows some examples of interventions performed by the clinical pharmacist. Anticancer drugs such as venetoclax, lenalidomide, and dasatinib were examples of potential drug-drug interactions. Table shows the adverse effects that occurred. Drug-related nephrotoxicity was the most common adverse effect. Table shows the results of the multivariate logistic regression analysis: factors most predictive of the presence of DRP. Polypharmacy and length of hospitalization were the most determinant factors in differentiating the groups with and without DRP, respectively. After adjustment for other factors, the likelihood of the presence of DRP was statistically significantly 7.921 folds (95% CI: 3.033–20.689) higher in patients with polypharmacy compared to patients without polypharmacy ( p < 0.001). On the other hand, each 5-day increase in the duration of hospitalization continued to increase the likelihood of the presence of DRP by a statistically significant (OR = 1.476, 95% CI: 1.125–1.938 p = 0.005). In our study, 152 DRPs were identified and 120 DRPs were totally solved. This reveals the importance of involving the clinical pharmacist in a multidisciplinary team. The most common DRPs in our study were possible or actual adverse drug events. Since the patient population was generally elderly and cancer patients, they were exposed to polypharmacy and drug-drug interactions. Additionally, this was not surprising since the risk of exposure to possible or actual adverse drug events was high due to the anticancer medications they use . Adverse drug events varied across studies. While this rate was 28.6% in the study conducted by Kim et al. in the hematology service, it was 78.6% in the study conducted by Umar et al. in the oncology service. Since Kim et al.‘s study was retrospective, the rate of possible or actual adverse effects may have been found to be low. Additionally, although both studies used the PCNE classification system, it was not mentioned in Kim et al.‘s study which drug-drug interaction tool was used and which risk ratio for drug-drug interaction was considered clinically significant. In our study, most of the causes of DRPs were related to drug selection and their subgroup, inappropriate combination of drugs. Drug-drug interaction rates in the studies were 14.3%, 7.4%, 13.6%, and 73.2%, respectively . Differences in this rate may be due to polypharmacy rates, differences in healthcare services, and different drug-drug interaction software . Most of the potential drug-drug interactions in our study were at risk C (monitor therapy). Therefore, in some drug-drug interactions that required monitoring, only the physician was informed, and in others, intervention was recommended to the prescriber. Drug-drug interactions were mostly related to supportive medications. In our study, anticancer drugs such as venetoclax, lenalidomide, bortezomib, and dasatinib had potential drug-drug interactions. Venetoclax had potential drug-drug interactions with verapamil-trandolapril at increased risk of D. Verapamil-trandolapril is a CYP3A4 inhibitor , and concomitant use with venetoclax increases the concentration of venetoclax. It is recommended that the dose of venetoclax be reduced by 50% . Also, there was a potential drug-drug interaction at risk X (avoid combination) between dasatinib and pantoprazole. Concomitant use of these two agents decreases the concentration of dasatinib . Bortezomib had potential drug-drug interactions at risk level C with antihypertensive drugs and drugs used in the treatment of benign prostatic hyperplasia, such as tamsulosin . Bortezomib may have a blood pressure-lowering effect, so if used concomitantly with an antihypertensive drug or another drug that can lower blood pressure, the patient should be monitored for hypotension . In our study, there was also a potential drug-drug interaction between bortezomib and diltiazem at risk level C. Diltiazem, as a CYP3A4 inhibitor, may increase bortezomib concentration . The bortezomib prescribing information emphasizes that in this case, it should be monitored for toxicity and dose reduction should be made if necessary . In our study, there was a potential drug-drug interaction between lenalidomide and dexamethasone. When lenalidomide and dexamethasone are used together, venous thromboembolism prophylaxis should be considered, as the thrombogenic activity of lenalidomide may increase . Additionally, potential drug-drug interactions with antiemetics and opioid-derived analgesics were frequently observed in our study. Identifying, monitoring, and intervening when necessary, drug-drug interactions are very important in cancer patients, and clinical pharmacists have important roles in this regard . Dose selection was the second important DRP in our study. Renal dosage adjustment of drugs is significant, especially in patients who develop acute kidney injury . Even if the drugs are started at the correct dose, the dose of the drugs should be monitored and adjusted when necessary in case of liver and renal dysfunction . In our study, antimicrobials were among the drugs that required dosage adjustment according to renal function. This was due to the fact that although infectious disease physicians started antimicrobials at the correct dose, these doses were sometimes not followed up later. Drug-induced nephrotoxicity was a common adverse event in our study, similar to other studies . Also, venetoclax-related hyperuricemia, hyperkalemia and neutropenia were observed in some patients. In a study investigating the incidence of venetoclax-related toxicity risk in British Columbia, hyperkalemia and hyperphosphatemia were observed in 9 patients (27%), and hyperuricemia was observed in 7 patients (21%) . In their study by Koehler et al., venetoclax-related hyperkalemia (31%) and hyperuricemia (5%) were observed . In our study, one acute lymphoblastic leukemia patient had vincristine-induced neuropathy. Vincristine-induced neuropathy is a common side effect and its incidence is between 30 and 40% . The clinical pharmacist’s acceptance rate of the interventions was good. In general, interventions regarding renal and hepatic dosing were accepted. The clinical pharmacist did not intervene in some cases that required monitoring (for example, category C drug interactions) and only informed the physician. These were evaluated as not possible or necessary to resolve the problem. One of the strengths of the study is that the acceptability of the interventions was higher than other studies . Additionally, our study was the first study in Turkey to reveal DRPs in detail in this vulnerable patient population in the hematology service. One of the limitations of our study is that it was conducted in a single center and with a small number of patients. In addition, the clinical pharmacist in the study was an academician and did not work full-time in the hospital, but worked at certain times of the day. This may have caused some DRPs not to be determined. According to our study, a high frequency of DRPs and possible or actual adverse drug events were detected in patients. Older age, longer hospital stay, presence of acute lymphoblastic leukemia, presence of comorbidities, higher number of medications used, and polypharmacy rate were statistically significantly higher in the DRP group than in the non-DRP group According to the results of multiple logistic regression analysis, polypharmacy and length of hospital stay were the most determining factors in distinguishing between groups with and without DRP. The most common DRP was related to possible or actual adverse drug events. The most common cause of DRPs was drug selection and its subgroup, inappropriate combination of drugs. Also, our study shows the importance of including a clinical pharmacist in a multidisciplinary team in identifying and preventing DRPs in the hematology service. |
Evaluation of the E-Psycho-Oncological Short-Term Intervention “By Your Side” to Reduce Cancer-Related Distress: A Pilot Study | c8fa4874-de09-495e-9da9-f4e15836af8c | 11863238 | Internal Medicine[mh] | Cancer is one of the greatest challenges for global healthcare in the 21st century. In Germany, the number of new cases of cancer was estimated at 510 200 for the year 2022. Although the number of cancer patients is high, cancer mortality is expected to decrease due to medical advances that enable early cancer detection and treatment. Despite the improved prognosis, cancer patients are faced with psychological problems as every second cancer patient was found to suffer from elevated distress. Moreover, patients with cancer are regularly found to require psychotherapeutic support to deal with depression, anxiety, pain, fatigue and sleep disruption, which negatively affects the quality of life. , A possible solution for reducing distress in cancer patients and overcoming current barriers to psycho-oncological treatment are psycho-oncological e-mental health interventions. Frequently named barriers include the patient’s fear of getting stigmatized, discomfort to ask for help and limited access to mental healthcare. If a high level of availability is offered, psycho-oncological e-mental health interventions can have the potential to fill an important gap in quality cancer stepped care. Further, users of such technologies could benefit from the opportunity to undergo treatment anonymously and remotely. Previous e-mental health interventions, based on cognitive behavioral therapy (CBT), were found to reduce distress, anxiety, and depressive symptoms in patients with cancer. , Moreover, CBT-based interventions increased the quality of life in patients with breast cancer. Acceptance and commitment therapy (ACT), which is part of the so-called third wave of CBT, can also be an effective method for e-mental health interventions. However, existing studies regarding the effectiveness of CBT trials predominantly focus on patients with non-advanced cancer and there was weaker evidence for those screened with lower levels of distress. In addition, mindfulness-based stress reduction (MBSR) provides another effective approach to promote coping with cancer-related distress. MBSR-based e-mental health interventions can improve self-efficacy and could even reduce symptoms of anxiety and depression. , Overall, third wave interventions including ACT and MBSR were found to address coping with behavioral impact of distress rather than directly reducing distress. Most of the existing MBSR programs have an average duration of 8 weeks. MBSR based short-term interventions, which last less than 8 weeks, have also been shown to have the potential for distress reduction, even in lung-cancer-patients. In addition, a 4-week psycho-oncological e-mental health intervention to increase quality of life in patients with breast cancer demonstrated high adherence and completion rates. Thus, the effectiveness of a 4-week intervention program for patients with different types of cancer is worth exploring, given the advantages of distress reduction, potentially high adherence and easy integration into somatic cancer care, as well as into patients’ daily lives. This longitudinal intervention pilot study was conducted to gather initial evidence regarding the effectiveness of the e-psycho-oncological short-term intervention “By your Side,” which is a self-guided psycho-oncological intervention, including elements of CBT, ACT, and MBSR, targeting people with cancer, and consisting of 16 interactive modules over a period of 4 weeks. Furthermore, the study was conceptualized to assess the satisfaction with and the usability of the e-psycho-oncological short-term intervention “By your Side,” as these parameters are of great importance for its actual use and integration into routine care.
Procedure and Participants To gather initial evidence of the effectiveness of the e-psycho-oncological short-term intervention “By your Side,” a longitudinal pilot study with a pre-post design was conducted between 4 March 2022 and 4 January 2023. Participants were recruited through the study website, public events, social media, flyers and by contacting cancer-related support groups. Inclusion criteria were a diagnosis of cancer, a minimum age of 18 years, good command of the German language and Internet access. Digital informed consent was given before the start of the study. The study was approved by the ethic committee of the Medical Faculties of the University of Tübingen (293/2018BO1). In error, this trial was not prospectively registered, but we have now registered it retrospectively at the German Clinical Trial Register ( https://www.drks.de/search/de/trial/DRKS00036001 ) with the registration number DRKS00036001. E-Psycho-Oncological Short-Term Intervention ‘By Your Side’ The e-psycho-oncological short-term intervention “By your Side” is a self-guided psycho-oncological intervention, which combines effective psychotherapeutic methods of CBT, ACT, and MBSR. It aims to support patients in developing psychological resources, managing emotions and coping with stress and anxiety. The intervention is offered in the form of a web app, which is accessible via any Internet browser and any web-connected device such as laptops, smartphones or tablets. Previously downloaded material can also be used offline. The intervention includes 16 modules that are directly accessible to the patients. The large number of modules ensures that as many relevant topics as possible are covered and that patients have as much choice as possible. provides an overview of the intervention topics and how they are covered in each module. The content of the modules is presented by various media such as explanatory and expert videos, audio-guided mindfulness exercises, an interactive skills training and an individual skills box. The average engagement time per module is calculated to be 20 to 30 minutes. The duration of the e-psycho-oncological short-term intervention “By your Side” is 1 month, with a minimum of 7 modules to be completed. The study participants were not informed in advance that they had to complete at least 7 modules in order to be counted as completers. The modules start with an assessment of the patient’s current level of distress, perceived mindfulness, coping skills and a brief personal entry in a gratitude journal. Following this, there are 3 main components of psychoeducation, skills training and mindfulness practice, which are included in every module. The skills box helps patients gather specific skills, supportive information and videos during each module. At the end of the training, patients can create a personalized summary of their completed modules and receive a mindfulness exercise plan. To motivate patients to continue the training, a reward system is integrated into the app and, after completing a module, patients can choose between a yoga video, or a cooking recipe adapted to the needs of cancer patients. During the intervention, e-mail reminders are sent when participants are inactive for more than 2 days to ensure adherence to the program. Assessment Instruments and Schedule Quantitative data were collected via the online survey platform Unipark. The baseline assessment (T0) was conducted before the start of the intervention. Participants were only able to access the intervention once they had completed the initial assessment. The post-intervention assessment (T1) was conducted after 1 month and the completion of a minimum of 7 modules. The follow-up assessment (T2) took place 1 month after the completion of the intervention. This study included an assessment of sociodemographic and medical data, as well as primary, secondary, and tertiary outcomes, which were collected using validated assessment instruments and self-generated items. All psychometric instruments used were applied in the German-language versions. The assessment schedule and the respective measuring instruments are shown in . Sociodemographic and Medical Data Sociodemographic (ie, age, gender, marital status, level of education and population size of place of residence) and medical data (ie, tumor location, treatment status, mental illness, psychological and psychopharmacological treatment) were collected at the baseline (T0). The participants’ attitudes toward online interventions were assessed as a covariate at baseline (T0) with the Attitudes Toward Psychological Online Interventions questionnaire (APOI). APOI scores range from 16 to 80, with higher scores indicating more positive attitudes toward online interventions. Internal consistency of this scale was good (Cronbach’s α = .77). Primary and Secondary Outcome Measures The primary and secondary outcomes were assessed at the baseline (T0), post-intervention (T1) and follow-up (T2). The primary outcome distress was examined with the German version of the Perceived Stress Questionnaire (PSQ-20). The PSQ-20 scores range between 20 and 80, with higher scores indicating a higher level of distress. Internal consistency was excellent (Cronbach’s α = .89-.93). For secondary outcomes, depression symptoms were measured with the Patient Health Questionnaire-2 (PHQ-2). PHQ-2 scores range from 0 to 6, with higher scores indicating higher levels of depression symptoms. Internal consistency ranged between questionable and good (Cronbach’s α = .42-.79). The Generalized Anxiety Disorder Scale-2 (GAD-2) , was used to examine generalized anxiety symptoms. GAD-2 scores range from 0 to 6, with higher scores indicating higher levels of anxiety symptoms. Internal consistency ranged between questionable and good (Cronbach’s α = .39-.80). Self-efficacy was assessed with the General Self-Efficacy Scale (GSES). GSES scores range from 10 to 40, with higher scores indicating more self-efficacy. Internal consistency was excellent (Cronbach’s α = .85-.95). The Freiburg Mindfulness Inventory (FMI) was administered as a measurement of mindfulness, with scores ranging from 14 to 56. Higher FMI scores indicate a higher level of mindfulness. Internal consistency was excellent (Cronbach’s α = .85-.92). Tertiary Outcome Measures The post-intervention (T1) usability of and satisfaction with the intervention were evaluated using the System Usability Scale (SUS) and Client Satisfaction Questionnaire adapted to Internet-based Interventions (CSQ-I). The SUS scores range between 0 and 100 and the CSQ-I scores range from 8 to 32 with higher scores indicating a higher level of usability and user satisfaction, respectively. Internal consistency of CSQ-I was good (Cronbach’s α = .84), while the internal consistency of SUS was questionable (Cronbach’s α = .52). Statistical Analyses Statistical analyses were performed using R (Version 4.3.1) and RStudio. While the sum scores were calculated for the relevant scales, SUS scores were calculated based on the sum scores. Descriptive statistics were applied to examine the sociodemographic, medical and outcome measures. Participants who completed the full assessment schedule and study dropouts of were compared regarding their sociodemographic and medical characteristics, as well as the outcome measures at baseline via t -tests, χ 2 -tests, and Fisher’s exact tests. The primary and secondary outcome measures and continuous covariates were standardized before the analysis. Repeated measure analysis of covariance (ANCOVA) and post-hoc tests were conducted to determine the difference in distress (PSQ-20) between the baseline (T0), post-intervention (T1) and follow-up (T2) while age, gender, education, mental illness, and attitudes toward online interventions (APOI) were added as covariates. The variables of age and APOI were included in the model in a standardized form. A Shapiro-Wilk test did not indicate no-normality, while Mauchly’s test for sphericity showed no violation of the assumption of sphericity. Outliers were checked with the “identify_outliers” function of the R package “rstatix” and 1 outlier was detected. ANCOVAS were calculated including and excluding the relevant outlier and no differences were detected in the results. Therefore, the ANCOVA including the outlier is reported. For the secondary outcomes, repeated-measure ANOVAs and post-hoc tests were conducted, comparing depression symptoms (PHQ-2), generalized anxiety symptoms (GAD-2), self-efficacy (GSES) and mindfulness (FMI) between the baseline (T0), post-intervention (T1) and follow-up (T2). Due to the outliers and violation of the assumption of normality, additional robust ANOVAs were calculated with the R package “WRS2” to verify the results. To evaluate the usability of and satisfaction with the distributions, the mean values and sum scores for SUS and CSQ-I were examined and the P-values were adjusted for multiple comparisons. Furthermore, η² and Cohen’s d were used as effect sizes, with η²-values around 0.01, 0.06, and 0.14 considered small, medium, and large effects, and d -values of 0.2, 0.5, and 0.8 considered small, medium, and large effects, respectively.
To gather initial evidence of the effectiveness of the e-psycho-oncological short-term intervention “By your Side,” a longitudinal pilot study with a pre-post design was conducted between 4 March 2022 and 4 January 2023. Participants were recruited through the study website, public events, social media, flyers and by contacting cancer-related support groups. Inclusion criteria were a diagnosis of cancer, a minimum age of 18 years, good command of the German language and Internet access. Digital informed consent was given before the start of the study. The study was approved by the ethic committee of the Medical Faculties of the University of Tübingen (293/2018BO1). In error, this trial was not prospectively registered, but we have now registered it retrospectively at the German Clinical Trial Register ( https://www.drks.de/search/de/trial/DRKS00036001 ) with the registration number DRKS00036001.
The e-psycho-oncological short-term intervention “By your Side” is a self-guided psycho-oncological intervention, which combines effective psychotherapeutic methods of CBT, ACT, and MBSR. It aims to support patients in developing psychological resources, managing emotions and coping with stress and anxiety. The intervention is offered in the form of a web app, which is accessible via any Internet browser and any web-connected device such as laptops, smartphones or tablets. Previously downloaded material can also be used offline. The intervention includes 16 modules that are directly accessible to the patients. The large number of modules ensures that as many relevant topics as possible are covered and that patients have as much choice as possible. provides an overview of the intervention topics and how they are covered in each module. The content of the modules is presented by various media such as explanatory and expert videos, audio-guided mindfulness exercises, an interactive skills training and an individual skills box. The average engagement time per module is calculated to be 20 to 30 minutes. The duration of the e-psycho-oncological short-term intervention “By your Side” is 1 month, with a minimum of 7 modules to be completed. The study participants were not informed in advance that they had to complete at least 7 modules in order to be counted as completers. The modules start with an assessment of the patient’s current level of distress, perceived mindfulness, coping skills and a brief personal entry in a gratitude journal. Following this, there are 3 main components of psychoeducation, skills training and mindfulness practice, which are included in every module. The skills box helps patients gather specific skills, supportive information and videos during each module. At the end of the training, patients can create a personalized summary of their completed modules and receive a mindfulness exercise plan. To motivate patients to continue the training, a reward system is integrated into the app and, after completing a module, patients can choose between a yoga video, or a cooking recipe adapted to the needs of cancer patients. During the intervention, e-mail reminders are sent when participants are inactive for more than 2 days to ensure adherence to the program.
Quantitative data were collected via the online survey platform Unipark. The baseline assessment (T0) was conducted before the start of the intervention. Participants were only able to access the intervention once they had completed the initial assessment. The post-intervention assessment (T1) was conducted after 1 month and the completion of a minimum of 7 modules. The follow-up assessment (T2) took place 1 month after the completion of the intervention. This study included an assessment of sociodemographic and medical data, as well as primary, secondary, and tertiary outcomes, which were collected using validated assessment instruments and self-generated items. All psychometric instruments used were applied in the German-language versions. The assessment schedule and the respective measuring instruments are shown in .
Sociodemographic (ie, age, gender, marital status, level of education and population size of place of residence) and medical data (ie, tumor location, treatment status, mental illness, psychological and psychopharmacological treatment) were collected at the baseline (T0). The participants’ attitudes toward online interventions were assessed as a covariate at baseline (T0) with the Attitudes Toward Psychological Online Interventions questionnaire (APOI). APOI scores range from 16 to 80, with higher scores indicating more positive attitudes toward online interventions. Internal consistency of this scale was good (Cronbach’s α = .77).
The primary and secondary outcomes were assessed at the baseline (T0), post-intervention (T1) and follow-up (T2). The primary outcome distress was examined with the German version of the Perceived Stress Questionnaire (PSQ-20). The PSQ-20 scores range between 20 and 80, with higher scores indicating a higher level of distress. Internal consistency was excellent (Cronbach’s α = .89-.93). For secondary outcomes, depression symptoms were measured with the Patient Health Questionnaire-2 (PHQ-2). PHQ-2 scores range from 0 to 6, with higher scores indicating higher levels of depression symptoms. Internal consistency ranged between questionable and good (Cronbach’s α = .42-.79). The Generalized Anxiety Disorder Scale-2 (GAD-2) , was used to examine generalized anxiety symptoms. GAD-2 scores range from 0 to 6, with higher scores indicating higher levels of anxiety symptoms. Internal consistency ranged between questionable and good (Cronbach’s α = .39-.80). Self-efficacy was assessed with the General Self-Efficacy Scale (GSES). GSES scores range from 10 to 40, with higher scores indicating more self-efficacy. Internal consistency was excellent (Cronbach’s α = .85-.95). The Freiburg Mindfulness Inventory (FMI) was administered as a measurement of mindfulness, with scores ranging from 14 to 56. Higher FMI scores indicate a higher level of mindfulness. Internal consistency was excellent (Cronbach’s α = .85-.92).
The post-intervention (T1) usability of and satisfaction with the intervention were evaluated using the System Usability Scale (SUS) and Client Satisfaction Questionnaire adapted to Internet-based Interventions (CSQ-I). The SUS scores range between 0 and 100 and the CSQ-I scores range from 8 to 32 with higher scores indicating a higher level of usability and user satisfaction, respectively. Internal consistency of CSQ-I was good (Cronbach’s α = .84), while the internal consistency of SUS was questionable (Cronbach’s α = .52).
Statistical analyses were performed using R (Version 4.3.1) and RStudio. While the sum scores were calculated for the relevant scales, SUS scores were calculated based on the sum scores. Descriptive statistics were applied to examine the sociodemographic, medical and outcome measures. Participants who completed the full assessment schedule and study dropouts of were compared regarding their sociodemographic and medical characteristics, as well as the outcome measures at baseline via t -tests, χ 2 -tests, and Fisher’s exact tests. The primary and secondary outcome measures and continuous covariates were standardized before the analysis. Repeated measure analysis of covariance (ANCOVA) and post-hoc tests were conducted to determine the difference in distress (PSQ-20) between the baseline (T0), post-intervention (T1) and follow-up (T2) while age, gender, education, mental illness, and attitudes toward online interventions (APOI) were added as covariates. The variables of age and APOI were included in the model in a standardized form. A Shapiro-Wilk test did not indicate no-normality, while Mauchly’s test for sphericity showed no violation of the assumption of sphericity. Outliers were checked with the “identify_outliers” function of the R package “rstatix” and 1 outlier was detected. ANCOVAS were calculated including and excluding the relevant outlier and no differences were detected in the results. Therefore, the ANCOVA including the outlier is reported. For the secondary outcomes, repeated-measure ANOVAs and post-hoc tests were conducted, comparing depression symptoms (PHQ-2), generalized anxiety symptoms (GAD-2), self-efficacy (GSES) and mindfulness (FMI) between the baseline (T0), post-intervention (T1) and follow-up (T2). Due to the outliers and violation of the assumption of normality, additional robust ANOVAs were calculated with the R package “WRS2” to verify the results. To evaluate the usability of and satisfaction with the distributions, the mean values and sum scores for SUS and CSQ-I were examined and the P-values were adjusted for multiple comparisons. Furthermore, η² and Cohen’s d were used as effect sizes, with η²-values around 0.01, 0.06, and 0.14 considered small, medium, and large effects, and d -values of 0.2, 0.5, and 0.8 considered small, medium, and large effects, respectively.
Study Population Baseline (T0) data were collected from n = 91 individuals, n = 82 (90.1%) of whom completed the first module. Of all participants, n = 34 (37.4%) completed 7 modules in 1 month, n = 28 (30.8%) took part in the post-intervention assessment (T1). Twenty-five participants could be assessed in the 1-month follow-up (T2), while complete data of all 3 measure points were available from n = 23 (25.3%) participants, which were included in the final data analysis. Of the N = 23 participants, 82.6% (n = 19) were female. The average age was M = 55.52 (SD = 11.05) years and the age ranged between 31 and 79 years. A total of 73.9% (n = 17) of the participants had children. Overall, 34.8% (n = 8) reported a diagnosis of mental illness and 65.2% (n = 15) received psychological treatment while 21.7% (n = 5) were currently treated with psychopharmacological medications. For additional sample characteristics, see . The n = 68 participants who dropped out of the study did not significantly differ from the participants that completed the full assessment schedule regarding the sociodemographic and medical variables reported in this section (all P > .05). Further, there were no significant differences regarding baseline values of the primary and secondary outcomes covered in the next sections (all P > .05). Primary Outcome Measure: Distress (PSQ-20) The PSQ-20 scores were higher at baseline (T0, M = 52.68, SD = 16.34) than at post-intervention (T1, M = 46.96, SD = 16.79) and at the one-month follow-up (T2, M = 43.84, SD = 17.59). visualizes the PSQ-20 scores stratified by measure point. A repeated-measure ANCOVA revealed a significant effect of time on distress between measure points after controlling for age, gender, education, mental illness, and attitudes toward online interventions ( F (2,36) = 4.10, p = .025, ω 2 p = 0.02. The post-hoc tests showed that distress was significantly lower at the 1-month follow-up (T2, t (22) = 3.14, p adj = .014, d = .654) compared to the baseline (T0). There was no difference in distress between the baseline and post-intervention assessment. Secondary Outcome Measures: Depressive Symptoms (PHQ-2), Anxiety (GAD-2), Self-Efficacy (GSES), and Mindfulness (FMI) The descriptive statistics of the secondary outcome measures are reported in . Repeated-measures ANOVAs were conducted to compare secondary outcome measures between the different assessment points. There were no significant differences in depressive symptoms ( F (2, 44) = 2.73, p = .077) between measure points. Time had a significant effect on anxiety symptoms ( F (2, 44) = 10.05, p < .001, ω 2 p = 0.11). The post-hoc tests revealed that anxiety was significantly higher at baseline (T0) than at post-intervention (T1, t (22) = 2.89, p adj = .025, d = .603) and follow-up (T2, t (22) = 3.84, p adj = .003, d = .801). Further, repeated-measures ANOVA revealed a significant effect of time on self-efficacy ( F (2, 44) = 10.26, p < .001, ω 2 p = 0.09). Self-efficacy was lower at the baseline (T0) than at post-intervention (T1, t (22) = −2.88, p adj = .026, d = .600) and follow-up (T2, t (22) = −3.84, p adj = .003, d = .800). Furthermore, time did also significantly effect mindfulness ( F (2, 44) = 13.84, p < .001, ω 2 p = 0.13). Post-hoc tests showed that mindfulness was significantly lower at the baseline (T0) than at post-intervention (T1, t (22) = −3.28, p adj = .010, d = .683) and follow-up (T2, t (22) = −4.29, p adj < .001, d = .895). Due to outliers and violation of the assumption of normality, additional robust ANOVAs were calculated to verify the results of the initial comparisons. Robust ANOVAs were conducted with the R package “WRS2.” Time had a significant effect on anxiety symptoms ( F (2, 24) = 7.72, p = .004), self-efficacy ( F (2,27) = 11.38, p < .001), and mindfulness ( F (2, 28) = 19.35, p < .001). Tertiary outcome measures: Usability (SUS) and user satisfaction (CSQ-I) Regarding the usability of the e-mental health short-term intervention “By your Side,” responses to the SUS were examined. The average participant’s SUS score was M = 89.57 (SD = 8.04, Mdn = 92.50), indicating good usability. Responses to the individual items are shown in . To investigate satisfaction with the intervention, the CSQ-I was inspected. For this sample, the CSQ-I was M = 28.00 (SD = 3.03, Mdn = 28.00), indicating high satisfaction with the intervention. The responses to the items of the CSQ-I are shown in .
Baseline (T0) data were collected from n = 91 individuals, n = 82 (90.1%) of whom completed the first module. Of all participants, n = 34 (37.4%) completed 7 modules in 1 month, n = 28 (30.8%) took part in the post-intervention assessment (T1). Twenty-five participants could be assessed in the 1-month follow-up (T2), while complete data of all 3 measure points were available from n = 23 (25.3%) participants, which were included in the final data analysis. Of the N = 23 participants, 82.6% (n = 19) were female. The average age was M = 55.52 (SD = 11.05) years and the age ranged between 31 and 79 years. A total of 73.9% (n = 17) of the participants had children. Overall, 34.8% (n = 8) reported a diagnosis of mental illness and 65.2% (n = 15) received psychological treatment while 21.7% (n = 5) were currently treated with psychopharmacological medications. For additional sample characteristics, see . The n = 68 participants who dropped out of the study did not significantly differ from the participants that completed the full assessment schedule regarding the sociodemographic and medical variables reported in this section (all P > .05). Further, there were no significant differences regarding baseline values of the primary and secondary outcomes covered in the next sections (all P > .05).
The PSQ-20 scores were higher at baseline (T0, M = 52.68, SD = 16.34) than at post-intervention (T1, M = 46.96, SD = 16.79) and at the one-month follow-up (T2, M = 43.84, SD = 17.59). visualizes the PSQ-20 scores stratified by measure point. A repeated-measure ANCOVA revealed a significant effect of time on distress between measure points after controlling for age, gender, education, mental illness, and attitudes toward online interventions ( F (2,36) = 4.10, p = .025, ω 2 p = 0.02. The post-hoc tests showed that distress was significantly lower at the 1-month follow-up (T2, t (22) = 3.14, p adj = .014, d = .654) compared to the baseline (T0). There was no difference in distress between the baseline and post-intervention assessment.
The descriptive statistics of the secondary outcome measures are reported in . Repeated-measures ANOVAs were conducted to compare secondary outcome measures between the different assessment points. There were no significant differences in depressive symptoms ( F (2, 44) = 2.73, p = .077) between measure points. Time had a significant effect on anxiety symptoms ( F (2, 44) = 10.05, p < .001, ω 2 p = 0.11). The post-hoc tests revealed that anxiety was significantly higher at baseline (T0) than at post-intervention (T1, t (22) = 2.89, p adj = .025, d = .603) and follow-up (T2, t (22) = 3.84, p adj = .003, d = .801). Further, repeated-measures ANOVA revealed a significant effect of time on self-efficacy ( F (2, 44) = 10.26, p < .001, ω 2 p = 0.09). Self-efficacy was lower at the baseline (T0) than at post-intervention (T1, t (22) = −2.88, p adj = .026, d = .600) and follow-up (T2, t (22) = −3.84, p adj = .003, d = .800). Furthermore, time did also significantly effect mindfulness ( F (2, 44) = 13.84, p < .001, ω 2 p = 0.13). Post-hoc tests showed that mindfulness was significantly lower at the baseline (T0) than at post-intervention (T1, t (22) = −3.28, p adj = .010, d = .683) and follow-up (T2, t (22) = −4.29, p adj < .001, d = .895). Due to outliers and violation of the assumption of normality, additional robust ANOVAs were calculated to verify the results of the initial comparisons. Robust ANOVAs were conducted with the R package “WRS2.” Time had a significant effect on anxiety symptoms ( F (2, 24) = 7.72, p = .004), self-efficacy ( F (2,27) = 11.38, p < .001), and mindfulness ( F (2, 28) = 19.35, p < .001).
Regarding the usability of the e-mental health short-term intervention “By your Side,” responses to the SUS were examined. The average participant’s SUS score was M = 89.57 (SD = 8.04, Mdn = 92.50), indicating good usability. Responses to the individual items are shown in . To investigate satisfaction with the intervention, the CSQ-I was inspected. For this sample, the CSQ-I was M = 28.00 (SD = 3.03, Mdn = 28.00), indicating high satisfaction with the intervention. The responses to the items of the CSQ-I are shown in .
The aim of this study was to gather initial evidence regarding the effectiveness of the e-psycho-oncological short-term intervention “By your Side” that incorporates CBT, MBSR, and ACT techniques and is tailored to reduce distress in patients with cancer. Participants reported a significant reduction in distress 1 month after completing the intervention. This effect was not dependent on age, gender, education, mental illness, or attitudes toward online interventions. In addition, the patients’ anxiety symptoms were significantly lower than before the intervention while mindfulness and self-efficacy were increased. However, there was no significant effect on symptoms of depression. The participants evaluated the usability of the digital intervention as high while user satisfaction was also high. The effect of distress reduction occurred steadily after the intervention (T1) and became significant 1 month after the intervention (T2). A meta-analysis of interventions promoting resilience in cancer patients clarified that interventions with more modules and a longer duration have stronger effects on resilience and that resilience, which is related to lower distress, can improve until 1 year after an intervention. , The findings indicate that distress reduction may evolve over time. This may be due to the time participants need to integrate new strategies into their daily routine, which might take longer than the intervention itself. A Chinese study published in 2023, in which 175 participants with lung cancer took part in a 4-week guided and supervised MBSR program, showed an immediate reduction of distress after completion. This shows that distress reduction through MBSR programs seems to be feasible after completing a short 4-week intervention. In addition to increased mindfulness, this result was also assumed by the social support from family and friends which participants received. The possibility of a stronger integration of social support into “By your Side” might therefore lead to an even faster reduction of distress, which is necessary given the high rates of distress in patients with cancer. The results of our study are in line with prior research on mindfulness-based online interventions and Internet-delivered CBT as effective approaches for distress reduction in psycho-oncological patient treatment. , Another finding about the effects of the e-psycho-oncological short-term intervention “By your Side” was a decrease in anxiety, while no significant effect was observed concerning the reduction of depressive symptoms. A meta-analysis found interventions with a duration over 6 weeks to show higher effects on reducing depressive symptoms . Therefore, the relatively short intervention duration of 4 weeks might have been insufficient for achieving a significant reduction in depression. A possible solution to the problem of reconciling short-term interventions and the reduction of depressive symptoms through longer intervention periods could be provided through targeted screening. For example, this could be achieved by using screening tools for depressive symptoms and offering targeted additional intervention time if psychometric depression scores reach levels of clinical relevance, as seen in the e-mental health intervention “I Can Do.” Further, patients who used the e-psycho-oncological short-term intervention “By your Side” for 4 weeks experienced an increase in self-efficacy and mindfulness. Based on Bandura et al’s theory, self-efficacy is the conviction to organize certain actions to achieve a certain goal at the end. This coping effect can result from setting and pursuing goals as well as from developing resilience to setbacks. An increase in self-efficacy through short-term online interventions is in line with another study, which revealed a significant increase in self-efficacy in patients with breast cancer through a 6-week e-mental health intervention, which included daily MBSR home training and a weekly guided session of 2 hours. Mindfulness has been described as a non-judgmental and accepting awareness of the present moment. The result of increasing mindfulness during a 4-week MBSR intervention is in line with recent studies. , Participants of the “By your Side” evaluation, who completed 7 modules of the intervention, reported high rates of usability and satisfaction with the intervention. Moreover, most completers could imagine future users being able to quickly learn how to use the intervention. In addition, all completers would recommend the intervention to friends affected by cancer. A possible reason for the high level of satisfaction and usability might be the intervention’s patient-centric process of development. In contrast, the initial assumption was that short interventions would be associated with higher adherence. This assumption is based on the study of Wang et al, who also tested a 4-week IMBSR intervention in patients with cancer and found good results in terms of completion rate and a 90% adherence rate. The e-psycho-oncological short-term intervention “By your Side” showed a comparatively poorer adherence with a high dropout rate, so that the advantage of an increased adherence with shorter chosen intervention periods, in this case 4 weeks, cannot be confirmed for this intervention. At least the majority of all participants (90%) of the e-psycho-oncological short-term intervention “By your Side” used the intervention at least minimally by completing the first module. Potential reasons for the high dropout rate are discussed in more detail in the Section “Study Limitation.” Study Limitations Some important limitations of the present study need to be considered. Firstly, the study was not preregistered which may have introduced some biases into the results. Secondly, the most important limitation is the small study sample. Only a few participants completed the minimum amount of 7 modules within the 4-week period and were therefore included in the final analysis. By establishing a minimum of 7 completed modules, it was ensured that the content of this short-term intervention complies with the recommended period of MBSR interventions. However, the resulting higher intensity of the intervention may have led participants to drop out. One possible reason for this could be that cancer patients might have less capacity to complete such intensive training due to treatment, emotional and socioeconomical burden. Another reason for the high dropout rate might have been the presentation format of the “By your Side” intervention: The use of digital treatment offers requires a basic level of digital knowledge and skills. Technical problems or dissatisfaction might have further contributed to dropouts or non-adherence. While user satisfaction was high, possible reasons for dropouts could not be collected from participants who did not complete the assessment. Another reason for the high dropout rate could be the study participants’ unawareness of the fact that at least 7 modules had to be completed in order to be considered a completer. Further studies should therefore investigate how these utilization skills can be better supported and how technological barriers of use and additional reasons for dissatisfaction can be removed in order to ensure higher adherence to digital interventions. Moreover, the sample of subjects was selected by convenience sampling due to cost-time effectiveness. It is therefore not suitable for deriving generalized statements. As the small study sample was mostly composed of women with higher education, the study sample can cannot be considered representative. Further, adherent participants may have been more self-motivated or more capable to self-guide themselves through the modules than those who dropped out. Therefore, selection bias needs to be considered when interpreting the results. To draw further conclusions about the intervention’s effect, a similar investigation conducted with a more representative sample would be necessary and individuals with a greater need for external guidance need to received targeted support. Thirdly, due to the study design of a pilot study, no control group was implemented. Therefore, it is not guaranteed that the positive effects observed in the study can be attributed to the intervention. Although the results of the e-psycho-oncological short-term intervention “By your Side” are promising, a precise evaluation of its efficacy is not possible based on the limited data gathered in this pilot study. In order to draw more reliable conclusions about the intervention’s effectiveness, a randomized controlled trial would be required. Finally, only a single follow-up assessment was conducted after 1 month. Additional follow-ups over longer periods of time would be helpful to draw conclusions concerning long-term effects. The presented data are based on self-report. Especially the diagnosis of cancer can therefore not be objectively verified. However, participants were recruited primarily in cancer-specific contexts to address this limitation. Nevertheless, this study gives a first promising indication of the intervention’s potential and adds to the evidence supporting the use of e-psycho-oncological short-term interventions in general.
Some important limitations of the present study need to be considered. Firstly, the study was not preregistered which may have introduced some biases into the results. Secondly, the most important limitation is the small study sample. Only a few participants completed the minimum amount of 7 modules within the 4-week period and were therefore included in the final analysis. By establishing a minimum of 7 completed modules, it was ensured that the content of this short-term intervention complies with the recommended period of MBSR interventions. However, the resulting higher intensity of the intervention may have led participants to drop out. One possible reason for this could be that cancer patients might have less capacity to complete such intensive training due to treatment, emotional and socioeconomical burden. Another reason for the high dropout rate might have been the presentation format of the “By your Side” intervention: The use of digital treatment offers requires a basic level of digital knowledge and skills. Technical problems or dissatisfaction might have further contributed to dropouts or non-adherence. While user satisfaction was high, possible reasons for dropouts could not be collected from participants who did not complete the assessment. Another reason for the high dropout rate could be the study participants’ unawareness of the fact that at least 7 modules had to be completed in order to be considered a completer. Further studies should therefore investigate how these utilization skills can be better supported and how technological barriers of use and additional reasons for dissatisfaction can be removed in order to ensure higher adherence to digital interventions. Moreover, the sample of subjects was selected by convenience sampling due to cost-time effectiveness. It is therefore not suitable for deriving generalized statements. As the small study sample was mostly composed of women with higher education, the study sample can cannot be considered representative. Further, adherent participants may have been more self-motivated or more capable to self-guide themselves through the modules than those who dropped out. Therefore, selection bias needs to be considered when interpreting the results. To draw further conclusions about the intervention’s effect, a similar investigation conducted with a more representative sample would be necessary and individuals with a greater need for external guidance need to received targeted support. Thirdly, due to the study design of a pilot study, no control group was implemented. Therefore, it is not guaranteed that the positive effects observed in the study can be attributed to the intervention. Although the results of the e-psycho-oncological short-term intervention “By your Side” are promising, a precise evaluation of its efficacy is not possible based on the limited data gathered in this pilot study. In order to draw more reliable conclusions about the intervention’s effectiveness, a randomized controlled trial would be required. Finally, only a single follow-up assessment was conducted after 1 month. Additional follow-ups over longer periods of time would be helpful to draw conclusions concerning long-term effects. The presented data are based on self-report. Especially the diagnosis of cancer can therefore not be objectively verified. However, participants were recruited primarily in cancer-specific contexts to address this limitation. Nevertheless, this study gives a first promising indication of the intervention’s potential and adds to the evidence supporting the use of e-psycho-oncological short-term interventions in general.
In summary, the present longitudinal pilot study provides prelimited evidence that the e-psycho-oncological short-term intervention “By your Side” seems to be an helpful tool for reducing distress and anxiety symptoms by simultaneously improving mindfulness and self-efficacy in cancer patients. Despite the positive effects, the high dropout rate (74.7%) and small convenience sample must be taken into account. Future research should address possible barriers to adherence in a larger, more representative sample. Nevertheless, the intervention shows promising indications for a practicable implementation into the “real world” healthcare system, as usability and satisfaction were rated high. E-psycho-oncological short-term interventions such as “By your Side” are one way to provide low-threshold support to cancer patients within a short period of time that is not tied to a specific location. This could be an important advantage to address existing barriers and gaps in cancer care. The results from this longitudinal intervention study are promising and indicate a good enhancement of current treatment approaches.
|
CGF with Bio-Oss collagen as grafting materials for simultaneous implant placement after osteotome sinus floor elevation: a prospective study | 5bc6c82f-0ec1-4b89-b472-3b31c4aadf7e | 11660636 | Dentistry[mh] | Residual ridge reduction is a common and fast phenomenon in the edentulous area of the maxilla. This area undergoes a chronic, progressive, and irreversible resorption leading to critical defects in the shape and function of the jaw . The process of bone resorption in the posterior maxilla following tooth loss, coupled with renewed pneumatization of the maxillary sinus, contributes to the vertical and horizontal deficiency of the alveolar bone. These deficiencies are observed in the bilateral maxillary posterior zone, extending from the second premolar to the pterygoid plates, located at the base of the maxillary sinuses. Often, vertical bone augmentation with a sinus lift procedure is required when dental implants are considered for placement in this zone. The bone in this region is also known to have compromised quality (types 3 and 4), which can increase the implant failure rate . Failures in rehabilitation with dental implants should be anticipated when osseointegrated implants are installed in areas of poor bone quality, such as the bilateral maxillary posterior zone, particularly in cases of severe bone resorption . Identifying the most efficient implantation procedure with successful long-term outcomes remains one of the primary concerns in the implant industry. Osteotome Sinus Floor Elevation (OSFE), as introduced by Summers , has proven to be an effective procedure for the atrophic maxilla . Furthermore, OSFE is a less invasive surgical technique compared to the previously used lateral approach, offering better cellular viability around the implant and thereby reducing the risk of adverse soft tissue responses . OSFE has been widely applied both with and without grafting materials . Recently, an increasing number of studies have focused on simultaneous sinus elevation and implantation . Regardless of whether grafting material was placed, the implant could help preserve the elevated sinus membrane and achieve new bone formation in the sinus . The simultaneous placement of implants contributes to an improved survival rate of implants in the maxilla and helps preserve the alveolar crest volume . Although significant advances have been made in the rehabilitation of the atrophic maxilla using the OSFE procedure, OSFE is not without its limitations, notably the potential risk of membrane perforation. Consequently, alternative methods have been developed to improve clinical outcomes, expand the indications for use, and reduce complications . The newest approaches include: osseodensification, which uses densifying burs to enhance bone density and implant stability; Minimally Invasive Antral Membrane Balloon Elevation (MIAMBE), which employs a balloon for precise and gentle sinus membrane elevation; and the hydraulic sinus lift, which uses fluid pressure for a controlled and less invasive approach . Furthermore, ongoing research into alternative grafting materials is essential for refining OSFE and its alternative techniques, with the goal of improving bone formation, enhancing implant stability, and optimizing clinical outcomes in sinus augmentation procedures . Bio-Oss® is a deproteinated bovine-derived xenograft product, which is widely used in dental practices for ridge augmentation in humans . However, concerns about complications and long-term safety associated with bovine-derived xenografts are increasing . Since it lacks osteoinduction and cellular activity, which poses a primary challenge in the field of osteoconductive graft enrichment research, Bio-Oss® presents further concerns. Conversely, Bio-Oss Collagen®, with its 10% collagen formulation, was developed to mimic the role of collagen in promoting rapid vascularization. Recent studies have demonstrated its role in enhancing new bone formation . Collagen is favorable due to its facile application. The interaction between soft tissue and collagen grafting has been extensively described in experimental and clinical research in both animal and human models . In a human study, sockets in the molar region of the maxilla and mandible that were filled with Bio-Oss collagen demonstrated new bone synthesis after a healing period of six weeks. This approach resulted in low complication rates and effectively preserved the area for subsequent implant placement . Moreover, bone resorption was significantly less in extraction sockets filled with Bio-Oss collagen combined with early implantation compared to untreated sockets . Recently, Bio-Oss collagen has been demonstrated to achieve a high implant success rate following the OSFE procedure . Alternatively, Concentrated Growth Factors (CGFs), which are obtained from the patient's own blood through specific sequences of centrifuge speed, are considered the newest generation of platelet concentrates. Given its crucial role in bone regeneration through the provision of various growth factors , the combination of CGF with other grafting materials has gained increasing interest in the field of bone regeneration . The use of CGF in maxillary sinus augmentation improves clinical outcomes by promoting vascularization and tissue regeneration at the surgical site, enhancing both the quality and quantity of newly formed sinus bone . When combined with graft materials, CGF helps preserve vertical bone height and significantly accelerates osteogenesis, while also reducing postoperative complications and improving implant survival rates . However, research on the combination of CGF and Bio-Oss collagen is limited and has not been reported in the context of OSFE in the existing literature. The purpose of this study was to investigate whether alternative grafting materials, specifically CGF and Bio-Oss collagen, for OSFE with simultaneous implantation can ensure high implant survival rate after long-term functional loading. Early implant failure typically occurs before the prosthesis is placed, whereas late implant failure is associated with functional loading after the prosthesis has been placed. In our study, the implant survival rate was investigated 6 months after first stage surgery and one year after functional loading with final prosthesis. Moreover, investigating viable techniques for sinus floor elevation and simultaneous implant placement in cases of edentulous posterior maxilla with RBH < 5 mm is of great clinical interest. In addition, this study contributes to the evaluation of patients' postoperative experiences. The null hypothesis is that there are no differences in endo-sinus bone synthesis indicators after OSFE with simultaneous implantation using different grafting materials. The specific objective of this study was to radiographically investigate the efficiency of the combination of Bio-Oss collagen and CGF as grafting materials for OSFE and simultaneous implant placement in the severe atrophic maxilla after one year of functional loading. To our knowledge, there has been no previous research in this area.
Study design/ sample This prospective study was conducted in the Affiliated Hospital of Stomatology, Nanjing Medical University, China. This study complies with the Declaration of Helsinki and the Good Clinical Practice guidelines, and was approved by Medical Ethics Committee, School of Stomatology, Nanjing Medical University, China (Approval number: PJ2020-141–001). Informed consent for data evaluation and publishing has been obtained from all included subjects. Patients meeting our inclusion criteria underwent OSFE with simultaneous implantation from March 2021 to January 2022. A detailed medical record was established for each patient, including medical and dental history, oral examination, surgical records, details of bone graft materials, and the type of prosthetic reconstruction. Inclusion criteria for enrollment in this study were as follows: Patient whom underwent OSFE with simultaneous implantation must be at least 18 years of age; have good general health; adequate oral hygiene, indicated by a bleeding index of less than 30% and a plaque score of less than 20%; have been extracted 1 or 2 maxillary molars or pre-molars because of failure of endodontic treatment, root fracture, or after suffering from severe caries for more than 3 months. Radiographic inclusion criteria are as follows: Residual Bone Height (RBH) of the alveolar bone crest measured on CBCT at each implant site is 2–5 mm; an adequate residual alveolar ridge width for implant placement of 6 mm or more; absence of any radiographic signs of maxillary sinus pathology; and length of the inserted implants ranging from 8–10 mm with diameters of 4.1/4.8 mm (Dentium implant/ ICX implant/ MIS). Exclusion criteria were as follows: postoperative CBCT imaging lacking clarity or unclear anatomic references, or incomplete medical history due to the failure of the patient to come for follow up appointments. Using the R program (R Foundation for Statistical Computing) , at least 107 subjects were required to evaluate the association between the primary outcome and three categorial variables. This calculation assumed a medium effect size of 0.3 (measured by Cohen's d), a P -value of 0.05, and a power of 80%. Considering the dropout rate of 20%, a total of 134 subjects were needed. Surgical procedure Patients were treated according to the designed treatment plan. They were assigned randomized numbers and subsequently divided into three different groups based on those numbers. The surgery consisted of simultaneous implant placement with sinus augmentation via the Osteotome technique (OSFE) by surgeon B.SH. All patients underwent a comprehensive oral examination and received a CBCT scan prior to surgery. Pre-operative CBCT image was used to evaluate RBH and crestal bone width. Those meeting our inclusion criteria were selected for the study. For all subjects, local anesthesia was performed, a mid-crestal incision was made and flap raised, and then the implant site was generated using a pilot drill, maintaining a distance of 1 mm from the sinus floor. Next, the bone of the sinus floor was fractured into the sinus cavity, elevating the Schneiderian membrane, with a vigilant tap of the mallet. The initial sinus elevation was performed with osteotomes, gradually advancing until the final depth was achieved. In the first group (Group A), the implant was placed without any grafting material. In Groups B and C, the elevated sinus was filled with either 0.25 g of Bio-Oss® Bone Graft or 100 mg of Bio-Oss collagen mixed with CGF, respectively. For all groups, the implant was then inserted simultaneously and more palatal, using a submerged technique and using a two-stage procedure. Postoperative CBCT scans were taken immediately for all patients. Postsurgical care following OSFE with simultaneous implant placement was performed according to standard postsurgical treatment protocols. Patients were instructed to rinse their mouths with a 0.12% chlorhexidine solution for 60 s, five times a day, for 14 days. Additionally, anti-inflammatory drugs and antibiotics were prescribed following the surgery. Grafting materials Geistlich Bio-Oss® small granules (0.25 – 1 mm) (Geistlich Pharma AG, Wolhusen, Switzerland); Geistlich Bio-Oss Collagen®, comprised of 90% Geistlich Bio-Oss® granules and 10% of porcine collagen (Geistlich Pharma AG, Wolhusen, Switzerland)); and CGF, which was prepared in the hospital directly before surgery using a previously described method , were used as grafting materials. For the CGF with the Bio-Oss group, the extracted CGF layer was then separated and divided into small fragments using sterile scissors, and mixed with 100 mg Geistlich Bio-Oss Collagen® (Fig. ). Prosthetic rehabilitation CBCT was taken for each patient at 3 months (T1) and at six months (T2) after OSFE, followed by the second-stage surgery. Following two weeks secondary healing period, dental impressions were made and implant stability was evaluated. Two weeks later, the final restorations were performed after the insertion of the prosthetic abutments. Each implant was used to hold a single crown. Variables The primary predictor variables were three groups according to the type of bone graft used: Group A (control group without any bone graft), Group B (with Bio-Oss bone graft), and Group C (Bio-Oss collagen mixed with CGF). The primary outcome variable was: Implant survival rate. The secondary outcome variables were the changes in the endo-sinus bone gain at different time points which were measured by the following parameters: Height of the apical bone (H, mm), Sinus Lift (SL, mm), Vertical Bone Resorption (VBR, mm), Grafting size (D, mm), Total Bone Resorption (TBR, mm), Implant Stability Quotient (ISQ, between 0 and 100), and Bone density (B, HU). Other secondary outcomes included the post-surgery patient’s pain based on VAS score (1 to 100, mm), post-surgery patient’s satisfaction VAS score (1 to 10, cm), and willingness to do this procedure again (Yes, No). Covariates included the age by years, sex (male, female), smoking (smoker, non-smoker), Residual Bone Height (RBH, mm), Alveolar Bone Width (ABW, mm), and Implant protrusion length (IPL). Data collection Implant survival rate Implants success was evaluated at two time points: six months after implant placement and one year after functional loading of the upper prosthesis. The survival rate was recorded using the following success criteria: no implant mobility detected during clinical examination; no pain or any unusual complaint from the patient; no peri-implant radiolucency, infection, or neuropathies; fully functional suited prosthetic constructions without positional change; and an absence of any occlusal malfunctions. Radiographic evaluation For each patient, CBCT imaging using GiANO (NewTom, Imola, Italy) with NewTom NNT analysis software was performed at each stage outlined in the study flowchart (Fig. ) due to its superior spatial resolution. The imaging was conducted at the following time points: preoperative, immediately after first stage surgery T0, 3 months after first stage surgery T1, before second stage surgery T2, and one year after loading with final restorations T3. All patients received full high-resolution scan: Voxel size 0.075 mm (12.6 mAs, 90 kVp, 3 mA), a field of view (FOV) of 10 (cm) × 10 (cm), and 360° rotation around patients in 3.6-s scan time. All the linear variables were measured on the coronal cross-sections parallel to the longitudinal direction of the implant, using the measuring tool of NewTom NNT analysis software. The precision of the measuring system is 0.01 mm. Measurements were assessed by a single operator three times and the average was calculated. The following are definitions of this study linear variables: RBH: the vertical distance between the alveolar crest and the floor of the maxillary sinus along the maxilla axis was used to measure RBH before surgery. ABW: the horizontal width of the alveolar bone was measured at 3 mm below the alveolar bone crest. RBH and ABW were only assessed before surgery using preoperative CBCT. IPL was calculated as the implant length IL minus RBH (IPL = IL-RBH). On postoperative CBCT taken for each implant at four time points, four planes have been indicated orthogonal to the long axis of inserted implant using the previous software, as shown in Fig. . Plane ‘A’ was established passing through highest point of the elevated sinus floor after OSFE, Plane ‘B’ was established by passing through the vertex of the implant, Plane ‘Cp’ was established by passing through the bottom level of the maxillary sinus from the palatal side, and Plane ‘Cb’ was established by passing through the bottom level of the maxillary sinus from the buccal side. The vertical distance (H), representing the height of the augmented sinus floor, between Plane ‘A’ and Plane ‘B’ was measured as the at four time points. Immediately after surgery, the maxillary sinus floor (Plane 'A') was positioned above the apex of all implants, resulting in H0 being greater than 0. At follow-up, when the maxillary sinus floor was on contact with the implant apex, plane ‘A’ and plane ‘B’ overlapped, and H was considered to be 0. SL was calculated as H + IPL at four time points. VBR represented the difference in SL between the different follow-up time points. It was calculated twice: first as the difference between SL0 and SL2 (VBR1), and second as the difference between SL0 and SL3 (VBR2). D was defined as the mean vertical distance between the initial sinus floor and the elevated sinus floor assessed at buccal and palatal sides. It was calculated as the average of buccal D (the vertical distance between plane ‘A’ and plane ‘Cb’), and palatal D (the vertical distance between plane ‘A’ and plane ‘Cp’). TBR represented the difference in D between the different follow-up time points. It was calculated twice: first as the difference between D0 and D2 (TBR1), and second as the difference between D0 and D3 (TBR2). H, buccal D, palatal D, D, and SL were measured and calculated four each implant at four time points postoperatively, immediately after the first stage surgery (T0), 3 months after the first stage surgery (T1), 6 months after the first stage surgery and before the second stage surgery (T2), and one year after functional loading (T3). Implant stability quotient Implant Stability was measured using The Osstell resonance frequency analyzer (Osstell, Göteborg, Sweden) for each specimen two weeks following the second stage surgery and before dental impression was made. The resonance frequency measurement, an indicator for mechanical implant stability, was assigned a value between 0 and 100. ISQ was measured 3 times for each specimen, and the median was calculated. Bone density The bone tissue density was analyzed using NewTom NNT software with a spot diameter of 1 mm at three regions around the center of the measured implant protrusion buccally and lingually. The mean value of the three measurements of the average bone volume was then calculated and expressed in Hounsfield units (HU). B was measured and calculated four each implant at three time points T1, T2, and T3. Patient’s satisfaction Patient satisfaction was assessed using a simple questionnaire administered three days postoperatively following the first stage of surgery. Initially, patients were asked to rate their pain using a 100-mm Visual Analog Scale (VAS) scale. The pain scores were categorized as follows: 0 to 4 mm indicated no pain, 5 to 44 mm indicated mild pain, 45 to 74 mm indicated moderate pain, and 75 to 100 mm indicated severe pain. Subsequently, patients were asked to rate their satisfaction using 10-cm VAS scale, ranging from 0 meaning not satisfied to 10 meaning very satisfied. Finally, patients were inquired about their willingness to undergo similar procedures in the future if they needed (yes or no question). Data analysis Statistical analysis of data was performed using IBM SPSS Statistics for Windows, Version 23.0 (IBM Corporation, Armonk, New York). The Shapiro–Wilk test was implemented to assess whether the data followed a normal distribution. ANOVA test followed by Tukey’s post hoc test was applied to analyze differences in RBH, ABW, IPL, SL, D, B, VBR, TBR, ISQ, pain VAS score and satisfaction VAS score measurements between three groups (A, B and C), and to look at how each group changed over time. The χ 2 test was used to compare ISR, sex, smoking status, and patients' willingness to undergo a similar procedure in the future among the three groups. Fisher’s exact test and independent t -test were used to conduct bivariate analyses of the covariates versus ISR. The Pearson’s correlation coefficient was calculated between primary outcome ISR and variables (RBH, B1 and B2). Additionally, the Pearson’s correlation coefficient and Eta correlation ratio were calculated between the pain VAS score and other variables. Three multiple logistic regressions were implemented to examine the relationship between radiographic measurements (H, D and SL) and ISR at three time points T0, T1, and T2. P < 0.05 was considered to indicate a statistically significant difference.
This prospective study was conducted in the Affiliated Hospital of Stomatology, Nanjing Medical University, China. This study complies with the Declaration of Helsinki and the Good Clinical Practice guidelines, and was approved by Medical Ethics Committee, School of Stomatology, Nanjing Medical University, China (Approval number: PJ2020-141–001). Informed consent for data evaluation and publishing has been obtained from all included subjects. Patients meeting our inclusion criteria underwent OSFE with simultaneous implantation from March 2021 to January 2022. A detailed medical record was established for each patient, including medical and dental history, oral examination, surgical records, details of bone graft materials, and the type of prosthetic reconstruction. Inclusion criteria for enrollment in this study were as follows: Patient whom underwent OSFE with simultaneous implantation must be at least 18 years of age; have good general health; adequate oral hygiene, indicated by a bleeding index of less than 30% and a plaque score of less than 20%; have been extracted 1 or 2 maxillary molars or pre-molars because of failure of endodontic treatment, root fracture, or after suffering from severe caries for more than 3 months. Radiographic inclusion criteria are as follows: Residual Bone Height (RBH) of the alveolar bone crest measured on CBCT at each implant site is 2–5 mm; an adequate residual alveolar ridge width for implant placement of 6 mm or more; absence of any radiographic signs of maxillary sinus pathology; and length of the inserted implants ranging from 8–10 mm with diameters of 4.1/4.8 mm (Dentium implant/ ICX implant/ MIS). Exclusion criteria were as follows: postoperative CBCT imaging lacking clarity or unclear anatomic references, or incomplete medical history due to the failure of the patient to come for follow up appointments. Using the R program (R Foundation for Statistical Computing) , at least 107 subjects were required to evaluate the association between the primary outcome and three categorial variables. This calculation assumed a medium effect size of 0.3 (measured by Cohen's d), a P -value of 0.05, and a power of 80%. Considering the dropout rate of 20%, a total of 134 subjects were needed. Surgical procedure Patients were treated according to the designed treatment plan. They were assigned randomized numbers and subsequently divided into three different groups based on those numbers. The surgery consisted of simultaneous implant placement with sinus augmentation via the Osteotome technique (OSFE) by surgeon B.SH. All patients underwent a comprehensive oral examination and received a CBCT scan prior to surgery. Pre-operative CBCT image was used to evaluate RBH and crestal bone width. Those meeting our inclusion criteria were selected for the study. For all subjects, local anesthesia was performed, a mid-crestal incision was made and flap raised, and then the implant site was generated using a pilot drill, maintaining a distance of 1 mm from the sinus floor. Next, the bone of the sinus floor was fractured into the sinus cavity, elevating the Schneiderian membrane, with a vigilant tap of the mallet. The initial sinus elevation was performed with osteotomes, gradually advancing until the final depth was achieved. In the first group (Group A), the implant was placed without any grafting material. In Groups B and C, the elevated sinus was filled with either 0.25 g of Bio-Oss® Bone Graft or 100 mg of Bio-Oss collagen mixed with CGF, respectively. For all groups, the implant was then inserted simultaneously and more palatal, using a submerged technique and using a two-stage procedure. Postoperative CBCT scans were taken immediately for all patients. Postsurgical care following OSFE with simultaneous implant placement was performed according to standard postsurgical treatment protocols. Patients were instructed to rinse their mouths with a 0.12% chlorhexidine solution for 60 s, five times a day, for 14 days. Additionally, anti-inflammatory drugs and antibiotics were prescribed following the surgery. Grafting materials Geistlich Bio-Oss® small granules (0.25 – 1 mm) (Geistlich Pharma AG, Wolhusen, Switzerland); Geistlich Bio-Oss Collagen®, comprised of 90% Geistlich Bio-Oss® granules and 10% of porcine collagen (Geistlich Pharma AG, Wolhusen, Switzerland)); and CGF, which was prepared in the hospital directly before surgery using a previously described method , were used as grafting materials. For the CGF with the Bio-Oss group, the extracted CGF layer was then separated and divided into small fragments using sterile scissors, and mixed with 100 mg Geistlich Bio-Oss Collagen® (Fig. ). Prosthetic rehabilitation CBCT was taken for each patient at 3 months (T1) and at six months (T2) after OSFE, followed by the second-stage surgery. Following two weeks secondary healing period, dental impressions were made and implant stability was evaluated. Two weeks later, the final restorations were performed after the insertion of the prosthetic abutments. Each implant was used to hold a single crown.
Patients were treated according to the designed treatment plan. They were assigned randomized numbers and subsequently divided into three different groups based on those numbers. The surgery consisted of simultaneous implant placement with sinus augmentation via the Osteotome technique (OSFE) by surgeon B.SH. All patients underwent a comprehensive oral examination and received a CBCT scan prior to surgery. Pre-operative CBCT image was used to evaluate RBH and crestal bone width. Those meeting our inclusion criteria were selected for the study. For all subjects, local anesthesia was performed, a mid-crestal incision was made and flap raised, and then the implant site was generated using a pilot drill, maintaining a distance of 1 mm from the sinus floor. Next, the bone of the sinus floor was fractured into the sinus cavity, elevating the Schneiderian membrane, with a vigilant tap of the mallet. The initial sinus elevation was performed with osteotomes, gradually advancing until the final depth was achieved. In the first group (Group A), the implant was placed without any grafting material. In Groups B and C, the elevated sinus was filled with either 0.25 g of Bio-Oss® Bone Graft or 100 mg of Bio-Oss collagen mixed with CGF, respectively. For all groups, the implant was then inserted simultaneously and more palatal, using a submerged technique and using a two-stage procedure. Postoperative CBCT scans were taken immediately for all patients. Postsurgical care following OSFE with simultaneous implant placement was performed according to standard postsurgical treatment protocols. Patients were instructed to rinse their mouths with a 0.12% chlorhexidine solution for 60 s, five times a day, for 14 days. Additionally, anti-inflammatory drugs and antibiotics were prescribed following the surgery.
Geistlich Bio-Oss® small granules (0.25 – 1 mm) (Geistlich Pharma AG, Wolhusen, Switzerland); Geistlich Bio-Oss Collagen®, comprised of 90% Geistlich Bio-Oss® granules and 10% of porcine collagen (Geistlich Pharma AG, Wolhusen, Switzerland)); and CGF, which was prepared in the hospital directly before surgery using a previously described method , were used as grafting materials. For the CGF with the Bio-Oss group, the extracted CGF layer was then separated and divided into small fragments using sterile scissors, and mixed with 100 mg Geistlich Bio-Oss Collagen® (Fig. ).
CBCT was taken for each patient at 3 months (T1) and at six months (T2) after OSFE, followed by the second-stage surgery. Following two weeks secondary healing period, dental impressions were made and implant stability was evaluated. Two weeks later, the final restorations were performed after the insertion of the prosthetic abutments. Each implant was used to hold a single crown.
The primary predictor variables were three groups according to the type of bone graft used: Group A (control group without any bone graft), Group B (with Bio-Oss bone graft), and Group C (Bio-Oss collagen mixed with CGF). The primary outcome variable was: Implant survival rate. The secondary outcome variables were the changes in the endo-sinus bone gain at different time points which were measured by the following parameters: Height of the apical bone (H, mm), Sinus Lift (SL, mm), Vertical Bone Resorption (VBR, mm), Grafting size (D, mm), Total Bone Resorption (TBR, mm), Implant Stability Quotient (ISQ, between 0 and 100), and Bone density (B, HU). Other secondary outcomes included the post-surgery patient’s pain based on VAS score (1 to 100, mm), post-surgery patient’s satisfaction VAS score (1 to 10, cm), and willingness to do this procedure again (Yes, No). Covariates included the age by years, sex (male, female), smoking (smoker, non-smoker), Residual Bone Height (RBH, mm), Alveolar Bone Width (ABW, mm), and Implant protrusion length (IPL).
Implant survival rate Implants success was evaluated at two time points: six months after implant placement and one year after functional loading of the upper prosthesis. The survival rate was recorded using the following success criteria: no implant mobility detected during clinical examination; no pain or any unusual complaint from the patient; no peri-implant radiolucency, infection, or neuropathies; fully functional suited prosthetic constructions without positional change; and an absence of any occlusal malfunctions. Radiographic evaluation For each patient, CBCT imaging using GiANO (NewTom, Imola, Italy) with NewTom NNT analysis software was performed at each stage outlined in the study flowchart (Fig. ) due to its superior spatial resolution. The imaging was conducted at the following time points: preoperative, immediately after first stage surgery T0, 3 months after first stage surgery T1, before second stage surgery T2, and one year after loading with final restorations T3. All patients received full high-resolution scan: Voxel size 0.075 mm (12.6 mAs, 90 kVp, 3 mA), a field of view (FOV) of 10 (cm) × 10 (cm), and 360° rotation around patients in 3.6-s scan time. All the linear variables were measured on the coronal cross-sections parallel to the longitudinal direction of the implant, using the measuring tool of NewTom NNT analysis software. The precision of the measuring system is 0.01 mm. Measurements were assessed by a single operator three times and the average was calculated. The following are definitions of this study linear variables: RBH: the vertical distance between the alveolar crest and the floor of the maxillary sinus along the maxilla axis was used to measure RBH before surgery. ABW: the horizontal width of the alveolar bone was measured at 3 mm below the alveolar bone crest. RBH and ABW were only assessed before surgery using preoperative CBCT. IPL was calculated as the implant length IL minus RBH (IPL = IL-RBH). On postoperative CBCT taken for each implant at four time points, four planes have been indicated orthogonal to the long axis of inserted implant using the previous software, as shown in Fig. . Plane ‘A’ was established passing through highest point of the elevated sinus floor after OSFE, Plane ‘B’ was established by passing through the vertex of the implant, Plane ‘Cp’ was established by passing through the bottom level of the maxillary sinus from the palatal side, and Plane ‘Cb’ was established by passing through the bottom level of the maxillary sinus from the buccal side. The vertical distance (H), representing the height of the augmented sinus floor, between Plane ‘A’ and Plane ‘B’ was measured as the at four time points. Immediately after surgery, the maxillary sinus floor (Plane 'A') was positioned above the apex of all implants, resulting in H0 being greater than 0. At follow-up, when the maxillary sinus floor was on contact with the implant apex, plane ‘A’ and plane ‘B’ overlapped, and H was considered to be 0. SL was calculated as H + IPL at four time points. VBR represented the difference in SL between the different follow-up time points. It was calculated twice: first as the difference between SL0 and SL2 (VBR1), and second as the difference between SL0 and SL3 (VBR2). D was defined as the mean vertical distance between the initial sinus floor and the elevated sinus floor assessed at buccal and palatal sides. It was calculated as the average of buccal D (the vertical distance between plane ‘A’ and plane ‘Cb’), and palatal D (the vertical distance between plane ‘A’ and plane ‘Cp’). TBR represented the difference in D between the different follow-up time points. It was calculated twice: first as the difference between D0 and D2 (TBR1), and second as the difference between D0 and D3 (TBR2). H, buccal D, palatal D, D, and SL were measured and calculated four each implant at four time points postoperatively, immediately after the first stage surgery (T0), 3 months after the first stage surgery (T1), 6 months after the first stage surgery and before the second stage surgery (T2), and one year after functional loading (T3). Implant stability quotient Implant Stability was measured using The Osstell resonance frequency analyzer (Osstell, Göteborg, Sweden) for each specimen two weeks following the second stage surgery and before dental impression was made. The resonance frequency measurement, an indicator for mechanical implant stability, was assigned a value between 0 and 100. ISQ was measured 3 times for each specimen, and the median was calculated. Bone density The bone tissue density was analyzed using NewTom NNT software with a spot diameter of 1 mm at three regions around the center of the measured implant protrusion buccally and lingually. The mean value of the three measurements of the average bone volume was then calculated and expressed in Hounsfield units (HU). B was measured and calculated four each implant at three time points T1, T2, and T3. Patient’s satisfaction Patient satisfaction was assessed using a simple questionnaire administered three days postoperatively following the first stage of surgery. Initially, patients were asked to rate their pain using a 100-mm Visual Analog Scale (VAS) scale. The pain scores were categorized as follows: 0 to 4 mm indicated no pain, 5 to 44 mm indicated mild pain, 45 to 74 mm indicated moderate pain, and 75 to 100 mm indicated severe pain. Subsequently, patients were asked to rate their satisfaction using 10-cm VAS scale, ranging from 0 meaning not satisfied to 10 meaning very satisfied. Finally, patients were inquired about their willingness to undergo similar procedures in the future if they needed (yes or no question).
Implants success was evaluated at two time points: six months after implant placement and one year after functional loading of the upper prosthesis. The survival rate was recorded using the following success criteria: no implant mobility detected during clinical examination; no pain or any unusual complaint from the patient; no peri-implant radiolucency, infection, or neuropathies; fully functional suited prosthetic constructions without positional change; and an absence of any occlusal malfunctions.
For each patient, CBCT imaging using GiANO (NewTom, Imola, Italy) with NewTom NNT analysis software was performed at each stage outlined in the study flowchart (Fig. ) due to its superior spatial resolution. The imaging was conducted at the following time points: preoperative, immediately after first stage surgery T0, 3 months after first stage surgery T1, before second stage surgery T2, and one year after loading with final restorations T3. All patients received full high-resolution scan: Voxel size 0.075 mm (12.6 mAs, 90 kVp, 3 mA), a field of view (FOV) of 10 (cm) × 10 (cm), and 360° rotation around patients in 3.6-s scan time. All the linear variables were measured on the coronal cross-sections parallel to the longitudinal direction of the implant, using the measuring tool of NewTom NNT analysis software. The precision of the measuring system is 0.01 mm. Measurements were assessed by a single operator three times and the average was calculated. The following are definitions of this study linear variables: RBH: the vertical distance between the alveolar crest and the floor of the maxillary sinus along the maxilla axis was used to measure RBH before surgery. ABW: the horizontal width of the alveolar bone was measured at 3 mm below the alveolar bone crest. RBH and ABW were only assessed before surgery using preoperative CBCT. IPL was calculated as the implant length IL minus RBH (IPL = IL-RBH). On postoperative CBCT taken for each implant at four time points, four planes have been indicated orthogonal to the long axis of inserted implant using the previous software, as shown in Fig. . Plane ‘A’ was established passing through highest point of the elevated sinus floor after OSFE, Plane ‘B’ was established by passing through the vertex of the implant, Plane ‘Cp’ was established by passing through the bottom level of the maxillary sinus from the palatal side, and Plane ‘Cb’ was established by passing through the bottom level of the maxillary sinus from the buccal side. The vertical distance (H), representing the height of the augmented sinus floor, between Plane ‘A’ and Plane ‘B’ was measured as the at four time points. Immediately after surgery, the maxillary sinus floor (Plane 'A') was positioned above the apex of all implants, resulting in H0 being greater than 0. At follow-up, when the maxillary sinus floor was on contact with the implant apex, plane ‘A’ and plane ‘B’ overlapped, and H was considered to be 0. SL was calculated as H + IPL at four time points. VBR represented the difference in SL between the different follow-up time points. It was calculated twice: first as the difference between SL0 and SL2 (VBR1), and second as the difference between SL0 and SL3 (VBR2). D was defined as the mean vertical distance between the initial sinus floor and the elevated sinus floor assessed at buccal and palatal sides. It was calculated as the average of buccal D (the vertical distance between plane ‘A’ and plane ‘Cb’), and palatal D (the vertical distance between plane ‘A’ and plane ‘Cp’). TBR represented the difference in D between the different follow-up time points. It was calculated twice: first as the difference between D0 and D2 (TBR1), and second as the difference between D0 and D3 (TBR2). H, buccal D, palatal D, D, and SL were measured and calculated four each implant at four time points postoperatively, immediately after the first stage surgery (T0), 3 months after the first stage surgery (T1), 6 months after the first stage surgery and before the second stage surgery (T2), and one year after functional loading (T3).
Implant Stability was measured using The Osstell resonance frequency analyzer (Osstell, Göteborg, Sweden) for each specimen two weeks following the second stage surgery and before dental impression was made. The resonance frequency measurement, an indicator for mechanical implant stability, was assigned a value between 0 and 100. ISQ was measured 3 times for each specimen, and the median was calculated.
The bone tissue density was analyzed using NewTom NNT software with a spot diameter of 1 mm at three regions around the center of the measured implant protrusion buccally and lingually. The mean value of the three measurements of the average bone volume was then calculated and expressed in Hounsfield units (HU). B was measured and calculated four each implant at three time points T1, T2, and T3.
Patient satisfaction was assessed using a simple questionnaire administered three days postoperatively following the first stage of surgery. Initially, patients were asked to rate their pain using a 100-mm Visual Analog Scale (VAS) scale. The pain scores were categorized as follows: 0 to 4 mm indicated no pain, 5 to 44 mm indicated mild pain, 45 to 74 mm indicated moderate pain, and 75 to 100 mm indicated severe pain. Subsequently, patients were asked to rate their satisfaction using 10-cm VAS scale, ranging from 0 meaning not satisfied to 10 meaning very satisfied. Finally, patients were inquired about their willingness to undergo similar procedures in the future if they needed (yes or no question).
Statistical analysis of data was performed using IBM SPSS Statistics for Windows, Version 23.0 (IBM Corporation, Armonk, New York). The Shapiro–Wilk test was implemented to assess whether the data followed a normal distribution. ANOVA test followed by Tukey’s post hoc test was applied to analyze differences in RBH, ABW, IPL, SL, D, B, VBR, TBR, ISQ, pain VAS score and satisfaction VAS score measurements between three groups (A, B and C), and to look at how each group changed over time. The χ 2 test was used to compare ISR, sex, smoking status, and patients' willingness to undergo a similar procedure in the future among the three groups. Fisher’s exact test and independent t -test were used to conduct bivariate analyses of the covariates versus ISR. The Pearson’s correlation coefficient was calculated between primary outcome ISR and variables (RBH, B1 and B2). Additionally, the Pearson’s correlation coefficient and Eta correlation ratio were calculated between the pain VAS score and other variables. Three multiple logistic regressions were implemented to examine the relationship between radiographic measurements (H, D and SL) and ISR at three time points T0, T1, and T2. P < 0.05 was considered to indicate a statistically significant difference.
This study included 135 patients, with 45 in each group. Twelve subjects were excluded due to unclear CBCT images while 123 patients (126 implants) completed the study assessment follow-up plan. The patients comprised of 79 men and 44 women with a mean age of 58 ± 11.2 (20–78) years. Radiographic assessment of pre-operative CBCT indicated RBH ≤ 5 mm (mean of 4.37 ± 0.75, range 2.6–5) and ABW ≥ 6 (mean of ABW = 11.53 ± 1.89, rang 6–16). Comparison of covariates versus study groups A, B and C is presented in Table . Total implant survival rate was 96%. All 5 failed cases were found at T2 and then removed and rescheduled for retreatment. Comparatively, no implant failure was found at T3 time points, after one year of functional loading. Bivariate analyses of the covariates versus ISR are presented in Table . No significant difference was found for ISR, 95.6% for Group A ( n = 45), 97.6% for Group B ( n = 41), and 95% for Group C ( n = 40) (Table ). During the follow-up period, four cases of acute sinusitis were documented, three in Group A and one in Group B. These infections were effectively treated with amoxicillin-clavulanate (2,000 mg/125 mg every 12 h for five days) and did not adversely affect the functional performance of the implants. Moreover, no occurrences of Schneiderian membrane perforation were observed throughout the duration of the study. Radiographic assessment of CBCT after surgery was taken at four time points and is recorded in Table . Significant increase of radiographic measurements for H, D and SL were recorded in both grafted Groups B and C compared to Group A at four time points after surgery T0, T1, T2 and T3 ( P < 0.001) with significantly increased D3 (n = 121) in Group C compared to Groups A and B. VBR1 ( n = 126) and VBR2 ( n = 121) had no significant difference between the groups. While TBR1 ( n = 126) and TBR2 ( n = 121) were significantly increased in Group A compared to Groups B and C ( P = 0.004, P = 0.000), TBR2 also significantly increased in Group B compared to Group C. ISQ measured at the T2 time point ( n = 121) significantly increased in Group B compared to Group A ( P = 0.000), and also increased in Group C compared to Group A as well, but without significant difference. B1 ( n = 126), B2 ( n = 126), and B3 ( n = 121) all significantly increased in Groups B and C compared to Group A ( P < 0.001). The bivariate association between some predictor variables and primary outcome ISR is presented in Table . RBH was indicated to have a significant positive relationship with ISR, [r (126) = 0.359, P = 0.000]. Moreover, there was a significant positive relationship between B1 and ISR, [r (126) = 0.271, P = 0.002], and there was a significant positive relationship between B2 and ISR, [r (126) = 0.359, P = 0.000]. Three multivariate logistic regression analyses for radiographic measurements (H, D and SL) at T0, T1, and T2 time points as predictors for ISR are presented in Table . In the first multivariate regression analyses, implants with higher D0 value were 8.06 times more likely to survive after 6 months of implantation (OR 8.06; 95% CI 1.59 to 38.24; P = 0.010). In the second multivariate regression analyses, implants with higher D1 value were 96.58 times more likely to survive after 6 months of implantation (OR 96.58; 95% CI 1.69 to 5.52; P = 0.027). In the third multivariate regression analyses, implants with higher D2 value were 4.97 times more likely to survive after 6 months of implantation (OR 4.97; 95% CI 1.29 to 19.19; P = 0.020). The results of the patient satisfaction questionnaire, as presented in Table , indicate a significant increase in pain VAS scores (mean of pain VAS = 45.18 ± 20.6, range 3–89) in Group B compared to Groups A and C (P = 0.000). The distribution of pain scores categories between groups are illustrated in Fig. . Additionally, there was a significant increase in patient satisfaction VAS scores (mean of satisfaction VAS = 6.4 ± 2.12, range 1–10) in Groups C and A compared to Group B ( P = 0.000). Furthermore, there was a significant increase in the refusal to undergo similar procedures in the future in Group B compared to Groups A and C ( P = 0.041 and P = 0.032, respectively). After studying the correlation between pain scores and various variables, we identified several significant relationships. There was a significant positive relationship between pain score and age [r (123) = 0.189, P = 0.036], a significant negative relationship between pain score and RBH [r (123) = -0.324, P = 0.000], and a significant positive relationship between pain score and IPL [r (123) = 0.207, P = 0.021]. A significant negative relationship was also found between pain score and satisfaction score [r (123) = -0.752, P = 0.000]. Furthermore, there was a medium association between pain score and patients' refusal to undergo similar procedures in the future [η (123) = 0.553, P = 0.000]. Lastly, there was a weak association between pain score and female gender [η (123) = 0.312, P = 0.000], and no association between pain score and smoking [η (123) = 0.189, P = 0.036].
The purpose of this radiological study was to investigate the ability of the combination of CGF with Bio-Oss collagen as grafting materials to induce osteogenesis in the maxillary sinus after OSFE and simultaneous implant placement in patients with severe vertical defects in the alveolar ridge RBH ≤ 5 mm. The null hypothesis posited that no significant difference exists in the long-term intra-sinus osteogenesis process among the various grafting materials utilized following OSFE with simultaneous implantation. The positive correlation between RBH and implant survival rate indicates combined with simultaneous implant placement in the atrophic maxilla can achieve high implant survival rates, under routine clinical practice conditions. In this study, we propose the combination of Bio-Oss Collagen and CGF as grafting materials following OSFE and simultaneous implant placement, a topic not previously investigated in the literature . As shown in our results and in previous studies, sinus floor elevation can be performed without graft material, resulting in sufficient bone development and implant longevity . However, a study conducted by Kim et al. in an animal model demonstrated that bone formation is significantly constrained when sinus lift surgery is performed without the use of grafting materials . Moreover, CGF has shown considerable potential in tissue regeneration, attributed to its capacity to enhance cell proliferation, migration, and differentiation, as well as its ability to stimulate angiogenesis and osteogenesis . The utilization of growth factors has been shown to enhance clinical outcomes by promoting improved vascularization at surgical sites. Furthermore, their application contributes to better postoperative recovery, significantly improving patients' quality of life . The application of CGF membrane as the sole grafting material in conjunction with OSFE and simultaneous implant placement in the atrophic maxilla has been reported to yield favorable outcomes, including significant vertical bone augmentation immediately following surgery . Chen, H., et al. reported that after a 24-month follow-up, there was no significant difference in marginal bone loss between OSFE performed using concentrated growth factors (CGF) alone and OSFE using CGF in combination with bone grafting materials. However, CGF alone was preferred due to its superior patient satisfaction and safety profile . ISR demonstrated a significant positive correlation with grafting size D across three follow-up time points. In the present study, all bone formation indicators showed significant improvement in both grafted groups during follow-up evaluations. While VBR1 and VBR2 did not differ significantly between the groups, TBR1 and TBR2 exhibited a notable increase in the non-grafted Group A. After one year of functional loading, D3 increased in Group C significantly compared to Groups A and B, while TBR2 significantly receded in Group C compared to Groups A and B. Additionally, it was observed that the initial D0 and H0 values were significantly lower in Group A compared to Groups B and C, with Group C exhibiting the highest values. This indicates a notable variation in baseline measurements across the groups, with Group C demonstrating superior initial parameters. Notably, this trend persisted after 12 months of follow-up, as evidenced by the D3 and H3 measurements. These findings suggest that, while collagen combined with CGF may undergo absorption by the six-month mark, as reported in previous research , endo-sinus bone gain begins to develop gradually thereafter, effectively slowing the resorption process by the one-year follow-up. This supports the hypothesis that the combination of Bio-Oss collagen with CGF promotes tissue repair and regeneration, offering a sustained and beneficial effect on bone formation over time. Furthermore, this combination may facilitate new bone formation in the sinus when applied following OSFE, potentially improving the outcomes of sinus augmentation procedures. Compatible with these findings, a recent radiographic study reported that the application of Bio-Oss collagen after OSFE and simultaneous implantation achieved a significantly increased initial endo-sinus bone gain compared to the non-grafted group . However, unlike our results, the difference in endo-sinus bone in the aforementioned study diminished after one year of functional loading. In contrast, in our study, bone gain indicators in Group C remained significantly higher than those in the non-grafted group at the T3 time point, suggesting a more sustained enhancement of bone formation over time. Moreover, our findings align with a previous study that successfully employed a combination of CGF and collagen in conjunction with bone grafts for alveolar ridge preservation . This study demonstrated the efficacy of this regenerative approach in promoting tissue healing and enhancing bone regeneration, further supporting the positive outcomes observed in our study. Moreover, simultaneous sinus augmentation and implantation with the presence of grafting material can help assist preserve the augmented sinus membrane and prevent marginal bone loss, as demonstrated in previous studies . This protective effect may contribute to the slower absorption of grafting materials observed in Group C. Most importantly, a reduction in bone absorption coupled with the synthesis of new bone was observed in the sinus. This finding aligns with a previous study, which indicated that alveolar bone levels in extraction sockets were better preserved using collagen in conjunction with early implantation, as compared to collagen alone without implantation . Furthermore, collagen helps maintain the space needed for the osteoinductive properties of the coagulum by enlarging the space, which promotes increased vascularization and stabilization of the coagulum. This favorable microenvironment supports the recruitment of osteoprogenitor cells and promotes effective bone regeneration. A recent study involving both radiographic and histomorphologic examination demonstrated that the use of collagen alone following OSFE was sufficient to induce proper new bone formation . In our study, the average final sinus lift achieved in Group C was 7.44 mm, which closely approximates the 7.75 mm reported by Yerko, et al., where autologous fibrin glue with a collagen carrier was used during lateral sinus augmentation with simultaneous implantation in the atrophic maxilla with less than 5 mm RBH . This similarity in results further underscores the effectiveness of collagen, either alone or in combination with other materials, in achieving substantial sinus lift and promoting successful bone regeneration in challenging clinical scenarios. To our knowledge, this is the first study that investigated the functional application of the combination of Bio-Oss collagen with CGF as grafting materials after OSFE. Although previous research has explored the use of collagen combined with CGF membranes for alveolar ridge preservation, demonstrating enhanced soft tissue healing , this study uniquely focuses on the application of this combination in the context of OSFE. By addressing this specific clinical scenario, our study contributes new insights into the potential benefits of this grafting combination for promoting bone regeneration and preserving the sinus membrane following sinus augmentation procedures. Moreover, a previous animal study confirmed the positive effects of combining the key growth factor in CGF with Bio-Oss collagen for bone formation . This combination was found to activate the PI3K/AKT signaling pathway in a rat cranial defect model, a mechanism known to play a critical role in regulating osteogenesis and bone remodeling. The activation of this signaling pathway suggests that the synergistic effects of CGF and Bio-Oss collagen may enhance cellular processes involved in bone regeneration, offering a promising therapeutic approach for improving bone healing in clinical settings. When RBH is severely defected, the implant initial stability is likely to be poor. The application of bone graft materials can strengthen the initial stability, ISQ measured after second stage surgery was significantly increased in Group B compared to Group A, but not compared to Group C. Nevertheless, bone density was increased in Groups B and C compared to Group A at three-time points follow-up, confirming new bone formation in Group C. Bone density was positively correlated with ISR at both T1 and T2. This result is compatible with a previous study which radiographically evaluated bone morphogenetic protein-2 loaded Bio-Oss collagen for OSFE with simultaneous implant placement in atrophic maxilla. The study recorded high levels of bone density, ranging from 643 to 1201 HU with minimal marginal bone loss and good implant stability after 3 years . Another radiographic study indicated a significant increase in bone density with the application of CGF alone after lateral sinus augmentation and simultaneous implantation compared to no grafting in the atrophic posterior maxilla . The promotion of osseointegration and reduction of bone resorption by adding CGF to Bio-Oss bone grafts has been demonstrated in a previous animal study on extracted socket preservation . However, concerns are rising about the long-term safety of bovine-derived xenografts . In this study, we propose the combination of Bio-Oss collagen with CGF to address these concerns and enhance osseointegration and bone formation. Collagen's soft, flexible, and spongy nature, combined with the mechanical solidity of inorganic bone particles, contributes to sustaining the achieved elevation after sinus augmentation. In addition, collagen's plastic cohesive feature after hydration makes it easier to form and use as a grafting material for OSFE, reducing the incidence of Schneiderian membrane perforation. OSFE with the application of Bio-Oss collagen graft has been proven to improve the quality of the postoperative patient experience . Collagen has proven to stimulate bone formation in bone defects, non-healing extraction sockets, and OSFE with simultaneous implant placement . Possibly, CGF may play a role to overcome the degradation and high absorption rate of collagen by stimulating subsequent osteogenesis and bone formation through the PI3K/AKT signaling pathway. A histological study by Ghasemirad et al. indicated a significantly increased amount of endo-sinus newly-formed bone with the application of CGF after lateral sinus augmentation in the atrophic maxilla compared to the bovine xenograft group at the 6-month time point . However, further investigations are needed to verify this pathway in future research. In our study, we sought to enhance the effectiveness of OSFE by introducing a novel combination of grafting materials. While several recent advancements focus on introducing new techniques to improve internal sinus lifting, our approach emphasizes optimizing the regenerative potential of the procedure through the use of innovative grafting materials. In comparison to these newly developed alternatives to OSFE, osseodensification, similarly to OSFE, demonstrated effective membrane lifting and superior endo-sinus bone gain when grafting materials were applied in cases of atrophic maxilla, as observed at the six-month follow-up . Notably, the previous study reported no instances of implant failure, further emphasizing the favorable clinical outcomes associated with this approach. Additionally, MIAMBE technique has been used in severely atrophic maxilla and proved its safety achieving high patient satisfactions and low pain score with relatively accepted bone gain . Furthermore, hydraulic sinus lift with immediate implant placement—without the use of grafting materials—has achieved a 5–6 mm bone gain at the six-month follow-up, accompanied by a 100% success rate, further underscoring its clinical efficacy and reliability . These findings collectively underscore the potential of various approaches to enhance sinus augmentation outcomes, with the grafting material combination proposed in our study offering a promising alternative for optimizing bone regeneration and improving implant success. In addition, our results indicated increased patient satisfaction VAS scores, decreased postoperative pain VAS scores, and significantly more positive responses regarding willingness to undergo a similar procedure in the future with the application of Bio-Oss collagen and CGF combination after OSFE in Group C compared to Group B. This is consistent with previous reports indicating that OSFE with Bio-Oss collagen application is accompanied by high patient satisfaction, minimal postoperative pain, and reduced fear of undergoing similar surgical procedures . Additionally, our findings closely align with those of Yan Dai et al.’s study , demonstrating that it is highly valuable for alleviating postoperative symptoms and providing pain relief. Moreover, our results indicate a critical need to find alternative procedures that improve patients' postoperative experiences, especially in cases of atrophic maxilla which require careful considerations. The combination of Bio-Oss collagen and CGF after OSFE and simultaneous implantation could be an effective alternative. After radiological evaluation, the indicators of endo-sinus bone diameters and bone density in this study suggest a potential positive impact of the combination of bone collagen with CGF on soft tissue healing and endo-sinus bone formation. This combination contributes to increased implant stability, survival, and longevity, with evidence of improving the postoperative patient experience and providing pain relief. This research has several constraints. Firstly, the study duration was one year after functional loading of final restorations; therefore, further studies with a larger sample size and a longer-term follow-up are needed to evaluate the lifespan of implants after the application of the Bio-Oss collagen and CGF combination. Secondly, while CBCT offers a precision of 0.01 mm, its capability to accurately represent internal soft tissue structures and lesions is limited. Additionally, there is a limited correlation between CBCT and Hounsfield Units for the standardized quantification of bone density. This discrepancy results in an overestimation of bone quantity when compared to the gold standard of micro-CT . Despite these limitations, the structural pattern of the alveolar bone, which is considered the second most important factor in assessing bone quality, remains consistent across CBCT machines with the highest resolution . Thirdly, this research was based on radiographic evaluation to measure endo-sinus bone formation. Further investigations are needed to elucidate the mechanisms and signaling pathways underlying the endo-sinus osteogenesis process. Another limitation of this study lies in its insufficient consideration of the variability in implant macrogeometry when assessing survival rates. Although three distinct implant designs were utilized, the study did not perform a detailed analysis of specific macrogeometric characteristics such as thread pitch, shape, or diameter and their potential impact on outcomes. This oversight may limit the applicability of the findings to other implant designs, reducing the study's external validity and its relevance to broader clinical contexts. Despite these limitations, the study offers valuable insights into endo-sinus bone formation and implant survival. While further investigations into osteogenesis mechanisms and implant macrogeometry are needed, the findings provide a strong foundation for future research and contribute meaningfully to the field.
Based on the findings and within the limitations of this prospective study, the combination of collagen and CGF as a grafting material has demonstrated reliability as a protocol for OSFE. This method has shown particular efficacy for simultaneous implant placement in the atrophic posterior maxilla with a RBH of 5 mm or less, yielding significant endo-sinus bone gain and high levels of patient satisfaction. Future studies are warranted to elucidate the intrinsic mechanisms driving new endo-sinus bone formation, thereby enhancing our understanding of this approach and its clinical applications.
|
Preliminary antifibrotic and vasoconstrictor effects of adrenaline in Schlemm’s canal and suprachoroidal minimally invasive glaucoma surgery in primary open-angle glaucoma | 30fc3ed4-54da-4ba8-8317-5c53c92a4ab1 | 11868348 | Surgical Procedures, Operative[mh] | Glaucoma is the leading cause of irreversible blindness and currently affects 70 million people worldwide . Nowadays, intraocular pressure (IOP) is no longer the only modifiable risk factor in the treatment of glaucoma. Numerous other risk factors can be treated, such as diabetes mellitus, arterial hypertension, and dyslipidemia, all of which have an unfavourable impact on primary open-angle glaucoma (POAG) neurodegeneration . Conventional glaucoma filtration surgery, such as trabeculectomy and glaucoma drainage implant surgery, has been considered the gold standard in glaucoma surgical treatment . Despite exhibiting efficacy at IOP reduction, these incisional surgeries are associated with potential postoperative blinding complications, such as hypotony, long-term risk of endophthalmitis and suprachoroidal haemorrhage . In order to provide a less invasive and safer method for reducing IOP, newer surgical techniques, called minimally invasive glaucoma surgery (MIGS), have been developed and have found their place in the glaucoma treatment paradigm in the last two decades. MIGS refers to a group of IOP-lowering surgical interventions that enhance aqueous humour outflow through Schlemm's canal, the suprachoroidal space or the subconjunctival space . Due to the small device geometry, which reduces trauma to ocular tissues and shortens operation time, MIGS offers enhanced safety profile, predictability, and minimal conjunctival manipulation . MIGS can also be combined with phacoemulsification surgery in patients with both cataract and glaucoma. There are multiple approaches to reduce IOP using MIGS devices. iStent® (Glaukos, Avedro) was the first approved ab interno MIGS implant for use in open-angle glaucoma . It reduces IOP by increasing aqueous outflow from trabecular meshwork and Schlemm’s canal stents . MINIject® (iSTAR Medical SA, Wavre, Belgium) is a new implant developed to target the suprachoroidal space . Its unique flexible design conforms to the shape of the eye and the micropores promote aqueous outflow through the device . Although the evidence supports a good safety profile of using MIGS devices in patients, fibrosis around the implants in the trabecular meshwork (TM) is the main cause of surgical failure in MIGS . The significant postoperative fibrotic tissue formation due to cell proliferation and adhesion limits the success of newly created outflow routes, and makes several new micro-incision devices fail to receive market approval for clinical use . Currently, mitomycin-C (MMC) and 5-fluorouracil (5-FU) are commonly used as non-specific antifibrotic drugs after trabeculectomy. MMC inhibits fibroblast proliferation and 5-FU interferes with cell growth . However, they also carry the risk of severe tissue damage, corneal decompensation and infections , and are therefore too toxic to be used inside the eye in MIGS. Therefore, novel and non-toxic antifibrotic therapies are needed to enhance the long-term effectiveness of MIGS devices. Adrenaline, also known as epinephrine, is endogenously produced by the adrenal glands and has been shown to exhibit distinct antifibrotic effects in glaucoma surgery . We recently investigated the antifibrotic effect of adrenaline by RNA-Sequencing technology and evaluated its impact on fibroblast contractility both in vitro and in three subconjunctival glaucoma surgeries (trabeculectomy, PreserFlo Microshunt, Baerveldt 350 tube surgery). Our results showed that adrenaline substantially decreased conjunctival fibroblast contractility without significant cytotoxicity even at high concentrations of 0.05%, and demonstrated that adrenaline may confer antifibrotic attributes in a concentration-dependent manner by affecting the expression of key cell cycle genes . Given the promising effects in subconjunctival glaucoma surgery, the aim of this study was to investigate the potential benefits of adrenaline on Schlemm's canal and suprachoroidal MIGS devices. By probing the potential antifibrotic effects of adrenaline in these specific locations, this research endeavours to shed light on new strategies to increase the long-term success rates of Schlemm’s canal and suprachoroidal MIGS devices.
Cell culture SV40-immortalized (NTM5) human TM cells were used in this study, and have been characterised and shown to be representative for TM cell studies . Technical replicates were also used in all assays. Human TM cells were cultured in an incubator at 37 °C with 5% CO 2 and 95% humidity. The media for cell culture consisted of Dulbecco’s modified Eagle’s medium (DMEM) (Gibco, ThermoScientific, UK), 10% fetal calf serum (Gibco, ThermoScientific, UK), 100 units/mL penicillin and 0.1 mg/mL streptomycin (Sigma Aldrich, Gillingham, UK). All experiments were conducted in accordance with the Declaration of Helsinki and approved by the West of Scotland Research Ethics Committee (REC 19/WS/0146). Collagen contraction assay A suspension containing 2 × 10 5 TM cells was centrifuged at 1500 rpm for 5 min. The supernatant was discarded, and the cell pellet was resuspended in 100 µL of fetal calf serum. The collagen gel solution was prepared by 1 mL of Type 1 collagen (Fist Link, Wolverhampton, UK) and 160 µL of concentrated media, which consisted of 1.4 mL of DMEM 10 × (Sigma Aldrich, Gillingham, UK), 140 µL of L-glutamine (ThermoScientific, Loughborough, UK), and 360 µL of 7.5% sodium bicarbonate (Sigma Aldrich, Gillingham, UK). The pH was adjusted to 7.0 by sodium hydroxide before the cells were mixed with the collagen solution. 150 µL of the cell-gel mixture was placed in each MatTek dish and allowed to set for 10 min at 37 °C. The gels were treated with 1.5 mL of different concentrations of adrenaline (0%, 0.0005%, 0.01%) after carefully releasing the gels. The cells in the gels were observed using an Olympus CKX41 inverted microscope, and the gel photos were taken daily over 7 days and analysed using the ImageJ software. The percentage of matrix contraction was calculated using the formula: [12pt]{minimal}
$$Contraction\;=[] 100$$ C o n t r a c t i o n = A r e a o f g e l a t D a y 0 - A r e a o f g e l a t D a y n A r e a o f g e l a t D a y 0 × 100 Light microscopy observation Human TM cells were seeded at a density of 1 × 10 5 cells per well in 6-well plates and treated with different concentrations of adrenaline (0%, 0.0001%, 0.0005%, 0.001%, 0.005%, 0.01%). After 1-day treatment, cells were imaged using an Olympus CKX41 inverted microscope with Olympus CellSens Standard 1.13 software. Cell viability assay Human TM cells were seeded at a density of 6.25 × 10 3 cells per well in a 96-well plate and treated with different concentrations of adrenaline (0%, 0.0001%, 0.0005%, 0.001%, 0.005%, 0.01%) for 1 day. The drug solutions in the 96-well plate were then replaced by 100 µL of fresh media, followed by the addition of 20 µL of the Cell Titer 96 Aqueous one solution (Promega, Southampton, UK). The plate was incubated for 2 h at 37 °C and the absorbance was measured at 490 nm using a PHERAstar FS instrument (BMG Labtech, Aylesbury, UK). The cell viability was calculated as a percentage of the value of untreated (0%) cells. Real-time quantitative PCR Human TM cells were seeded at a density of 1 × 10 5 cells per well in 6-well plates and treated with different concentrations of adrenaline (0%, 0.0005%, 0.01%) for 1 day. The total RNA was extracted using a RNeasy mini kit (Qiagen, Crawley, UK) and the cDNA was synthesised using a high-capacity cDNA reverse transcription kit (ThermoScientific, Loughborough, UK). RT-qPCR was performed using the QuantiFast SYBR Green PCR kit (Qiagen, Crawley, UK) on a QuantStudio 7 Real-time PCR system. The reaction settings for 40 cycles were as follows: Holding stage: 50 °C for 2 min and 95 °C for 5 min; PCR stage: 95 °C for 5 min and 60 °C for 30 s. The sequences for the forward and reverse primers are presented in Table . The relative gene expression was calculated using the formula: [12pt]{minimal}
$$Relative\;Gene\;Expression={2}^{-( {Ct}_{target}- {Ct}_{GAPDH})}$$ R e l a t i v e G e n e E x p r e s s i o n = 2 - ( Δ Ct target - Δ Ct GAPDH ) Intracameral injection of adrenaline during MIGS implantation We tested the effects of adrenaline 0.05% intraoperatively in five patients with POAG undergoing a MIGS device combined with phacoemulsification. Two MIGS devices were included in this study: iStent inject® (Glaukos, Avedro) as a Schlemm’s canal MIGS and MINIject® (iSTAR Medical SA, Wavre, Belgium) as a suprachoroidal MIGS. Informed consent was obtained from all patients. Phacoemulsification surgery was performed as per standard practice. Miochol was injected intracamerally at the end of phacoemulsification to constrict the pupil in preparation for MIGS implantation. All patients then received 0.1 mL of intracameral injection of adrenaline 0.05% before the anterior chamber was filled with viscoelastic and the MIGS device was inserted. Clinical examination Clinical data, including best-corrected visual acuity (BCVA), IOP, central corneal thickness (CCT), cup-to-disc ratio (C/D), previous glaucoma surgery, previous glaucoma laser, lens status and anti-glaucoma medications were recorded preoperatively and postoperatively in all patients. Blood pressure, heart rate, and blood oxygen saturation were also measured preoperatively, intraoperatively, and postoperatively. Pupil diameter was measured using a measuring scale before and after surgery. To record the positioning of the MIGS, the intraoperative insertion of the implant was captured by video recording during surgery. Postoperatively, gonioscopy was performed in clinic and photographs of the implant in the angle were taken by slit lamp camera (ARC slit lamp imaging camera, Carleton optical, Chesham, Buckinghamshire, UK). We also performed angle scans with an anterior segment OCT (ANTERION® Heidelberg Engineering, Germany) to visualise the implant transversally. Statistical analyses GraphPad Prism was used for graph generation and statistical analysis. All graphs display mean and standard error of the mean (SEM). To determine statistical significance, a one-way ANOVA was performed and P values were calculated. P values, which were statistically significant, were expressed as follows: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.
SV40-immortalized (NTM5) human TM cells were used in this study, and have been characterised and shown to be representative for TM cell studies . Technical replicates were also used in all assays. Human TM cells were cultured in an incubator at 37 °C with 5% CO 2 and 95% humidity. The media for cell culture consisted of Dulbecco’s modified Eagle’s medium (DMEM) (Gibco, ThermoScientific, UK), 10% fetal calf serum (Gibco, ThermoScientific, UK), 100 units/mL penicillin and 0.1 mg/mL streptomycin (Sigma Aldrich, Gillingham, UK). All experiments were conducted in accordance with the Declaration of Helsinki and approved by the West of Scotland Research Ethics Committee (REC 19/WS/0146).
A suspension containing 2 × 10 5 TM cells was centrifuged at 1500 rpm for 5 min. The supernatant was discarded, and the cell pellet was resuspended in 100 µL of fetal calf serum. The collagen gel solution was prepared by 1 mL of Type 1 collagen (Fist Link, Wolverhampton, UK) and 160 µL of concentrated media, which consisted of 1.4 mL of DMEM 10 × (Sigma Aldrich, Gillingham, UK), 140 µL of L-glutamine (ThermoScientific, Loughborough, UK), and 360 µL of 7.5% sodium bicarbonate (Sigma Aldrich, Gillingham, UK). The pH was adjusted to 7.0 by sodium hydroxide before the cells were mixed with the collagen solution. 150 µL of the cell-gel mixture was placed in each MatTek dish and allowed to set for 10 min at 37 °C. The gels were treated with 1.5 mL of different concentrations of adrenaline (0%, 0.0005%, 0.01%) after carefully releasing the gels. The cells in the gels were observed using an Olympus CKX41 inverted microscope, and the gel photos were taken daily over 7 days and analysed using the ImageJ software. The percentage of matrix contraction was calculated using the formula: [12pt]{minimal}
$$Contraction\;=[] 100$$ C o n t r a c t i o n = A r e a o f g e l a t D a y 0 - A r e a o f g e l a t D a y n A r e a o f g e l a t D a y 0 × 100
Human TM cells were seeded at a density of 1 × 10 5 cells per well in 6-well plates and treated with different concentrations of adrenaline (0%, 0.0001%, 0.0005%, 0.001%, 0.005%, 0.01%). After 1-day treatment, cells were imaged using an Olympus CKX41 inverted microscope with Olympus CellSens Standard 1.13 software.
Human TM cells were seeded at a density of 6.25 × 10 3 cells per well in a 96-well plate and treated with different concentrations of adrenaline (0%, 0.0001%, 0.0005%, 0.001%, 0.005%, 0.01%) for 1 day. The drug solutions in the 96-well plate were then replaced by 100 µL of fresh media, followed by the addition of 20 µL of the Cell Titer 96 Aqueous one solution (Promega, Southampton, UK). The plate was incubated for 2 h at 37 °C and the absorbance was measured at 490 nm using a PHERAstar FS instrument (BMG Labtech, Aylesbury, UK). The cell viability was calculated as a percentage of the value of untreated (0%) cells.
Human TM cells were seeded at a density of 1 × 10 5 cells per well in 6-well plates and treated with different concentrations of adrenaline (0%, 0.0005%, 0.01%) for 1 day. The total RNA was extracted using a RNeasy mini kit (Qiagen, Crawley, UK) and the cDNA was synthesised using a high-capacity cDNA reverse transcription kit (ThermoScientific, Loughborough, UK). RT-qPCR was performed using the QuantiFast SYBR Green PCR kit (Qiagen, Crawley, UK) on a QuantStudio 7 Real-time PCR system. The reaction settings for 40 cycles were as follows: Holding stage: 50 °C for 2 min and 95 °C for 5 min; PCR stage: 95 °C for 5 min and 60 °C for 30 s. The sequences for the forward and reverse primers are presented in Table . The relative gene expression was calculated using the formula: [12pt]{minimal}
$$Relative\;Gene\;Expression={2}^{-( {Ct}_{target}- {Ct}_{GAPDH})}$$ R e l a t i v e G e n e E x p r e s s i o n = 2 - ( Δ Ct target - Δ Ct GAPDH )
We tested the effects of adrenaline 0.05% intraoperatively in five patients with POAG undergoing a MIGS device combined with phacoemulsification. Two MIGS devices were included in this study: iStent inject® (Glaukos, Avedro) as a Schlemm’s canal MIGS and MINIject® (iSTAR Medical SA, Wavre, Belgium) as a suprachoroidal MIGS. Informed consent was obtained from all patients. Phacoemulsification surgery was performed as per standard practice. Miochol was injected intracamerally at the end of phacoemulsification to constrict the pupil in preparation for MIGS implantation. All patients then received 0.1 mL of intracameral injection of adrenaline 0.05% before the anterior chamber was filled with viscoelastic and the MIGS device was inserted.
Clinical data, including best-corrected visual acuity (BCVA), IOP, central corneal thickness (CCT), cup-to-disc ratio (C/D), previous glaucoma surgery, previous glaucoma laser, lens status and anti-glaucoma medications were recorded preoperatively and postoperatively in all patients. Blood pressure, heart rate, and blood oxygen saturation were also measured preoperatively, intraoperatively, and postoperatively. Pupil diameter was measured using a measuring scale before and after surgery. To record the positioning of the MIGS, the intraoperative insertion of the implant was captured by video recording during surgery. Postoperatively, gonioscopy was performed in clinic and photographs of the implant in the angle were taken by slit lamp camera (ARC slit lamp imaging camera, Carleton optical, Chesham, Buckinghamshire, UK). We also performed angle scans with an anterior segment OCT (ANTERION® Heidelberg Engineering, Germany) to visualise the implant transversally.
GraphPad Prism was used for graph generation and statistical analysis. All graphs display mean and standard error of the mean (SEM). To determine statistical significance, a one-way ANOVA was performed and P values were calculated. P values, which were statistically significant, were expressed as follows: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.
Adrenaline significantly decreased matrix contraction in human TM cells A three-dimensional cell-populated collagen contraction assay was carried out to assess the contractility of TM cells after adrenaline treatment, as it has been shown to be a good in vitro model and functional assay to study tissue contraction in the eye . Increasing concentrations of adrenaline decreased the cell proliferation and cell density of TM cells in gels at both day 4 and day 7 (Fig. a). Figure b shows representative gel areas at day 4 and day 7. A dose-dependent decrease in collagen contraction was observed with increasing concentrations of adrenaline throughout the 7 days (Fig. c). At day 4, the TM cell-populated collagen gels treated with 0%, 0.0005%, and 0.01% adrenaline contracted 50.6%, 32.2%, and 5.7%, respectively, while at day 7, they contracted 87.8%, 80.6%, and 7.9%, respectively. The differences in collagen contraction were statistically significant for both 0.0005% adrenaline ( P < 0.05 on day 2, P < 0.0001 on days 3–5, P < 0.001 on days 6–7) and 0.01% adrenaline ( P < 0.05 on day 1, P < 0.01 on day 2, P < 0.0001 on days 3–7), when compared to no drug control 0%. Adrenaline did not affect cell viability and cell morphology in human TM cells The potential toxicity of adrenaline on TM cells was assessed by light microscopy observation and a cell viability assay. After 1-day treatment, the cell viability of adrenaline 0%, 0.0001%, 0.0005%, 0.001%, 0.005%, and 0.01% was 100.0%, 108.0%, 99.3%, 97.9%, 92.0%, and 100.7%, respectively. The TM cells on the 6-well plates did not display noticeable changes in cell morphology (Fig. a). All adrenaline concentrations (0.0001%, 0.0005%, 0.001%, 0.005%, 0.01%) did not exhibit a significant decrease in cell viability compared to no drug control 0% ( P > 0.05) (Fig. b). Adrenaline significantly decreased the expression of key cell cycle genes in human TM cells We tested the effects of adrenaline on the expression of key cell cycle genes in human TM cells to investigate whether adrenaline would have similar effects as previously shown in human Tenon’s fibroblasts . The expression of key cell cycle and fibrosis genes was measured by real-time qPCR after adrenaline treatment (Fig. ). In the G2 phase, ASPM ( P < 0.05), CDCA3 ( P < 0.01), CDCA8 ( P < 0.01), DBF4 ( P < 0.001), and MKI67 ( P < 0.05) genes were significantly downregulated after 0.01% adrenaline treatment when compared to no drug control 0%. The expression of DBF4 gene also showed a significant decrease with 0.0005% adrenaline ( P < 0.01). In the M phase, PLK1 ( P < 0.01), CCNB1 ( P < 0.001), CDC20 ( P < 0.01), and PRR11 ( P < 0.0001) genes were significantly downregulated after 0.01% adrenaline treatment when compared to no drug control 0%, but no significant differences were observed in the 0.0005% adrenaline group. In addition, MRTFB showed a significantly reduced gene expression with 0.01% adrenaline ( P < 0.05), and ACTA2 showed a decreased gene expression with both 0.01% and 0.0005% adrenaline ( P < 0.001). Intracameral adrenaline injection did not cause any ocular adverse effects during and after MIGS implantation We next validated our results by injecting adrenaline 0.05% intracamerally in patients with POAG receiving a Schlemm’s canal MIGS (iStent inject) or suprachoroidal MIGS (MINIject), combined with phacoemulsification surgery. Demographically, we gathered data from different age and ethnic groups, which were evenly distributed between both MIGS groups (Table ). All patients in the study were using multiple topical antiglaucoma medications, but none were on oral acetazolamide tablets. Adrenaline exhibited vasoconstrictive effects and significantly reduced intraoperative bleeding in both iStent inject and MINIject patients (Fig. a). None of the patients experienced any intraoperative or postoperative adverse effects due to the intracameral adrenaline 0.05% injection. The MINIject patients (patients 1, 2, 3) had preoperative BCVA (logMAR) of 0.0, 0.6 and 0.1, respectively. For these patients, preoperative IOP was 17, 14 and 14 mmHg, respectively, on at least two anti-glaucoma drops. CCT was found to be 513, 484 and 593 µm at baseline for these patients, and their C/D ratio was 0.8, 0.7 and 0.7, respectively (Table ). One week postoperatively, BCVA was 0.2, 0.6 and 0.0 for patients 1, 2 and 3, respectively. Postoperative IOP was 9, 15 and 16 mmHg on no antiglaucoma medications. The implants were well positioned on gonioscopy (Fig. b) and the surrounding trabecular meshwork tissues were healthy (Fig. c). The iStent inject patients (patients 4 and 5) had preoperative BCVA (logMAR) of 0.5 and 0.2, respectively. For these patients, preoperative IOP was 16 and 27 mmHg, respectively, on at least two anti-glaucoma drops. Their CCT was 546 and 583 µm at baseline, and their C/D ratio was 0.9 and 0.8, respectively (Table ). One week postoperatively, BCVA was 0.5 and 0.2, and IOP was 16 and 14 mmHg, respectively, on no antiglaucoma medications. The implants were correctly positioned (Fig. b) and the surrounding trabecular meshwork tissues were also healthy (Fig. c). Intracameral adrenaline injection did not significantly affect blood pressure, heart rate, oxygen saturation and pupil size during MIGS implantation All patients had stable blood pressures, heart rates, and oxygen saturations preoperatively, during surgery and postoperatively. Before surgery, the MINIject patients (patients 1, 2, 3) showed a blood pressure of 142/85, 124/84 and 176/80 mmHg, respectively. Their preoperative heart rate was 73, 90 and 72 beats per minute, and their oxygen saturation was 98%, 96% and 99%, respectively. The iStent inject patients (patients 4 and 5) had a preoperative blood pressure of 185/105 and 179/94 mmHg. Their heart rate was 93 and 77 beats per minute, and their oxygen saturation was 96% and 99%, respectively. During surgery, the MINIject patients (patients 1, 2, 3) showed a blood pressure of 143/74, 154/71 and 176/76 mmHg, respectively. Intraoperatively, their heart rate was 60, 93 and 70 beats per minute, and their oxygen saturation was 100%, 97% and 97%, respectively. For the patients who received iStent inject (patients 4 and 5), the blood pressure was 185/105 and 179/94 mmHg during surgery. Intraoperatively, their heart rate was 84 and 59 beats per minute, and their oxygen saturation was 96% and 99%, respectively. After surgery, the MINIject patients (patients 1, 2, 3) showed a blood pressure of 140/73, 154/88 and 176/76 mmHg, respectively. Postoperatively, their heart rate was 60, 93 and 66 beats per minute for the same patients, respectively. Their oxygen saturation after surgery was 98%, 97% and 96%, respectively. For patients who received iStent inject (patients 4 and 5), the blood pressure was 185/100 and 179/94 mmHg after surgery. Postoperatively, their heart rate was 95 and 89 beats per minute, and their oxygen saturation was 96% and 95%, respectively. All patients received intracameral miochol as standard preoperative drug before MIGS implantation. There was no significant change in pupil diameter observed intraoperatively and postoperatively after intracameral adrenaline 0.05% injection, remaining between 6 and 7 mm in diameter in all patients.
A three-dimensional cell-populated collagen contraction assay was carried out to assess the contractility of TM cells after adrenaline treatment, as it has been shown to be a good in vitro model and functional assay to study tissue contraction in the eye . Increasing concentrations of adrenaline decreased the cell proliferation and cell density of TM cells in gels at both day 4 and day 7 (Fig. a). Figure b shows representative gel areas at day 4 and day 7. A dose-dependent decrease in collagen contraction was observed with increasing concentrations of adrenaline throughout the 7 days (Fig. c). At day 4, the TM cell-populated collagen gels treated with 0%, 0.0005%, and 0.01% adrenaline contracted 50.6%, 32.2%, and 5.7%, respectively, while at day 7, they contracted 87.8%, 80.6%, and 7.9%, respectively. The differences in collagen contraction were statistically significant for both 0.0005% adrenaline ( P < 0.05 on day 2, P < 0.0001 on days 3–5, P < 0.001 on days 6–7) and 0.01% adrenaline ( P < 0.05 on day 1, P < 0.01 on day 2, P < 0.0001 on days 3–7), when compared to no drug control 0%.
The potential toxicity of adrenaline on TM cells was assessed by light microscopy observation and a cell viability assay. After 1-day treatment, the cell viability of adrenaline 0%, 0.0001%, 0.0005%, 0.001%, 0.005%, and 0.01% was 100.0%, 108.0%, 99.3%, 97.9%, 92.0%, and 100.7%, respectively. The TM cells on the 6-well plates did not display noticeable changes in cell morphology (Fig. a). All adrenaline concentrations (0.0001%, 0.0005%, 0.001%, 0.005%, 0.01%) did not exhibit a significant decrease in cell viability compared to no drug control 0% ( P > 0.05) (Fig. b).
We tested the effects of adrenaline on the expression of key cell cycle genes in human TM cells to investigate whether adrenaline would have similar effects as previously shown in human Tenon’s fibroblasts . The expression of key cell cycle and fibrosis genes was measured by real-time qPCR after adrenaline treatment (Fig. ). In the G2 phase, ASPM ( P < 0.05), CDCA3 ( P < 0.01), CDCA8 ( P < 0.01), DBF4 ( P < 0.001), and MKI67 ( P < 0.05) genes were significantly downregulated after 0.01% adrenaline treatment when compared to no drug control 0%. The expression of DBF4 gene also showed a significant decrease with 0.0005% adrenaline ( P < 0.01). In the M phase, PLK1 ( P < 0.01), CCNB1 ( P < 0.001), CDC20 ( P < 0.01), and PRR11 ( P < 0.0001) genes were significantly downregulated after 0.01% adrenaline treatment when compared to no drug control 0%, but no significant differences were observed in the 0.0005% adrenaline group. In addition, MRTFB showed a significantly reduced gene expression with 0.01% adrenaline ( P < 0.05), and ACTA2 showed a decreased gene expression with both 0.01% and 0.0005% adrenaline ( P < 0.001).
We next validated our results by injecting adrenaline 0.05% intracamerally in patients with POAG receiving a Schlemm’s canal MIGS (iStent inject) or suprachoroidal MIGS (MINIject), combined with phacoemulsification surgery. Demographically, we gathered data from different age and ethnic groups, which were evenly distributed between both MIGS groups (Table ). All patients in the study were using multiple topical antiglaucoma medications, but none were on oral acetazolamide tablets. Adrenaline exhibited vasoconstrictive effects and significantly reduced intraoperative bleeding in both iStent inject and MINIject patients (Fig. a). None of the patients experienced any intraoperative or postoperative adverse effects due to the intracameral adrenaline 0.05% injection. The MINIject patients (patients 1, 2, 3) had preoperative BCVA (logMAR) of 0.0, 0.6 and 0.1, respectively. For these patients, preoperative IOP was 17, 14 and 14 mmHg, respectively, on at least two anti-glaucoma drops. CCT was found to be 513, 484 and 593 µm at baseline for these patients, and their C/D ratio was 0.8, 0.7 and 0.7, respectively (Table ). One week postoperatively, BCVA was 0.2, 0.6 and 0.0 for patients 1, 2 and 3, respectively. Postoperative IOP was 9, 15 and 16 mmHg on no antiglaucoma medications. The implants were well positioned on gonioscopy (Fig. b) and the surrounding trabecular meshwork tissues were healthy (Fig. c). The iStent inject patients (patients 4 and 5) had preoperative BCVA (logMAR) of 0.5 and 0.2, respectively. For these patients, preoperative IOP was 16 and 27 mmHg, respectively, on at least two anti-glaucoma drops. Their CCT was 546 and 583 µm at baseline, and their C/D ratio was 0.9 and 0.8, respectively (Table ). One week postoperatively, BCVA was 0.5 and 0.2, and IOP was 16 and 14 mmHg, respectively, on no antiglaucoma medications. The implants were correctly positioned (Fig. b) and the surrounding trabecular meshwork tissues were also healthy (Fig. c).
All patients had stable blood pressures, heart rates, and oxygen saturations preoperatively, during surgery and postoperatively. Before surgery, the MINIject patients (patients 1, 2, 3) showed a blood pressure of 142/85, 124/84 and 176/80 mmHg, respectively. Their preoperative heart rate was 73, 90 and 72 beats per minute, and their oxygen saturation was 98%, 96% and 99%, respectively. The iStent inject patients (patients 4 and 5) had a preoperative blood pressure of 185/105 and 179/94 mmHg. Their heart rate was 93 and 77 beats per minute, and their oxygen saturation was 96% and 99%, respectively. During surgery, the MINIject patients (patients 1, 2, 3) showed a blood pressure of 143/74, 154/71 and 176/76 mmHg, respectively. Intraoperatively, their heart rate was 60, 93 and 70 beats per minute, and their oxygen saturation was 100%, 97% and 97%, respectively. For the patients who received iStent inject (patients 4 and 5), the blood pressure was 185/105 and 179/94 mmHg during surgery. Intraoperatively, their heart rate was 84 and 59 beats per minute, and their oxygen saturation was 96% and 99%, respectively. After surgery, the MINIject patients (patients 1, 2, 3) showed a blood pressure of 140/73, 154/88 and 176/76 mmHg, respectively. Postoperatively, their heart rate was 60, 93 and 66 beats per minute for the same patients, respectively. Their oxygen saturation after surgery was 98%, 97% and 96%, respectively. For patients who received iStent inject (patients 4 and 5), the blood pressure was 185/100 and 179/94 mmHg after surgery. Postoperatively, their heart rate was 95 and 89 beats per minute, and their oxygen saturation was 96% and 95%, respectively. All patients received intracameral miochol as standard preoperative drug before MIGS implantation. There was no significant change in pupil diameter observed intraoperatively and postoperatively after intracameral adrenaline 0.05% injection, remaining between 6 and 7 mm in diameter in all patients.
Fibrosis is the most important cause of failure after MIGS. Although the antimetabolites MMC and 5-FU are commonly used to modulate wound healing after subconjunctival glaucoma filtration surgery, they are too toxic to be used intraocularly in MIGS due to the severe sight-threatening side effects, such as hypotonous maculopathy , corneal melting and perforation , and scleral calcification . Therefore, there is a large unmet clinical need to develop an alternative and non-toxic antifibrotic drug that can be used in MIGS. Adrenaline is a safe, cost-effective, and widely available drug in ophthalmic surgery, and has been shown to have a beneficial antifibrotic effect in human Tenon’s fibroblasts . Severe fibrosis of the TM tissue can lead to sustained extracellular matrix accumulation and distortion of the TM framework, resulting in increased resistance to aqueous humour outflow and elevated IOP . It also has an adverse impact on the implantation and is the primary cause of MIGS failure . Our in vitro results indicate that adrenaline also exhibits concentration-dependent antifibrotic effects in human TM cells. Adrenaline 0.0005% and 0.01% significantly decreased the contractility of TM cells by 7.2% and 79.8%, respectively, after a 7-day contraction assay. Meanwhile, no apparent cytotoxicity was observed in TM cells after 1-day treatment with different concentrations of adrenaline (0.0001%, 0.0005%, 0.001%, 0.005%, 0.01%). These findings highlight the potential application of adrenaline in MIGS for its antifibrotic and vasoconstrictor effects, thereby enhancing the surgical success rates of MIGS. Our previous RNA-Sequencing results provide compelling evidence that high concentrations of adrenaline have an impact on cell cycle genes in human Tenon’s fibroblasts . We further examined the expression level of key cell cycle genes and fibrosis-related genes in TM cells after adrenaline treatment. In the G2 phase, ASPM , CDCA3 , CDCA8 , DBF4 , and MKI67 genes were significantly downregulated by adrenaline 0.01%. ASPM is involved in the formation of the mitotic spindle . CDCA3 is a crucial regulator of cell cycle progression from the G2 phase to mitosis . CDCA8 is also a component of the chromosomal passenger complex and is important in mitosis . DBF4 plays a vital role in DNA replication , while MKI67 is closely related to cell proliferation . In the M phase, adrenaline 0.01% significantly decreased the gene expression of PLK1 , CCNB1 , CDC20 , PRR11 . PLK1 is a significant target of the DNA damage checkpoint, allowing cell-cycle arrests at multiple points in the G2 phase and mitosis . CCNB1 is essential for the proper control of the G2/M transition phase. CDC20 also plays a vital role in both nuclear movement prior to anaphase and chromosome separation . PRR11 plays a pivotal role in the accurate regulation from the late S phase to mitosis . The downregulation of these genes indicates that adrenaline inhibits cell cycle progression and suppresses the proliferation of TM cells, causing them to enter a growth arrest phase without undergoing cell death. Moreover, adrenaline 0.01% significantly reduced the expression of MRTFB and ACTA2 genes, while ACTA2 also exhibited decreased level after treatment with adrenaline 0.0005%. MRTFB is one of the major regulators of cytoskeletal gene expressions and is essential for TGF-β-induced fibrosis . ACTA2 is an important downstream gene of the MRTFB/SRF pathway and encodes α-SMA, which plays an important role in myofibroblast differentiation . In this study, the lower expression of MRTFB and ACTA2 genes after adrenaline treatment was consistent with the decreased matrix contraction, indicating a decrease in fibrosis development in TM cells. To validate the feasibility of adrenaline application in MIGS, we further examined the safety and efficacy of adrenaline in POAG patients undergoing Schlemm’s canal MIGS (iStent inject) or suprachoroidal MIGS (MINIject), when combined with phacoemulsification surgery. Different concentrations of adrenaline have been applied in the clinical setting. Adrenaline can be injected into the anterior chamber or added in the infusion solution during cataract surgery , in order to achieve adequate and sustained pupillary dilation. When used in combination with atropine, adrenaline is also effective in the management of intraoperative floppy-iris syndrome due to a powerful synergistic effect on iris dilation. Adrenaline 0.1% exhibits a good safety profile with no impact on blood pressure and heart rate , and adrenaline 0.02% does not increase the risk of postoperative macular oedema . Based on these previous studies, we selected an adrenaline concentration of 0.05% for intracameral injection (0.1 mL) in patients just prior to MIGS implantation in this study. None of the patients experienced any intraoperative or postoperative adverse reactions. Patients receiving these two MIGS devices all exhibited stable blood pressures, heart rates, and oxygen saturations during and after surgery, demonstrating a good safety profile of intracameral adrenaline 0.05%. Patient 2 showed a 30 mmHg increase in blood pressure from preoperative stage to intraoperative stage, which was related to the patient being very anxious. Although adrenaline may induce adverse cardiovascular effects, its systemic absorption is limited , likely due to the presence of the blood-retinal barrier. Nonetheless, it remains contraindicated in patients with hypertension, heart disease, arrhythmias, and narrow angle glaucoma, and the use of adrenaline in these patients requires thorough patient assessment and careful patient selection . The study had a few limitations. While transformed TM cells were used in this study, they may differ from primary TM cells in specific responses. Thus, it would be valuable to include primary TM cells in future in vitro studies. Although the data were carefully collected and analysed, this study was a proof-of-concept study and the sample size of patients was relatively small. With a known genetic variation in the alpha1B-adrenergic receptor , ethnic differences in adrenaline sensitivity also require further consideration. Additionally, the findings presented are specific to POAG patients and two MIGS devices (iStent inject and MINIject), which may not be directly transferable to other MIGS devices. Including individuals of different ethnicities and patients with other types of glaucoma, as well as incorporating a broader range of MIGS devices available on the market, could provide a more comprehensive evaluation for the use of adrenaline in MIGS in the future. Furthermore, this is a short-term study and the early incidence of stent failure is lower than the later failure rate. Currently, it is unknown whether adrenaline affects the cell cycle of other eye cells, such as potentially blocking any stem cell repopulation effects. Further research is needed to determine if this effect could lead to adverse consequences. It also remains unclear whether adrenaline is associated with a better reduction in IOP over time, which requires longer prospective studies. This study is focused on the early fibrotic responses, and moving beyond short-term effectiveness will be crucial to examine longer-term impacts. Given that endothelin-1 (ET-1) is a vasoconstrictor produced by endothelial cells, and that adrenaline can enhance its secretion, measuring ET-1 levels in the aqueous humour might help further understand the mechanism of adrenaline’s vasoconstrictive effects and its impact on aqueous humour outflow in the future. In conclusion, this study showed that adrenaline reduced the contractility of TM cells in a dose-dependent manner, and suppressed the expression of key cell cycle and fibrosis genes with no significant cytotoxicity. The intracameral injection of adrenaline 0.05% in patients undergoing MIGS implantation also demonstrated a good safety profile during and after surgery. Unless contraindicated, we therefore recommend intracameral injections of adrenaline 0.05% as a cheap and safe drug to be used just before MIGS insertion, as it decreases the risk of bleeding from the trabecular meshwork and also exhibits antifibrotic effects by arresting the cell cycle, thereby increasing the postoperative success rates in MIGS.
|
Complex challenges of estimating the age and vitality of muscle wounds: a study with matrix metalloproteinases and their inhibitors on animal and human tissue samples | 299e1f2d-919c-40d2-a904-917d99185422 | 8354971 | Pathology[mh] | Reconstructing events of physical violence by evaluating the point of time when wounds have been inflicted is a recurring task in forensic routine work. Not surprisingly, wound age estimation has been a central field of forensic research for a long time. Once injured by force, the body tissues respond with numerous molecular and cellular reactions in order to fix the damage: Wound healing is activated. The chronological course of wound healing can be divided in different phases, which are characterized by various processes and mediators . Estimating the wound age and/or the vitality of wounds is based on identifying the phases of wound healing by detecting these processes/mediators and putting them into a temporal context. Besides evaluating the macroscopic appearance of an injury, wound age estimation also comprises an assessment of microscopic and molecular alterations. A recent review by Li et al. stated that progress towards a more precise estimation of vitality and the age of wounds has been made during the last years. However, the authors also point out that an unrestrictedly reliable marker or set of markers has not been identified yet. Research problems are mainly seen in the availability and quantity of human tissue samples with sufficient information on wound age and wound vitality, as well as the results’ reproducibility, the examiner’s experience and methodological limitations. Casse et al. arrive at a similar conclusion. According to their study, research on wound age estimation and wound vitality demands the consideration of various factors to ensure not only high specificity and sensitivity, but also the reproducibility of the results. They underline the necessity of control groups and the differentiation between antemortem and postmortem wounds with a simultaneously adequate number of samples. This necessity derives from the phenomenon of the so-called biological death, meaning the death of each single cell a certain time after the death of the organism, the “individual death”. In 2010, Alaeddini et al. stated that the interval of survival varies in different body tissues due to their own survival mechanisms. In this time span, some physiologic processes might still go on and bias the estimation of wound age. They might even suggest vitality of an injury that has actually been inflicted postmortem. A review by Dunjić et al. also suggests that the activity of cells after the individual death depends on the type of tissue. Consequently, transferring research findings from a specific type of wound to other body tissues is nearly impossible . A recent study of our own research group on myocardial infarctions came up with positive immunohistochemical staining for matrix metalloproteases (MMP) 2 and 9 and their inhibitor TIMP-1 (tissue inhibitor of matrix metalloproteases) not only in ischemic areas, but also adjacent to wounds inflicted mechanically by electrodes or vessel ligations in rats’ hearts . In contrast, myocardial samples without an injury, ischemia, or other determinable harms (increased workload due to pulmonary embolism, cardiac resuscitation) represented no (increased) staining. MMPs are zinc-requiring proteolytic enzymes that are synthetized as zymogens, meaning in a proactive form, and are activated after an injury . In healthy tissue, MMPs are expressed continuously on a low level. Their expression increases when tissues are being remodeled , in physiologic as well as pathologic processes . They play a central role especially in the early phase of wound healing by degrading extracellular matrix (EM) . In humans, more than twenty different, class-divided MMPs have been identified , one important sub-group are the gelatinases (MMP-2 and MMP-9). Due to repeats homologous to fibronectin type II in their catalytic domains, they specifically degrade different types of collagens in the EM . High levels of MMP-2 and MMP-9 were detected in wounds after surgery and in chronic wounds . Furthermore, studies on skin samples and studies on ovarian carcinoma cells have shown that MMP-9 influences wound healing by activating TGF-beta via proteolysis and inducing the expression of vascular endothelial growth factor (VEGF) . A study on MMP-2 knockout mice was able to show that MMP-2 plays a key role in angiogenesis and tumor progression . The activation of MMPs is complex and strictly regulated on multiple stages . The primary control instance are TIMPs, a family of four enzymes (TIMP-1 to TIMP-4). They inhibit MMPs and thus inhibit the degradation of EM. TIMP-1 has a high affinity to MMP-9 and its inactive form, progelatinase B . The interaction of MMPs and TIMPs seems to be important for wound healing . In a review, Conlon et al. stated that MMPs not only act as proteolytic enzymes and inductors of the expression of signal molecules; moreover, they induce the activation of other MMPs. The different ways of MMP activation seem to intertwine. Previous studies also indicate that the ratio between MMPs and TIMPs is of great importance for wound healing. Ladwig et al. showed that the ratio of MMP-9 and TIMP-1 could be used as an indicator of wound healing in wound fluid of pressure ulcers. A dysregulation of MMPs and TIMPs seems to be one of the reasons why wound healing is defective in chronic wounds . In a study on dermal wounds, Gillard et al. discovered elevated expression of MMP-9, MMP-2, and TIMP-1 especially in early phases of wound healing, concluding that MMP-9 might be important for angiogenesis, whereas MMP-2 might play a role in tissue transformation. In addition to our own research results , all these findings suggest that a closer look on the applicability of MMPs and TIMPs in the context of wound age estimation and wound vitality could be worthwhile. Though promising, our findings in the preceding study only gave a hint that MMPs and TIMPS can be detected immunohistochemically in the early phase of wound healing of myocardial tissue—other important questions, however, remained unanswered: Can MMP-2, MMP-9, and TIMP-1 be detected immunohistochemically not only in injured myocardium, but also in skeletal muscle? If so, does their occurrence depend on the “age” of the examined wounds? Are there differences between the two types of muscle tissue? Can the markers help to differentiate between vital and postmortem-inflicted wounds? We aimed on addressing these questions with a broad and complex approach. The immunohistochemical detectability of MMP-2, MMP-9, and TIMP-1 was examined in two types of injured muscle tissue, myocardium and skeletal muscle. Moreover, we worked with postmortem drawn, human tissue samples and with an animal model, the isolated perfused Langendorff heart. This model allowed us to generate myocardial wounds with a defined “age” as well as postmortem-inflicted wounds. Animal experiments were performed in accordance with the German legislation on protection of animals and the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health (NIH Publication No. 85–23, revised 1996). The protocol for the Langendorff system was approved by the local Animal Ethics Committee (project no. O 27/11). The examination of human myocardium and skeletal muscle samples drawn during autopsies was approved by the ethical committee of the Medical Faculty of the Heinrich-Heine-University Düsseldorf (project no. 5833). Human study samples A total of 208 tissue samples of muscle wounds, skeletal muscle (140 samples) and myocardium (68 samples), were selected from 141 autopsies at the Department of Legal Medicine at the University Hospital Düsseldorf, Germany, in the period from 2006 to 2017. We included different types of violence (strangulation, blunt force, sharp force, polytrauma, myocardial injuries due to surgery, and myocardial injuries due to infarction). The age of the tissue donors ranged between 16 months and 94 years and both sexes were included. In a first step, wound age of each sample was estimated roughly according to the available data (assumed time period between infliction of wound and death of the individual). The estimate was refined by additionally considering hematoxylin&eosin (HE) staining results and classifying the findings according to Cummings et al. : A: very short survival time, few min max.—no signs of inflammation, no neutrophilic infiltration B: few min up to 4 h—single perivascular neutrophils C: 4 h up to 8 h—enhanced neutrophilic infiltration D: 8 h up to 12 h—infiltration of neutrophils, macrophages, and fibroblasts Isolated perfused Langendorff heart We used white male Wistar rats aged 2–3 months. The weight ranged between 250 and 350 g. The preparation of the rats’ hearts was performed as described before : The rats were kept on a 12:12 light/dark schedule (lights on at 0600 h) with food and water ad libitum. The animals were anesthetized by intraperitoneal injection of Pentobarbital (90 mg kg −1 ) and Heparin (0.2 ml). The depth of sedation was verified by the absence of reactions to pain. In this state, the rats were decapitated, an immediate thoracotomy was conducted and hearts were excised and mounted onto the Langendorff system. The hearts were perfused with modified Krebs–Henseleit-Buffer: 118 mM sodium chloride (VWR Chemicals Prolabo) 4.7 mM potassium chloride (Fluka) 1.2 mM magnesium sulfate hepta-hydrate (Sigma-Aldrich) 1.2 mM potassium dihydrogen phosphate (Merck) 25 mM sodium hydrogen carbonate (Roth) 0.5 mM ethylenediaminetetraacetic acid (Roth) 11 mM D-glucose (VWR Life Science and Roth) 1 mM L-lactic acid sodium salt (Serva) 2.25 mM calcium chloride (Merck) Heart function was monitored by observing heart rate, intraventricular pressure, and electrocardiogram (ECG). For data digitalization, we used an analog to digital converter (PowerLab/8SP, ADInstruments Pty Ltd., Castle hill, Australia) with a sampling rate of 500 Hz. Data documentation was carried out frequently by using Chart for Windows v5.0 (AD-Instruments). Rats’ hearts with vital wounds After a stabilization period of about 20 min on the Langendorff system, 16 hearts were injured by stabbing the wall of the left chamber with a scalpel. After defined time intervals of 5, 10, 15, 30, 60, 120, 180, and 240 min, the hearts were removed from the Langendorff system and directly immersed in 4% formalin. An overview of the study protocol is given in Fig. . Rats’ hearts with postmortem-inflicted wounds Eight hearts were excised after decapitation without being attached to the Langendorff system. After a defined time interval, they were injured by stabbing the wall of the left chamber with a scalpel: Two hearts each were stitched immediately, 10 min and 180 min after they had stopped beating; after another 180 min, the six hearts were fixed in 4% formalin. Furthermore, two hearts were injured 20 min after they had stopped beating and were fixed in 4% formalin after another 240 min (see also Fig. ). Immunohistochemical analysis Identical staining methods were used for both human study samples and rat study samples and were performed as described before by Mayer et al. : Tissue sections were deparaffinized, washed in distilled water three times for 5 min and washed in TBS buffer with 0.5% Tween 20 two times for 5 min. MMP-2: Slides were boiled in citrate buffer pH 6.0 for 10–15 min, cooled, and then washed in distilled water two times for 5 min. Primary antibody against MMP-2 (Medac, rabbit, E 18,012) was used in a concentration of 1:200, and the slides were incubated over night at + 4 C. Slides were washed in TBS buffer with 0.5% Tween 20 two times for 5 min. Endogenous peroxidase was blocked with 0.03% H2O2 for 20–25 min. MMP-9: Slides were treated with proteinase K (Dako, S3020) for 7 min and washed two times in TBS buffer with 0.5% Tween 20 for 5 min. Primary antibody against MMP-9 (Biorbyt orb, rabbit, 13,583) was used in a concentration of 1:300, and the slides were incubates over night at + 4 C. Slides were washed in TBS buffer with 0.5% Tween 20 two times for 5 min. Endogenous peroxidase was blocked with 0.03% H2O2 for 10–15 min. TIMP-1: Primary antibody against TIMP-1 (Biorbyt orb, rabbit, 195,994) was used in a concentration of 1:300, and the slides were incubated over night at + 4 C. Slides were then washed in TBS buffer with 0.5% Tween 20 two times for 5 min. Endogenous peroxidase was blocked with 0.03% H2O2 for 10–15 min. After blocking of endogenous peroxidase, all slides were washed in distilled water two times for 5 min and then in TBS buffer with 0.5% Tween 20 two times for 5 min. Afterwards, all slides were incubated with a peroxidase-marked polymer (Medac, Histofine1 Simple Stain MAX PO against rabbit, 414,142) for 30 min. Slides were stained with AEC (3-Amino-9-Ethylcarbazole, Cohesion Biosciences) and counterstained with Mayers hematoxylin (Merck, HX87717149). Evaluation system for immunohistochemical analysis To standardize the results of the immunohistochemical analysis, we used the following evaluation system as published before : MMP-9 and TIMP-1: 0: No visible staining I: Positive staining of single cells II: Positive staining of cell groups III: Positive staining of large tissue areas MMP-2: 0: No visible staining I: Positive staining of EM in the perivascular regions II: Positive staining of EM in larger areas III: Positive staining not only of EM but also of intracellular A total of 208 tissue samples of muscle wounds, skeletal muscle (140 samples) and myocardium (68 samples), were selected from 141 autopsies at the Department of Legal Medicine at the University Hospital Düsseldorf, Germany, in the period from 2006 to 2017. We included different types of violence (strangulation, blunt force, sharp force, polytrauma, myocardial injuries due to surgery, and myocardial injuries due to infarction). The age of the tissue donors ranged between 16 months and 94 years and both sexes were included. In a first step, wound age of each sample was estimated roughly according to the available data (assumed time period between infliction of wound and death of the individual). The estimate was refined by additionally considering hematoxylin&eosin (HE) staining results and classifying the findings according to Cummings et al. : A: very short survival time, few min max.—no signs of inflammation, no neutrophilic infiltration B: few min up to 4 h—single perivascular neutrophils C: 4 h up to 8 h—enhanced neutrophilic infiltration D: 8 h up to 12 h—infiltration of neutrophils, macrophages, and fibroblasts We used white male Wistar rats aged 2–3 months. The weight ranged between 250 and 350 g. The preparation of the rats’ hearts was performed as described before : The rats were kept on a 12:12 light/dark schedule (lights on at 0600 h) with food and water ad libitum. The animals were anesthetized by intraperitoneal injection of Pentobarbital (90 mg kg −1 ) and Heparin (0.2 ml). The depth of sedation was verified by the absence of reactions to pain. In this state, the rats were decapitated, an immediate thoracotomy was conducted and hearts were excised and mounted onto the Langendorff system. The hearts were perfused with modified Krebs–Henseleit-Buffer: 118 mM sodium chloride (VWR Chemicals Prolabo) 4.7 mM potassium chloride (Fluka) 1.2 mM magnesium sulfate hepta-hydrate (Sigma-Aldrich) 1.2 mM potassium dihydrogen phosphate (Merck) 25 mM sodium hydrogen carbonate (Roth) 0.5 mM ethylenediaminetetraacetic acid (Roth) 11 mM D-glucose (VWR Life Science and Roth) 1 mM L-lactic acid sodium salt (Serva) 2.25 mM calcium chloride (Merck) Heart function was monitored by observing heart rate, intraventricular pressure, and electrocardiogram (ECG). For data digitalization, we used an analog to digital converter (PowerLab/8SP, ADInstruments Pty Ltd., Castle hill, Australia) with a sampling rate of 500 Hz. Data documentation was carried out frequently by using Chart for Windows v5.0 (AD-Instruments). After a stabilization period of about 20 min on the Langendorff system, 16 hearts were injured by stabbing the wall of the left chamber with a scalpel. After defined time intervals of 5, 10, 15, 30, 60, 120, 180, and 240 min, the hearts were removed from the Langendorff system and directly immersed in 4% formalin. An overview of the study protocol is given in Fig. . Eight hearts were excised after decapitation without being attached to the Langendorff system. After a defined time interval, they were injured by stabbing the wall of the left chamber with a scalpel: Two hearts each were stitched immediately, 10 min and 180 min after they had stopped beating; after another 180 min, the six hearts were fixed in 4% formalin. Furthermore, two hearts were injured 20 min after they had stopped beating and were fixed in 4% formalin after another 240 min (see also Fig. ). Identical staining methods were used for both human study samples and rat study samples and were performed as described before by Mayer et al. : Tissue sections were deparaffinized, washed in distilled water three times for 5 min and washed in TBS buffer with 0.5% Tween 20 two times for 5 min. MMP-2: Slides were boiled in citrate buffer pH 6.0 for 10–15 min, cooled, and then washed in distilled water two times for 5 min. Primary antibody against MMP-2 (Medac, rabbit, E 18,012) was used in a concentration of 1:200, and the slides were incubated over night at + 4 C. Slides were washed in TBS buffer with 0.5% Tween 20 two times for 5 min. Endogenous peroxidase was blocked with 0.03% H2O2 for 20–25 min. MMP-9: Slides were treated with proteinase K (Dako, S3020) for 7 min and washed two times in TBS buffer with 0.5% Tween 20 for 5 min. Primary antibody against MMP-9 (Biorbyt orb, rabbit, 13,583) was used in a concentration of 1:300, and the slides were incubates over night at + 4 C. Slides were washed in TBS buffer with 0.5% Tween 20 two times for 5 min. Endogenous peroxidase was blocked with 0.03% H2O2 for 10–15 min. TIMP-1: Primary antibody against TIMP-1 (Biorbyt orb, rabbit, 195,994) was used in a concentration of 1:300, and the slides were incubated over night at + 4 C. Slides were then washed in TBS buffer with 0.5% Tween 20 two times for 5 min. Endogenous peroxidase was blocked with 0.03% H2O2 for 10–15 min. After blocking of endogenous peroxidase, all slides were washed in distilled water two times for 5 min and then in TBS buffer with 0.5% Tween 20 two times for 5 min. Afterwards, all slides were incubated with a peroxidase-marked polymer (Medac, Histofine1 Simple Stain MAX PO against rabbit, 414,142) for 30 min. Slides were stained with AEC (3-Amino-9-Ethylcarbazole, Cohesion Biosciences) and counterstained with Mayers hematoxylin (Merck, HX87717149). To standardize the results of the immunohistochemical analysis, we used the following evaluation system as published before : MMP-9 and TIMP-1: 0: No visible staining I: Positive staining of single cells II: Positive staining of cell groups III: Positive staining of large tissue areas MMP-2: 0: No visible staining I: Positive staining of EM in the perivascular regions II: Positive staining of EM in larger areas III: Positive staining not only of EM but also of intracellular General observations Whereas positive staining for MMP-9 and TIMP-1 was found strictly intracellular, positive staining reactions for MMP-2 were also found in the EM. These findings correlate with the results of Mayer et al. . Slides from human samples that had been stored in formalin for a longer time showed less distinct staining results compared to those samples collected more recently (2015–2017). In order to exclude a relevant influence of storage time on our results, we evaluated the slides of the “older” cases separately and compared them to the “younger” ones without finding any differences. We also checked, if the causative type of violence has an impact on the occurrence of the markers, again, no differences were found. Therefore, the results in this publication comprise all collected samples without separating them into the different types of violence for a clearer depiction. Human skeletal muscle Table is enclosed for detailed results; examples for staining results are presented in Fig. . MMP-9Positive staining results for MMP-9 were found in all wound age groups, even in samples with very short survival times. The intensity of staining was mostly equivalent to grade II or even III. However, there was also a considerable number of samples showing no positive staining at all. The share of samples with negative staining was especially large in wound age group D. TIMP-1The majority of samples of wound age groups A and B presented positive staining results with an intensity according to grade II or III. In wound age groups C and especially D, the share of samples with negative staining was larger. MMP-2Strong positive staining results for MMP-2 according to grades II and III were found in all samples regardless of the wound age group. Negative staining was found in age groups B to D but their share was rather small. Human myocardium Table presents the detailed results for human myocardium injuries. Examples for staining results can be found in Fig. . MMP-9A high number of samples in wound age group A showed staining results with intensities grade II and III. The share of grade III was especially high in wounds that followed an infarction. However, nearly half of the samples in group A showed no positive staining. Similar findings could be observed for group B. The single sample in group D presented no positive staining. TIMP-1The majority of samples of wounds in group A showed results with staining intensities grade II and III. Again, the share of grade III was especially high in wounds that followed an infarction. Also, a considerable number of samples showed no positive staining at all. The same accounted for samples in group B. The one sample in group D presented a staining intensity grade I. MMP-2Staining results in group A mainly presented intensities grade II and III with a considerable high share of grade III in infarction-derived wounds. In group B, the share of samples with a staining intensity grade I was greater. The sample in group D also presented positive staining with intensity grade I. Negative staining results for MMP-2 were only found in single cases. Rats’ hearts—vital wounds Table shows the detailed results of rat hearts with vital wounds. MMP-9 Only two positive staining results with intensity grade I were observed in one heart with a survival time of 30 min and in one heart with a survival time of 3 h. In all other cases, staining was negative. TIMP-1 There was only one heart with a survival time of 1 h that showed discreet positive staining. All other hearts showed no positive staining at all. MMP-2 Nine hearts showed positive staining results with intensities grade I and especially grade II. The shortest survival time with positive staining results was 15 min. From the hearts with longer survival times, only one heart with a survival time of 1 h stained positive. Rats’ hearts—postmortem-inflicted wounds Table shows the detailed results of rats’ hearts with postmortem-inflicted wounds, and examples of staining results are presented in Fig. . MMP-9 Three hearts presented positive staining results with intensities grade I and II. The time spans between the end of heartbeat and the infliction of the wounds varied between 0 min and 3 h In all cases, and the time span after wound infliction was 3 h. MMP-2 Three hearts presented negative staining results, including the two hearts with wounds inflicted 20 min after the heart had stopped beating. All the other hearts showed positive staining with intensities grade I and especially grade II. TIMP-1 None of the hearts presented positive staining results. Whereas positive staining for MMP-9 and TIMP-1 was found strictly intracellular, positive staining reactions for MMP-2 were also found in the EM. These findings correlate with the results of Mayer et al. . Slides from human samples that had been stored in formalin for a longer time showed less distinct staining results compared to those samples collected more recently (2015–2017). In order to exclude a relevant influence of storage time on our results, we evaluated the slides of the “older” cases separately and compared them to the “younger” ones without finding any differences. We also checked, if the causative type of violence has an impact on the occurrence of the markers, again, no differences were found. Therefore, the results in this publication comprise all collected samples without separating them into the different types of violence for a clearer depiction. Table is enclosed for detailed results; examples for staining results are presented in Fig. . MMP-9Positive staining results for MMP-9 were found in all wound age groups, even in samples with very short survival times. The intensity of staining was mostly equivalent to grade II or even III. However, there was also a considerable number of samples showing no positive staining at all. The share of samples with negative staining was especially large in wound age group D. TIMP-1The majority of samples of wound age groups A and B presented positive staining results with an intensity according to grade II or III. In wound age groups C and especially D, the share of samples with negative staining was larger. MMP-2Strong positive staining results for MMP-2 according to grades II and III were found in all samples regardless of the wound age group. Negative staining was found in age groups B to D but their share was rather small. Table presents the detailed results for human myocardium injuries. Examples for staining results can be found in Fig. . MMP-9A high number of samples in wound age group A showed staining results with intensities grade II and III. The share of grade III was especially high in wounds that followed an infarction. However, nearly half of the samples in group A showed no positive staining. Similar findings could be observed for group B. The single sample in group D presented no positive staining. TIMP-1The majority of samples of wounds in group A showed results with staining intensities grade II and III. Again, the share of grade III was especially high in wounds that followed an infarction. Also, a considerable number of samples showed no positive staining at all. The same accounted for samples in group B. The one sample in group D presented a staining intensity grade I. MMP-2Staining results in group A mainly presented intensities grade II and III with a considerable high share of grade III in infarction-derived wounds. In group B, the share of samples with a staining intensity grade I was greater. The sample in group D also presented positive staining with intensity grade I. Negative staining results for MMP-2 were only found in single cases. Table shows the detailed results of rat hearts with vital wounds. MMP-9 Only two positive staining results with intensity grade I were observed in one heart with a survival time of 30 min and in one heart with a survival time of 3 h. In all other cases, staining was negative. TIMP-1 There was only one heart with a survival time of 1 h that showed discreet positive staining. All other hearts showed no positive staining at all. MMP-2 Nine hearts showed positive staining results with intensities grade I and especially grade II. The shortest survival time with positive staining results was 15 min. From the hearts with longer survival times, only one heart with a survival time of 1 h stained positive. Table shows the detailed results of rats’ hearts with postmortem-inflicted wounds, and examples of staining results are presented in Fig. . MMP-9 Three hearts presented positive staining results with intensities grade I and II. The time spans between the end of heartbeat and the infliction of the wounds varied between 0 min and 3 h In all cases, and the time span after wound infliction was 3 h. MMP-2 Three hearts presented negative staining results, including the two hearts with wounds inflicted 20 min after the heart had stopped beating. All the other hearts showed positive staining with intensities grade I and especially grade II. TIMP-1 None of the hearts presented positive staining results. We addressed the challenges that go along with forensic wound age estimation as illustrated above by a unique sample collection comprised in the study: Not only were we able to include two different types of tissues (skeletal muscle and myocardium) from two species (humans, rats), the sample set also includes wounds with a defined wound age and reliably postmortem-inflicted wounds thanks to the Langendorff system. Thus, we not only took advantage of controlled experimental conditions in the rat model but also examined the applicability of the results gained on human tissue samples. Despite this broad approach, the presented results are very inhomogeneous and show great “scattering”. We found positive staining for all the tested markers in a considerable number of human samples regardless of their origin and the wound age. However, the same accounts for negative staining results (see Fig. for examples of human skeletal muscle). If any, there was a slight tendency of more intense staining results towards cases with a “younger” wound age: Apparently, the number of samples showing no positive staining or less intense staining increased with higher wound age, implying that the markers we evaluated occur shortly after the infliction of a wound and disappear rather fast. Similar findings have also been published by Wang et al. : In skin wounds of mice, high levels of MMP-9/MMP-2 seem to suggest an earlier stage of wound healing. The authors came to the conclusion that an increased expression and activation of MMP-2 might be important for the inflammation phase following an injury rather than being crucial for the process of wound healing. Regarding the results of human samples with wound age group A (very short survival time, few min max.), the share of those with staining intensities grade III was slightly higher in heart muscle samples than in skeletal muscle samples. This might lead to the assumption that wound healing and the emergence of the evaluated markers kick in faster in myocardium. However, a closer look on the myocardium samples revealed that the high staining intensities are mainly found in infarction-derived injuries. Though the differences were rather discreet, it is still obvious that staining of MMP-2, MMP-9, and TIMP-1 in younger wound age groups was more intense in cases with an underlying infarction compared to those with injuries of other origins. Therefore, we assume that in infarctions, the already existing inflammation might have caused an earlier expression and/or activation of the markers, which goes along with other research findings . Since particularly inner organs, but also skin in certain cases, might present acute or chronic illnesses, statements on wound age based on the detection of an inflammation-related marker have to be made cautiously. Referring to the review by Li et al. , it appears that “[…] wound age estimation is an intricate and multifactorial problem […]” which means that numerous intrinsic and extrinsic factors must be taken into account when trying to determine the age and vitality of a wound. Compared to human samples, rat hearts with vital wounds presented rather few positive staining results especially for markers MMP-9 and TIMP-1. In addition, positive staining was found earliest in cases with a wound age of 15 min. In many of the human cases, a shorter survival time has to be assumed. At least in part this alleged contradiction might be connected to the fact that tissues/cells can “survive” the death of the individual for some time. In case of the rats’ hearts, “wound age” refers to the time between infliction of wounds and fixation of the hearts in formalin. Under such circumstances, the death of all cells occurs almost simultaneously and at the same point of time as the death of the individual, i.e. the heart. Since such a sudden stop of all intracellular activities does not apply for the human cases, wound age processes that have been triggered at or shortly before the time of death might still have proceeded and caused the emergence of the investigated markers some time later—simulating a faster occurrence in human tissue. In addition, the lack of blood in the Langendorff system might result in a delayed activation and/or expression of MMPs and TIMPs since relevant mediators might be missing. Dunjic et al. already stated that after the individual death of a person, some cells are still active. Referring to this review, the time span during which the cells are active also seems to depend upon the type of tissue. According to Tsujimoto et al. , the time of cell death is also influenced by ATP levels. Fibroblasts in human skin samples could be analyzed several days postmortem . In this context, supravital reactions even several hours after the individual death can be explained. White blood cells seem to remain active for up to 12 h postmortem, which questions the presence of an inflammation as a vital reaction after injury. Additionally, Alaeddini et al. and Jennings et al. also described different intervals of survival due to different tissue mechanisms and stated that necrosis starts in defined regions of every organ, such as the subendocardial regions in the human heart. Compared to other organs, skeletal muscle tissue seems to show postmortem ultrastructural changes quite late . In a study on lamb muscle, Sylvestre et al. were able to provide evidence of postmortem activity of MMP-2. High levels of pro-MMP-2 and of active MMP-2, but also of active MMP-2, were detected not only on the day of slaughter, but also 21 days later in samples that had been stored at 4 °C. The high levels of MMP-2 led to the assumption that MMP-2 is involved in the degradation of the tissue. The problem of distinguishing between vital wounds and postmortem-inflicted wounds was also described in a review by Cecchi et al. in 2010 who pointing out that a variability in their detection makes many markers unreliable when it comes to this question. Furthermore, the methods used for detecting a marker, e.g. polymerase chain reaction or IHC, seem to have an impact upon the results. Against this background, the behavior of the markers tested in our study is not surprising. The occurrence of MMP-2 and MMP-9 in vital and postmortem-inflicted wounds does not show obvious differences. Merely for TIMP-1, some interesting results were obtained: Although there were many positive staining results in human samples, rats’ hearts with vital wounds presented positive staining for TIMP-1 only in one case with a wound age of 1 h. Furthermore, there were no positive staining results for TIMP-1 in rats’ hearts with postmortem-inflicted wounds. The overall picture of these findings suggests that TIMP-1 is not as sensitive as MMP-2 and MMP-9, implying a possible use as a vitality marker. To verify this hypothesis, TIMP-1 needs to be tested on human muscle samples with reliably postmortem-inflicted wounds. Unfortunately, such samples are difficult to obtain and were therefore not comprised in the study at hand. Our study is subjected to some limitations: Our collected samples of human tissue only include wounds with a wound age up to 12 h. The maximum post-infliction time span of rats’ hearts accounted for 240 min. We therefore have no information about the behavior of the markers when used on wounds that are days or even weeks old. In addition, the available data for the tissue samples drawn during autopsies underlie some uncertainties, especially with a view to the exact time of the infliction of the wounds. Uninjured control samples were already included in the preceding study , whereas samples of human skeletal muscle with postmortem-inflicted wounds and a reasonable postmortem interval are difficult to obtain and therefore could not be examined. Furthermore, the number of rat tissue samples examined in this study might seem comparably low. We resigned from increasing the number due to ethical aspects. Our study again demonstrates the challenges that go along with forensic wound age estimation and the establishment of new markers. Despite our complex approach of examining MMP-9, MMP-2, and TIMP-1 on both human and rat muscle tissue, as well as on vital and postmortem-inflicted wounds, we were faced with disappointing results. Though unexpected findings in a preceding study on rat hearts were quite promising, the results of the far more comprehensive sample collection show a very inhomogeneous picture leading to the conclusion that the markers do not meet the complex requirements of forensic wound age and wound vitality estimation. Only TIMP-1 might be of use when trying to differentiate between vital and postmortem-inflicted wounds but it needs to be tested on postmortem-inflicted wounds of human muscle samples. Overall, it became clear again that a profound understanding of the usefulness of potential markers can only be achieved by examining a variety of samples. The sample collection needs to include vital wounds and postmortem-inflicted wounds. When working with an animal model, human control samples are indispensable; otherwise, the transferability of results remains questionable. The same accounts for different types of tissues. Finally, potential influences of acute or chronic illnesses have to be kept in mind when interpreting analytical results. |
Effect of Surface Treatments of Polyetherketoneketone as a Post Material on Shear Bond Strength to Root Dentin using Two Types of Resin Cement | ad9aad74-bb93-48ba-8ed5-77bb2e46ac9b | 11734299 | Dentistry[mh] | Preparation of Shear Bond Strength Test Specimens shows a schematic diagram of test specimen preparation for the SBS test. In this study, 100 bovine anterior teeth were used. Frozen anterior teeth were thawed, and soft tissues still attached to the surfaces of the roots were removed. First, the bovine anterior tooth was split into a crown and a root portion at the cementoenamel junction, the tooth pulp was removed from the root canal, and the root was cut in half along its long axis . Second, the bovine root half was placed into a 2.54-cm epoxy resin ring, with the cut plane facing the top of the ring, and embedded in epoxy resin (Scandiplex, Fritsch; Hagen, Germany; ). The root was polished with a 120-grit waterproof abrasive paper until the surface of the root dentin was flattened, then further polished with abrasive papers up to 600 grit . The dentin surface was treated with 18% ethylenediaminetetraacetic acid solution (Ultradent EDTA18%, Ultradent; South Jordan, UT, USA) for 20 s and then washed in distilled water. The root was dried with air. All treated specimens were stored in a moist environment at 37°C and 95% humidity for 1 week; thereafter, the specimens were taken out and washed in distilled water. lists the materials used in this study and their composition. A 12-mm-diameter Pekkton ivory press ingot (Pekkton, Cendres + Metaux SA; Biel/Bienne, Switzerland), as the PEKK adherend was cut into 1-mm-thick slices using a blade with a thickness of 300 μm (Serge Microtome, SP1600, Leica; Wetzlar, Germany) to prepare 100 specimens. The prepared specimens were polished with waterproof abrasive papers up to 600 grit. Then, the specimens were ultrasonically cleaned and left to air dry. Of the 100 specimens, 60 were mechanically pretreated by sandblasting with 50-µm Al 2 O 3 at a pressure of 0.20–0.25 MPa for 10 s. The remaining 40 specimens were left untreated. Two self-adhesive resin cements (G-CEM [GCM, GC; Tokyo, Japan] and RelyX Unicem 2 AutoMix [UNA, 3M Oral Care; St Paul, MN, USA]) and two conventional resin cements requiring pretreatment (RelyX Ultimate Adhesive Resin Cement [ULR, 3M Oral Care] and Panavia V5 [PAF, Kuraray Noritake; Tokyo, Japan]) were used to create test specimens. Then, a double-sided tape with a 4-mm-diameter hole was affixed to the adhesive surface of the dentin to define the adhesive area. presents the various cements and pretreatments applied, and illustrates the experimental procedure used in this study. The self-adhesive cement specimens (GCM and UNA) were placed into 3 groups by pretreatment method: (1) untreated PEKK specimens bonded to dentin, (2) sandblasted PEKK specimens bonded to dentin, and (3) PEKK specimens (adherends) bonded to sandblasted and chemically-pretreated dentin. In the latter group, chemical pretreatment of dentin surfaces consisted of applying Scotchbond Universal (3M Oral Care) for 20 s, followed by air drying for 5 s. This was subsequently irradiated with light (BlueShot, Shofu; Kyoto, Japan) for 10 s before bonding. The conventional resin cements, ULR and PAF, requiring pretreatment were processed by following the same steps (1–3) as those for the self-adhesive cements (light irradiation was performed on ULR during treatment with Scotchbond Universal). To allow PEKK specimens to adhere to root dentin, adhesive resin cement was applied to the PEKK specimens, which were then manually pressed against the root dentin. After excessive cement was removed, it was irradiated with light from four directions for 10 s. The prepared specimens were stored in a moist environment at 37°C for 1 week. Each group included ten specimens. Surface Observations In addition to the SBS test specimens, two PEKK specimens were prepared for each of the three groups: untreated PEKK specimens, sandblasted PEKK specimens, and PEKK specimens mechanically treated by sandblasting and chemically treated with a universal adhesive. The surfaces of the PEKK specimens were observed under a field-emission scanning electron microscope (FE-SEM´ SU6600, Hitachi; Tokyo, Japan). The specimens were Au-Pd sputter-coated before FE-SEM observation. The elemental composition of the specimen surfaces was analyzed using energy-dispersive x-ray spectroscopy (EDX) at a 10-mm working distance and 15.0-kV operating voltage. Shear Bond Strength Test After 1 week of storage, the test specimens were removed from the moist environment and mounted on the SBS testing jig of a universal testing machine (EZ Graph, Shimadzu; Kyoto, Japan). Shear force was applied at a crosshead speed of 0.5 mm/min from the root apex toward the crown. SBS was calculated from the obtained maximum fracture loads. After testing, the fracture surfaces were observed at 2.5X magnifcation under a stereomicroscope equipped with a digital camera (Stemi 508, Zeiss; Oberkochen, Germany). Compressive Strength Test To verify the mechanical strength of the cements used in this study, the compressive strength test was performed on individual specimens created by injecting the respective cement into 4.0-mm (diameter) x 8.0-mm (height) transparent acrylic tubes. These were light irradiated perpendicular to the long axis of the specimen for a period specified by the manufacturer. After curing, the acrylic tubes were removed from the resulting cylindrical specimens. The specimens were then stored in a moist environment at 37°C for 1 week. After 1 week of storage, compressive strength was tested in a universal testing machine (EZ Graph, Shimadzu) at a crosshead speed of 0.5 mm/min. The loads at fracture were recorded and the compressive strength was calculated from these. Statistical Analysis One-way ANOVA was performed to determine whether the mean SBS and the compressive strength of each specimen differed significantly, followed by Tukey’s multiple comparisons test (SPSS v 25, IBM; Armonk, NY, USA). Statistical significance was set at p < 0.05. shows a schematic diagram of test specimen preparation for the SBS test. In this study, 100 bovine anterior teeth were used. Frozen anterior teeth were thawed, and soft tissues still attached to the surfaces of the roots were removed. First, the bovine anterior tooth was split into a crown and a root portion at the cementoenamel junction, the tooth pulp was removed from the root canal, and the root was cut in half along its long axis . Second, the bovine root half was placed into a 2.54-cm epoxy resin ring, with the cut plane facing the top of the ring, and embedded in epoxy resin (Scandiplex, Fritsch; Hagen, Germany; ). The root was polished with a 120-grit waterproof abrasive paper until the surface of the root dentin was flattened, then further polished with abrasive papers up to 600 grit . The dentin surface was treated with 18% ethylenediaminetetraacetic acid solution (Ultradent EDTA18%, Ultradent; South Jordan, UT, USA) for 20 s and then washed in distilled water. The root was dried with air. All treated specimens were stored in a moist environment at 37°C and 95% humidity for 1 week; thereafter, the specimens were taken out and washed in distilled water. lists the materials used in this study and their composition. A 12-mm-diameter Pekkton ivory press ingot (Pekkton, Cendres + Metaux SA; Biel/Bienne, Switzerland), as the PEKK adherend was cut into 1-mm-thick slices using a blade with a thickness of 300 μm (Serge Microtome, SP1600, Leica; Wetzlar, Germany) to prepare 100 specimens. The prepared specimens were polished with waterproof abrasive papers up to 600 grit. Then, the specimens were ultrasonically cleaned and left to air dry. Of the 100 specimens, 60 were mechanically pretreated by sandblasting with 50-µm Al 2 O 3 at a pressure of 0.20–0.25 MPa for 10 s. The remaining 40 specimens were left untreated. Two self-adhesive resin cements (G-CEM [GCM, GC; Tokyo, Japan] and RelyX Unicem 2 AutoMix [UNA, 3M Oral Care; St Paul, MN, USA]) and two conventional resin cements requiring pretreatment (RelyX Ultimate Adhesive Resin Cement [ULR, 3M Oral Care] and Panavia V5 [PAF, Kuraray Noritake; Tokyo, Japan]) were used to create test specimens. Then, a double-sided tape with a 4-mm-diameter hole was affixed to the adhesive surface of the dentin to define the adhesive area. presents the various cements and pretreatments applied, and illustrates the experimental procedure used in this study. The self-adhesive cement specimens (GCM and UNA) were placed into 3 groups by pretreatment method: (1) untreated PEKK specimens bonded to dentin, (2) sandblasted PEKK specimens bonded to dentin, and (3) PEKK specimens (adherends) bonded to sandblasted and chemically-pretreated dentin. In the latter group, chemical pretreatment of dentin surfaces consisted of applying Scotchbond Universal (3M Oral Care) for 20 s, followed by air drying for 5 s. This was subsequently irradiated with light (BlueShot, Shofu; Kyoto, Japan) for 10 s before bonding. The conventional resin cements, ULR and PAF, requiring pretreatment were processed by following the same steps (1–3) as those for the self-adhesive cements (light irradiation was performed on ULR during treatment with Scotchbond Universal). To allow PEKK specimens to adhere to root dentin, adhesive resin cement was applied to the PEKK specimens, which were then manually pressed against the root dentin. After excessive cement was removed, it was irradiated with light from four directions for 10 s. The prepared specimens were stored in a moist environment at 37°C for 1 week. Each group included ten specimens. In addition to the SBS test specimens, two PEKK specimens were prepared for each of the three groups: untreated PEKK specimens, sandblasted PEKK specimens, and PEKK specimens mechanically treated by sandblasting and chemically treated with a universal adhesive. The surfaces of the PEKK specimens were observed under a field-emission scanning electron microscope (FE-SEM´ SU6600, Hitachi; Tokyo, Japan). The specimens were Au-Pd sputter-coated before FE-SEM observation. The elemental composition of the specimen surfaces was analyzed using energy-dispersive x-ray spectroscopy (EDX) at a 10-mm working distance and 15.0-kV operating voltage. After 1 week of storage, the test specimens were removed from the moist environment and mounted on the SBS testing jig of a universal testing machine (EZ Graph, Shimadzu; Kyoto, Japan). Shear force was applied at a crosshead speed of 0.5 mm/min from the root apex toward the crown. SBS was calculated from the obtained maximum fracture loads. After testing, the fracture surfaces were observed at 2.5X magnifcation under a stereomicroscope equipped with a digital camera (Stemi 508, Zeiss; Oberkochen, Germany). To verify the mechanical strength of the cements used in this study, the compressive strength test was performed on individual specimens created by injecting the respective cement into 4.0-mm (diameter) x 8.0-mm (height) transparent acrylic tubes. These were light irradiated perpendicular to the long axis of the specimen for a period specified by the manufacturer. After curing, the acrylic tubes were removed from the resulting cylindrical specimens. The specimens were then stored in a moist environment at 37°C for 1 week. After 1 week of storage, compressive strength was tested in a universal testing machine (EZ Graph, Shimadzu) at a crosshead speed of 0.5 mm/min. The loads at fracture were recorded and the compressive strength was calculated from these. One-way ANOVA was performed to determine whether the mean SBS and the compressive strength of each specimen differed significantly, followed by Tukey’s multiple comparisons test (SPSS v 25, IBM; Armonk, NY, USA). Statistical significance was set at p < 0.05. Observations of Treated Surfaces shows the surfaces of the untreated PEKK specimens (a1, a2), sandblasted specimens (b1, b2), and specimens sandblasted and primed using Scotch Bond Universal Adhesive (c1, c2). The surfaces of the untreated PEKK specimens (a1, a2) were relatively smooth and even, except for a few tool marks. The sandblasted PEKK specimens (b1, b2) had irregular, rough surfaces. The surfaces of the PEKK specimens that were sandblasted and primed using Scotch Bond Universal Adhesive (c1, c2) were relatively smooth, except for a slight unevenness because the sandblasted surfaces were covered with the priming agent. shows the EDX spectra and element compositions of the surfaces of the PEKK specimens obtained. C, O, and Ti were detected on the surfaces of the untreated PEKK specimens (a). Al was detected on the surfaces of the PEKK specimens sandblasted with Al 2 O 3 (b). C, O, Si, and small amounts of Al and P were found on the surfaces of the PEKK specimens sandblasted and primed using Scotch Bond Universal (c). Shear Bond Strength One-way ANOVA showed a significant difference in SBS between the groups (p < 0.05). shows the SBS at the PEKK/root-dentin interfaces bonded with different adhesive resin cements. No significant difference in SBS was observed between the cement products used in the untreated and sandblasted PEKK specimens. SBS was higher in PEKK specimens with conventional ULR than in some untreated PEKK specimens with self-adhesive cement (p < 0.05). In contrast, SBS was higher in PEKK specimens with self-adhesive cements, which were both mechanically and chemically pretreated, than that of PEKK specimens with other cements and pretreated by other techniques (p < 0.05). The modes of failure of the PEKK specimens after testing are shown in . Regardless of the cement type, many adhesive failures were observed at the PEKK-cement interfaces in all untreated specimens. SUNA (priming using Scotchbond Universal on sandblasted PEKK specimens and root dentin, then bonded to PEKK and root dentin with RelyX Unicem2 AutoMix), which had the highest bond strength, showed no adhesive failures between PEKK and cement, indicating either cement-dentin adhesive failure or mixed failure. Compressive Strength shows the compressive strength of the individual cement products. The mean compressive strengths ranged from 246 to 272 MPa, and no significant difference was observed between the cements (p > 0.05). shows the surfaces of the untreated PEKK specimens (a1, a2), sandblasted specimens (b1, b2), and specimens sandblasted and primed using Scotch Bond Universal Adhesive (c1, c2). The surfaces of the untreated PEKK specimens (a1, a2) were relatively smooth and even, except for a few tool marks. The sandblasted PEKK specimens (b1, b2) had irregular, rough surfaces. The surfaces of the PEKK specimens that were sandblasted and primed using Scotch Bond Universal Adhesive (c1, c2) were relatively smooth, except for a slight unevenness because the sandblasted surfaces were covered with the priming agent. shows the EDX spectra and element compositions of the surfaces of the PEKK specimens obtained. C, O, and Ti were detected on the surfaces of the untreated PEKK specimens (a). Al was detected on the surfaces of the PEKK specimens sandblasted with Al 2 O 3 (b). C, O, Si, and small amounts of Al and P were found on the surfaces of the PEKK specimens sandblasted and primed using Scotch Bond Universal (c). One-way ANOVA showed a significant difference in SBS between the groups (p < 0.05). shows the SBS at the PEKK/root-dentin interfaces bonded with different adhesive resin cements. No significant difference in SBS was observed between the cement products used in the untreated and sandblasted PEKK specimens. SBS was higher in PEKK specimens with conventional ULR than in some untreated PEKK specimens with self-adhesive cement (p < 0.05). In contrast, SBS was higher in PEKK specimens with self-adhesive cements, which were both mechanically and chemically pretreated, than that of PEKK specimens with other cements and pretreated by other techniques (p < 0.05). The modes of failure of the PEKK specimens after testing are shown in . Regardless of the cement type, many adhesive failures were observed at the PEKK-cement interfaces in all untreated specimens. SUNA (priming using Scotchbond Universal on sandblasted PEKK specimens and root dentin, then bonded to PEKK and root dentin with RelyX Unicem2 AutoMix), which had the highest bond strength, showed no adhesive failures between PEKK and cement, indicating either cement-dentin adhesive failure or mixed failure. shows the compressive strength of the individual cement products. The mean compressive strengths ranged from 246 to 272 MPa, and no significant difference was observed between the cements (p > 0.05). PEKK has the potential to be an attractive post-and-core material because of its adequate mechanical strength and shock-absorption properties, as well as its ability to be customized through various processing methods. A previous study that evaluated the biomechanical behavior and long-term performance of PEKK post-and-core, metal, and fiber-post/resin-core using the finite element method found that PEKK, which has the lowest elastic modulus, was the least likely to cause root fracture. On the other hand, PEKK post-and-core systems showed more debonding under long-term cyclic loading than did metal or fiberglass post-and-core systems. This indicates that strong adhesion to root canal dentin is important for the long-term performance of PEKK posts. This study showed that both treatment techniques – mechanical treatment by sandblasting and chemical treatment with Scotchbond Universal – might improve the bond strength at the PEKK-dentin interface. Thus, the null hypothesis was rejected. Pretreatment of PEKK Specimens Various types of mechanical and chemical treatment techniques have been experimentally applied to PEKK specimens. PEKK possesses chemical inertness, low surface energy, and resistance to surface modification, as do other types of PAEK materials, which allow it to achieve durable bonding at the interface between the resin and PEKK materials. , , , , , , One study reported that chemical treatment of PEKK with H 2 SO 4 improved bond strength. However, PEKK surface treatment with H 2 SO 4 at chairside may harm the patient. In contrast, mechanical pretreatment by sandblasting using Al 2 O 3 particles is common in clinical settings, and some studies have reported that this technique improved the bond strength of PEKK. , , , , , , , The SEM observations in this study revealed irregular, rough surfaces on PEKK specimens. Failure mode analysis showed that the sandblasted PEKK group had fewer adhesive failures between PEKK and cement than did the untreated group. This suggests that mechanical pretreatment by sandblasting resulted in a certain degree of mechanical retention. However, no significant difference in SBS was observed between any cement group pretreated by sandblasting vs the untreated cement experimental groups. Stawarczyk et al reported that when PEKK was mechanically pretreated by sandblasting at a pressure of 0.2 MPa and chemically pretreated with a dimethacrylate pretreatment agent (PEKKbond, anaxdent North America; Ardmore, PA, USA), the tensile bond strength (TBS) was insufficient. These results indicated that obtaining a suitable PEKK surface morphology to yield high bond strength to resin cement may depend on the pressure of sandblasting. One study reported that sandblasting at a pressure of 0.2 MPa did not affect the bond strength to PEKK, whereas other studies reported that sandblasting at a pressure 0.5 MPa improved the bond strength to PEKK. , Sandblasting in this study was performed at 0.2–0.25 MPa. These pressures are relatively common when sandblasting fiber posts. , However, it was suggested that this did not contribute to the improvement of the bond strength to PEKK. Stawarczyk et al used a methacrylate pretreatment agent, Visio.Link (Bredent; Senden, Germany), and observed higher TBS in PEKK specimens chemically pretreated with Visio.Link than in PEKK specimens pretreated with PEKKbond, even under the same sandblasting pressure of 0.2 MPa. They further suggested that the components of Visio.Link effectively dissolved the PEKK surfaces, improving the bond strength without being affected by the sandblasting pressure. In this study, Scotchbond Universal was applied to PEKK specimens as chemical pretreatment. One study found that Scotchbond Universal showed higher bond strength to PEKK without mechanical and/or chemical pretreatment with sulfuric acid compared to other pretreatment materials used in another study on PEKK. Similarly, in this study, ULR chemically pretreated with Scotchbond Universal demonstrated higher SBS than did the self-adhesive GCM or UNA resin cements. Additionally, self-adhesive cements mechanically pretreated with sandblasting and chemically pretreated with Scotchbond Universal showed higher SBS than other types of cements pretreated using any other techniques. Scotchbond Universal contains silane and 10-methacryloyloxydecyl dihydrogen phosphate (10-MDP) monomer. Si and P were detected by EDX on the PEKK surfaces; this demonstrated that the surfaces of the PEKK specimens were appropriately chemically pretreated. Pretreatment materials containing 10-MDP monomer were reported to have bond strengths similar to those of methacrylate pretreatment materials, which are known to provide high bond strengths to PAEK materials. To verify the finding, the present authors conducted an additional experiment. PEKK specimens were pretreated with a 10-MDP-containing pretreatment agent (Panavia V5 Tooth Primer, Kuraray Noritake) or a silane-containing pretreatment agent (Porcelain Primer, Shofu), and the SBS to UNA was measured, as it showed the highest bond strength in this study (n = 10). Higher bond strength was observed to PEKK specimens pretreated with 10-MDP and silane monomers than in untreated PEKK specimens . This suggests that not only 10-MDP monomer but also silane monomer contributes to the bond strength of the surfaces of PEKK specimens. The results of EDS analysis showed Ti (originating from PEKK) and Al (originating from Al 2 O 3 ) on the PEKK surface. The chemical bonding to Ti and Al of the phosphate ester groups of 10-MDP and the silanol groups produced by hydrolysis of silane coupling agents may have contributed to the improvement in bond strength. PEKK is chemically stable but hydrophobic. Therefore, PEEK may have a high affinity for hydrophobic structures such as methacryloyloxy groups in the pretreatment agents. However, since several monomers with these functional groups are contained in each pretreatment, further research is needed to determine which mechanism contributes to the improvement in bond strength. Bond Strength of PEKK to Root Dentin To the best of our knowledge, only Wang et al investigated the SBS at the PEKK-dentin interface. They focused on the bond strength of a dental prosthesis to crown dentin. In this study, root dentin was used as the adherend to investigate the applicability of adhesive resin cement to post material. Root dentin has fewer exposed dentinal tubules than does crown dentin, suggesting that the bond strength decreases. Several studies have reported the bond strengths to dentin of several adhesive resin cements. , , , , , Many of these studies found that the adhesive resin cements requiring pretreatment had a slightly higher bond strength than did self-adhesive resin cements. , , Similarly, in this study, pretreated ULR and PAF had slightly higher bond strengths. This finding is in agreement with the results of the study by Someya et al on the SBS between dentin and adhesive resin cement. They assumed that self-adhesive cements take time to fully cure, which affects their mechanical properties. However, no significant difference in compressive strength was observed between the cements stored under the same conditions and used in this study. In contrast, Someya et al also performed a pull-out test as well as a SBS test on the root dentin using a fiber post and reported lower retention of self-adhesive cements compared to those requiring pretreatment. According to those authors, the post had a low retentive force because resin tags were not fully formed in root dentin by the self-adhesive resin cement. The same findings were reported previously. In the present study, fracture surface observations demonstrated that when PEKK specimens were pretreated appropriately, adhesive failure was observed in relatively large numbers at the cement-dentin interface. As with previous studies, resin tags were probably not fully formed in this study. Various types of mechanical and chemical treatment techniques have been experimentally applied to PEKK specimens. PEKK possesses chemical inertness, low surface energy, and resistance to surface modification, as do other types of PAEK materials, which allow it to achieve durable bonding at the interface between the resin and PEKK materials. , , , , , , One study reported that chemical treatment of PEKK with H 2 SO 4 improved bond strength. However, PEKK surface treatment with H 2 SO 4 at chairside may harm the patient. In contrast, mechanical pretreatment by sandblasting using Al 2 O 3 particles is common in clinical settings, and some studies have reported that this technique improved the bond strength of PEKK. , , , , , , , The SEM observations in this study revealed irregular, rough surfaces on PEKK specimens. Failure mode analysis showed that the sandblasted PEKK group had fewer adhesive failures between PEKK and cement than did the untreated group. This suggests that mechanical pretreatment by sandblasting resulted in a certain degree of mechanical retention. However, no significant difference in SBS was observed between any cement group pretreated by sandblasting vs the untreated cement experimental groups. Stawarczyk et al reported that when PEKK was mechanically pretreated by sandblasting at a pressure of 0.2 MPa and chemically pretreated with a dimethacrylate pretreatment agent (PEKKbond, anaxdent North America; Ardmore, PA, USA), the tensile bond strength (TBS) was insufficient. These results indicated that obtaining a suitable PEKK surface morphology to yield high bond strength to resin cement may depend on the pressure of sandblasting. One study reported that sandblasting at a pressure of 0.2 MPa did not affect the bond strength to PEKK, whereas other studies reported that sandblasting at a pressure 0.5 MPa improved the bond strength to PEKK. , Sandblasting in this study was performed at 0.2–0.25 MPa. These pressures are relatively common when sandblasting fiber posts. , However, it was suggested that this did not contribute to the improvement of the bond strength to PEKK. Stawarczyk et al used a methacrylate pretreatment agent, Visio.Link (Bredent; Senden, Germany), and observed higher TBS in PEKK specimens chemically pretreated with Visio.Link than in PEKK specimens pretreated with PEKKbond, even under the same sandblasting pressure of 0.2 MPa. They further suggested that the components of Visio.Link effectively dissolved the PEKK surfaces, improving the bond strength without being affected by the sandblasting pressure. In this study, Scotchbond Universal was applied to PEKK specimens as chemical pretreatment. One study found that Scotchbond Universal showed higher bond strength to PEKK without mechanical and/or chemical pretreatment with sulfuric acid compared to other pretreatment materials used in another study on PEKK. Similarly, in this study, ULR chemically pretreated with Scotchbond Universal demonstrated higher SBS than did the self-adhesive GCM or UNA resin cements. Additionally, self-adhesive cements mechanically pretreated with sandblasting and chemically pretreated with Scotchbond Universal showed higher SBS than other types of cements pretreated using any other techniques. Scotchbond Universal contains silane and 10-methacryloyloxydecyl dihydrogen phosphate (10-MDP) monomer. Si and P were detected by EDX on the PEKK surfaces; this demonstrated that the surfaces of the PEKK specimens were appropriately chemically pretreated. Pretreatment materials containing 10-MDP monomer were reported to have bond strengths similar to those of methacrylate pretreatment materials, which are known to provide high bond strengths to PAEK materials. To verify the finding, the present authors conducted an additional experiment. PEKK specimens were pretreated with a 10-MDP-containing pretreatment agent (Panavia V5 Tooth Primer, Kuraray Noritake) or a silane-containing pretreatment agent (Porcelain Primer, Shofu), and the SBS to UNA was measured, as it showed the highest bond strength in this study (n = 10). Higher bond strength was observed to PEKK specimens pretreated with 10-MDP and silane monomers than in untreated PEKK specimens . This suggests that not only 10-MDP monomer but also silane monomer contributes to the bond strength of the surfaces of PEKK specimens. The results of EDS analysis showed Ti (originating from PEKK) and Al (originating from Al 2 O 3 ) on the PEKK surface. The chemical bonding to Ti and Al of the phosphate ester groups of 10-MDP and the silanol groups produced by hydrolysis of silane coupling agents may have contributed to the improvement in bond strength. PEKK is chemically stable but hydrophobic. Therefore, PEEK may have a high affinity for hydrophobic structures such as methacryloyloxy groups in the pretreatment agents. However, since several monomers with these functional groups are contained in each pretreatment, further research is needed to determine which mechanism contributes to the improvement in bond strength. To the best of our knowledge, only Wang et al investigated the SBS at the PEKK-dentin interface. They focused on the bond strength of a dental prosthesis to crown dentin. In this study, root dentin was used as the adherend to investigate the applicability of adhesive resin cement to post material. Root dentin has fewer exposed dentinal tubules than does crown dentin, suggesting that the bond strength decreases. Several studies have reported the bond strengths to dentin of several adhesive resin cements. , , , , , Many of these studies found that the adhesive resin cements requiring pretreatment had a slightly higher bond strength than did self-adhesive resin cements. , , Similarly, in this study, pretreated ULR and PAF had slightly higher bond strengths. This finding is in agreement with the results of the study by Someya et al on the SBS between dentin and adhesive resin cement. They assumed that self-adhesive cements take time to fully cure, which affects their mechanical properties. However, no significant difference in compressive strength was observed between the cements stored under the same conditions and used in this study. In contrast, Someya et al also performed a pull-out test as well as a SBS test on the root dentin using a fiber post and reported lower retention of self-adhesive cements compared to those requiring pretreatment. According to those authors, the post had a low retentive force because resin tags were not fully formed in root dentin by the self-adhesive resin cement. The same findings were reported previously. In the present study, fracture surface observations demonstrated that when PEKK specimens were pretreated appropriately, adhesive failure was observed in relatively large numbers at the cement-dentin interface. As with previous studies, resin tags were probably not fully formed in this study. This study assessed the effects of mechanical and chemical surface pretreatment and the effects of different resin cements on the SBS at the PEKK-root dentin interface. The SBS at the PEKK-root dentin interface did not improve with mechanical pretreatment by sandblasting with 50-μm Al 2 O 3 at a pressure of 0.20-0.25 MPa for 10 s. However, the SBS at the PEKK-root dentin interface significantly improved with a combination of sandblasting and chemical pretreatment with 10-MDP and silane monomers. Thus, PEKK prepared with these pretreatment techniques has potential for use as a post material. However, further studies are warranted to determine the optimal sandblasing pressure for PEKK and its affinity for chemical solvents. In the future, studies should be conducted involving thermocycling and fatigue tests to investigate the long-term potential for retention on dentin. |
Harmonizing Quality Improvement Metrics Across Global Trial Networks to Advance Paediatric Clinical Trials Delivery | 7f333a53-5324-43cb-8958-e1f99691420b | 11335960 | Pediatrics[mh] | Significant challenges remain in how paediatric clinical trials are conducted: upto 19% of trials have been reported to discontinue, with up to 38% of trials reporting patient recruitment as the main reason . These results are attributed to issues with the design and operational execution of these trials, including lengthy study start-up times, inability to meet target enrolment goals and poor patient retention rates . The clinical research enterprise needs to transform involving collaboration among diverse public and private stakeholders, innovative re-engineering of the current delivery of clinical trials, and novel methodologies to integrate existing expertise, resources, and infrastructure . These challenges apply to paediatric trials . Clinical trial networks can support optimizing trial delivery by implementing quality and performance metrics in alignment with sponsors and sites . Adopting rigorous, harmonized systems and procedures to capture operational metrics and compare them with performance targets can support tracking, evaluating, benchmarking and predicting performance. Metrics should measure the right factors accurately, with standard definitions and data points, to provide actionable information to support planning and decisions . Metrics can be used to track outcomes, processes, and performance in clinical trial delivery by sponsors and research networks and can focus on the individual site and protocol levels but also at a portfolio level, across trials and sites . The objectives of this paper are to: Describe approaches used by contributing networks to identify and develop key metrics/indicators Describe common metrics and challenges in identifying network metrics Identify a preliminary set of interoperable metrics An international quality initiative “think tank” was convened with representatives from three paediatric trial networks from different jurisdictions that focus on novel drugs. These networks are specialty-agnostic with wide geographical coverage and work with the Pharmaceutical industry and academic Sponsors. This group, derived from the ongoing dialogue and collaboration between these networks, focused on improving the pediatric research enterprise and infrastructure. The group met remotely from 2021 to 2023 with at least quarterly meetings. Open discussions were driven by sharing of approaches, processes, documents, and experiences from each network. Metrics and their methods of collections were identified through discussion and sharing within the think tank. The metrics were identified by a survey of the networks. Sources of alignment and divergence and opportunities for shared metrics were identified by consensus between members of the think tank. This work used process data from the networks excluding personal data. Accordingly, review by research ethics committees or Institutional Review Boards was not needed. Contributing Networks Institute for Advanced Clinical Trials (I-ACT) for Children I-ACT was created by a consortium of key stakeholders in paediatrics, including the Critical Path Institute, the American Academy of Pediatrics and others in academia, industry, and the regulatory world. I-ACT is a 501c3 non-profit organization with a mission to serve as a neutral and independent organization on behalf of children everywhere. I-ACT is designed to advance innovative medicines and device development and labelling to improve child health . I-ACT engages public and private stakeholders through research and education to ensure that healthcare for children is continually improved by enhancing the awareness, quality, and support for paediatric clinical trials. I-ACT currently includes 74 U.S. and international network sites committed to performing paediatric research to support regulatory approval by industry and academic Sponsors. I-ACT supports the network sites by providing clinical trial opportunities, a peer-to-peer mentoring program, educational webinars, professional education grants, and supports sites to improve paediatric research conduct. I-ACT launched the Pediatric Improvement Collaborative for Clinical Trials & Research (PICTR ® ), a quality improvement program to help identify and mitigate the challenges sites face when conducting paediatric clinical trials. The PICTR program collects and analyses paediatric clinical trial operations data at site level to determine best practices and process improvement. The data is shared across the site network. This exchange creates a continuous learning environment to maximize trial speed, quality and efficiency. conect4children (c4c)-Collaborative Network for European Clinical Trials for Children C4c is an action under the Innovative Medicines Initiative 2 (IMI2) Joint Understanding, Grant Agreement 777389 from 2018 to 2024 . The c4c consortium includes 10 large pharmaceutical companies and 37 non-industry partners, including academia, hospitals, third-sector organizations and patient advocacy groups. The consortium aims to set up and evaluate a pan-European paediatric-focused clinical trial infrastructure tailored to meet the needs of children involved in clinical trials. c4c is focused on four main areas of services, including: strategic feasibility expert advice on study design and/or paediatric development programmes, including patient/parent involvement; a network of over 250 clinical sites across 21 European countries coordinated by 20 National Hubs and a central Network Infrastructure Office, with local knowledge and expertise and aligned processes across the entire network; a Training Academy providing standardized training to all study sites and site personnel; and a Data focused work package to support management of data and metrics used by the network and the development of a standardised paediatric data dictionary. A new legal entity, the conect4children Stichting has been established to ensure sustainability of this project’s results. Maternal Infant Child and Youth Research Network (MICYRN) MICYRN is a Canadian federal not-for-profit, charitable organization founded in 2006 to build capacity for high-quality applied health research. MICYRN is governed by a Board comprising member research organizations and members at large, who represent specific research foci and expertise. Oversight of the network is maintained through an executive team consisting of the Board chair, vice-chair, scientific directors, and executive directors. The network formally links 21 maternal and child health research member organizations based at academic health centres in Canada; is affiliated with more than 25 practice-based research networks; provides support to new and emerging teams; and has established strong national and international partnerships such as I-ACT and c4c. The mission of MICYRN is to catalyze advances in maternal and child healthcare by connecting minds and removing barriers to high-quality health research. MICYRN is working towards building a national infrastructure to attract and facilitate the conduct of maternal-child investigator-initiated and industry-sponsored multicenter clinical trials and functions as a de-centralized Academic Research Organization. MICYRN prioritizes quality improvement initiatives, supports training and mentorship programs for emerging investigators and new trainees, and leverages national partnerships to lead advocacy initiatives for regulatory and ethical pathways in Canada. In collaboration with I-ACT, MICYRN is working on a Quality Improvement and Performance Metrics Initiative to collect information on key indicators to improve maternal/child health in Canada. Institute for Advanced Clinical Trials (I-ACT) for Children I-ACT was created by a consortium of key stakeholders in paediatrics, including the Critical Path Institute, the American Academy of Pediatrics and others in academia, industry, and the regulatory world. I-ACT is a 501c3 non-profit organization with a mission to serve as a neutral and independent organization on behalf of children everywhere. I-ACT is designed to advance innovative medicines and device development and labelling to improve child health . I-ACT engages public and private stakeholders through research and education to ensure that healthcare for children is continually improved by enhancing the awareness, quality, and support for paediatric clinical trials. I-ACT currently includes 74 U.S. and international network sites committed to performing paediatric research to support regulatory approval by industry and academic Sponsors. I-ACT supports the network sites by providing clinical trial opportunities, a peer-to-peer mentoring program, educational webinars, professional education grants, and supports sites to improve paediatric research conduct. I-ACT launched the Pediatric Improvement Collaborative for Clinical Trials & Research (PICTR ® ), a quality improvement program to help identify and mitigate the challenges sites face when conducting paediatric clinical trials. The PICTR program collects and analyses paediatric clinical trial operations data at site level to determine best practices and process improvement. The data is shared across the site network. This exchange creates a continuous learning environment to maximize trial speed, quality and efficiency. conect4children (c4c)-Collaborative Network for European Clinical Trials for Children C4c is an action under the Innovative Medicines Initiative 2 (IMI2) Joint Understanding, Grant Agreement 777389 from 2018 to 2024 . The c4c consortium includes 10 large pharmaceutical companies and 37 non-industry partners, including academia, hospitals, third-sector organizations and patient advocacy groups. The consortium aims to set up and evaluate a pan-European paediatric-focused clinical trial infrastructure tailored to meet the needs of children involved in clinical trials. c4c is focused on four main areas of services, including: strategic feasibility expert advice on study design and/or paediatric development programmes, including patient/parent involvement; a network of over 250 clinical sites across 21 European countries coordinated by 20 National Hubs and a central Network Infrastructure Office, with local knowledge and expertise and aligned processes across the entire network; a Training Academy providing standardized training to all study sites and site personnel; and a Data focused work package to support management of data and metrics used by the network and the development of a standardised paediatric data dictionary. A new legal entity, the conect4children Stichting has been established to ensure sustainability of this project’s results. Maternal Infant Child and Youth Research Network (MICYRN) MICYRN is a Canadian federal not-for-profit, charitable organization founded in 2006 to build capacity for high-quality applied health research. MICYRN is governed by a Board comprising member research organizations and members at large, who represent specific research foci and expertise. Oversight of the network is maintained through an executive team consisting of the Board chair, vice-chair, scientific directors, and executive directors. The network formally links 21 maternal and child health research member organizations based at academic health centres in Canada; is affiliated with more than 25 practice-based research networks; provides support to new and emerging teams; and has established strong national and international partnerships such as I-ACT and c4c. The mission of MICYRN is to catalyze advances in maternal and child healthcare by connecting minds and removing barriers to high-quality health research. MICYRN is working towards building a national infrastructure to attract and facilitate the conduct of maternal-child investigator-initiated and industry-sponsored multicenter clinical trials and functions as a de-centralized Academic Research Organization. MICYRN prioritizes quality improvement initiatives, supports training and mentorship programs for emerging investigators and new trainees, and leverages national partnerships to lead advocacy initiatives for regulatory and ethical pathways in Canada. In collaboration with I-ACT, MICYRN is working on a Quality Improvement and Performance Metrics Initiative to collect information on key indicators to improve maternal/child health in Canada. I-ACT was created by a consortium of key stakeholders in paediatrics, including the Critical Path Institute, the American Academy of Pediatrics and others in academia, industry, and the regulatory world. I-ACT is a 501c3 non-profit organization with a mission to serve as a neutral and independent organization on behalf of children everywhere. I-ACT is designed to advance innovative medicines and device development and labelling to improve child health . I-ACT engages public and private stakeholders through research and education to ensure that healthcare for children is continually improved by enhancing the awareness, quality, and support for paediatric clinical trials. I-ACT currently includes 74 U.S. and international network sites committed to performing paediatric research to support regulatory approval by industry and academic Sponsors. I-ACT supports the network sites by providing clinical trial opportunities, a peer-to-peer mentoring program, educational webinars, professional education grants, and supports sites to improve paediatric research conduct. I-ACT launched the Pediatric Improvement Collaborative for Clinical Trials & Research (PICTR ® ), a quality improvement program to help identify and mitigate the challenges sites face when conducting paediatric clinical trials. The PICTR program collects and analyses paediatric clinical trial operations data at site level to determine best practices and process improvement. The data is shared across the site network. This exchange creates a continuous learning environment to maximize trial speed, quality and efficiency. C4c is an action under the Innovative Medicines Initiative 2 (IMI2) Joint Understanding, Grant Agreement 777389 from 2018 to 2024 . The c4c consortium includes 10 large pharmaceutical companies and 37 non-industry partners, including academia, hospitals, third-sector organizations and patient advocacy groups. The consortium aims to set up and evaluate a pan-European paediatric-focused clinical trial infrastructure tailored to meet the needs of children involved in clinical trials. c4c is focused on four main areas of services, including: strategic feasibility expert advice on study design and/or paediatric development programmes, including patient/parent involvement; a network of over 250 clinical sites across 21 European countries coordinated by 20 National Hubs and a central Network Infrastructure Office, with local knowledge and expertise and aligned processes across the entire network; a Training Academy providing standardized training to all study sites and site personnel; and a Data focused work package to support management of data and metrics used by the network and the development of a standardised paediatric data dictionary. A new legal entity, the conect4children Stichting has been established to ensure sustainability of this project’s results. MICYRN is a Canadian federal not-for-profit, charitable organization founded in 2006 to build capacity for high-quality applied health research. MICYRN is governed by a Board comprising member research organizations and members at large, who represent specific research foci and expertise. Oversight of the network is maintained through an executive team consisting of the Board chair, vice-chair, scientific directors, and executive directors. The network formally links 21 maternal and child health research member organizations based at academic health centres in Canada; is affiliated with more than 25 practice-based research networks; provides support to new and emerging teams; and has established strong national and international partnerships such as I-ACT and c4c. The mission of MICYRN is to catalyze advances in maternal and child healthcare by connecting minds and removing barriers to high-quality health research. MICYRN is working towards building a national infrastructure to attract and facilitate the conduct of maternal-child investigator-initiated and industry-sponsored multicenter clinical trials and functions as a de-centralized Academic Research Organization. MICYRN prioritizes quality improvement initiatives, supports training and mentorship programs for emerging investigators and new trainees, and leverages national partnerships to lead advocacy initiatives for regulatory and ethical pathways in Canada. In collaboration with I-ACT, MICYRN is working on a Quality Improvement and Performance Metrics Initiative to collect information on key indicators to improve maternal/child health in Canada. Approaches Used by Contributing Networks to Identify and Develop Key Metrics/Indicators Pediatric Improvement Collaborative for Clinical Research and Trials (PICTR®) In 2018, PICTR worked closely with members of the site network to assess current paediatric clinical trial research operations. Sites completed surveys about their operations and met frequently to share gaps in their processes. Based on site feedback and subject matter experts (SME), a preliminary list of measurable goals and metrics was developed for improving the clinical trials process within sites. To ensure the program’s goals and metrics aligned across the industry, PICTR hosted an SME meeting in Chicago in 2019, bringing together key stakeholders to discuss the conduct of clinical trials including pharmaceutical companies, federal agencies, academia, research sites, other global paediatric networks, and patients and families. The meeting outcome was a draft set of six metrics used to identify gaps in the clinical research operations process at site level. Following the SME meeting, 14 sites participated in a pilot project collecting research operations metrics focused on the institutional review board and contracts process. The pilot aided in validating the program goals and identifying additional metrics after which, there was an ongoing collaboration with key stakeholders resulting in a final set of 11 core research operation metrics (Appendix A). Quality Improvement initiatives for sites were based on these metrics. connect4children-Collaborative Network for European Clinical Trial for Children C4c collects metrics to measure quality and performance of processes and network. Implementing a Performance Measurement System has a positive organisational effect, improves results over the long term, drives organisational strategy, supports planning and decision-making, and acts as an effective tool for communicating achieved results to stakeholders Within c4c, a methodological model was developed to identify a list of metrics and underlying data points to be suggested for adoption by c4c. The model considered metrics-specific issues, including: Terminology. Common practice and use of metrics—collected from examples of national networks and sponsors. Lean Management approach in clinical research (e.g. “time” as one of the key performance measures). Goal-Question-Metric Paradigm (defining goals behind the processes to be measured and using these to decide precisely what to measure). Multi-Criteria Decision Analysis (to aggregate several simple metrics into one meaningful combined metric). Target setting. A cross-work stream collaboration between c4c partners led to the selection of an initial core set of 13 metrics (Appendix B) from a list of 126 proposed metrics. The core set, prioritised by function and business case, is used to measure the performance of the studies used to define the network’s processes and to test its viability (so-called proof-of-viability studies), thereby testing the usefulness and actionability of this core set. Each metric has a target (value or range) and several attributes defined, including Name and Code, Process (mapped to Network or Clinical Trial processes), Definition, Data Points, as well as prioritisation for collection. The subset was chosen after a three-month consultation process across all c4c National Hubs and Industry partners of the consortium. The c4c Network Committee approved the metrics after a pilot phase of utilising with academic proof-of-viability studies. These metrics are critical to the c4c network and trial performance management framework and are continuously reviewed and evaluated. MICYRN—Maternal Infant Child and Youth Research Network In early 2019, MICYRN collaborated with I-ACT to learn about the PICTR initiative and metrics collected in the United States. Following the discussions with I-ACT, MICYRN engaged with its clinical trials consortium (CTC) comprised of scientific and operational representatives across 16 clinical trial units at MICYRN’s member research organizations to discuss the Quality Improvement (QI) and Performance Metrics initiative. Buy-in from the CTC was achieved and deemed important to the maternal-child health research community in Canada. The MICYRN leadership team conducted individual teleconferences with CTC sites to identify a list of meaningful indicators across the 3 domains of quality, efficiency, and timeliness; 11 interviews were completed. Using the interview data, an electronic survey was created with the compiled list of 14 indicators and disseminated to the 16 consortium sites for completion. Sites were asked to rank each indicator in order of importance to their site (1–14). 11/16 CTC sites completed the survey. The survey results were analysed, reducing the list to the top 6 indicators identified by the CTC sites. The 6 indicators were reviewed by the MICYRN leadership team and in terms of tangible action items that MICYRN could support and facilitate. The MICYRN Annual General Assembly brought together the CTC to collectively generate common data elements and definitions, inclusion/exclusion criteria, timeframe, methods of data collection, frequency of reporting, and unit analysis, further reducing the indicators to 5 (Appendix C). The CTC and MICYRN leadership team are currently working on metrics collection and action items for each of the 5 defined indicators. In summary, metric selection was driven by site quality improvement in iACT (11 metrics), by network performance in c4c (13 metrics), and by both in MICYRN (5 metrics). Commonalities and Challenges in Identifying Network Metrics Appendix A–D describe the metrics provided by the participating networks. The metrics developed are broadly at either trial level, or at site, and/or country level. They are related to individual services developed, and/or network/infrastructure Figure summarizes the commonalities of the approach to identifying and developing these metrics across the three networks. All networks used a staged evidence-based approach based on existing evidence and wide internal stakeholder consultation and co-creation, keeping in mind the expected implementation of metrics across sites and organizations. Appendix D summarises metrics related to each phase across each contributing network. The network driven by site quality improvement did not have indicators for capacity/capability or identification/feasibility (Table ). 15 metrics for trial start up and conduct were identified. Metrics related to approvals were found in all three networks. Topics relating to protocol review were only included by the network driven by site quality improvement. Topics relating to numbers of paediatric interventional clinical trials and investigators participating in these at country level were only included by the network focussing on country-wide approach. Site identification/feasibility indicators were only included by the network that was driven only by network management. The challenges faced when reviewing and identifying common metrics reported by the three networks were: Technical differences: c4c, I-ACT and MICYRN use (and source data from organisations that may use) different technical standards and systems, making it difficult to exchange data and information. Measurement and semantic differences: All three networks use different terminology, definitions for each data point and metrics, and coding systems, making it difficult to compare data across organizations. Each of the three networks used slightly different reference points and definitions to capture similar metrics. For example, specific definitions used for site “initiation”, “activation” and “ready for enrolment” timelines were different between networks, impacting how the dates for these steps were captured. The same was noted in recruitment dates related to patient screening, consent, or enrolment. The source of information also varies; c4c collects detailed information from sponsors, whereas I-ACT and MICYRN collect the information from sites. Organizational policies: Parent and partner organisations have different policies and regulations regarding data sharing and use; these need to be addressed to establish common guidelines for data exchange. These differences often arise because of the characteristics of health systems. Our findings and discussion suggest that metrics are identified, defined, and developed according to each stakeholder's goals and the processes they can measure or influence. Working Towards a Common Interoperable Set of Metrics By comparing the identified metrics across the networks, we found specific shared metrics measured across all three networks that can form the basis of comparators for the service/support that the networks provide across the trial lifecycle. Shared metrics could measure the effectiveness of interoperable networks. An example of a shared metric is shown in Table , illustrating challenges with terminologies and data point/measures alignment. Pediatric Improvement Collaborative for Clinical Research and Trials (PICTR®) In 2018, PICTR worked closely with members of the site network to assess current paediatric clinical trial research operations. Sites completed surveys about their operations and met frequently to share gaps in their processes. Based on site feedback and subject matter experts (SME), a preliminary list of measurable goals and metrics was developed for improving the clinical trials process within sites. To ensure the program’s goals and metrics aligned across the industry, PICTR hosted an SME meeting in Chicago in 2019, bringing together key stakeholders to discuss the conduct of clinical trials including pharmaceutical companies, federal agencies, academia, research sites, other global paediatric networks, and patients and families. The meeting outcome was a draft set of six metrics used to identify gaps in the clinical research operations process at site level. Following the SME meeting, 14 sites participated in a pilot project collecting research operations metrics focused on the institutional review board and contracts process. The pilot aided in validating the program goals and identifying additional metrics after which, there was an ongoing collaboration with key stakeholders resulting in a final set of 11 core research operation metrics (Appendix A). Quality Improvement initiatives for sites were based on these metrics. connect4children-Collaborative Network for European Clinical Trial for Children C4c collects metrics to measure quality and performance of processes and network. Implementing a Performance Measurement System has a positive organisational effect, improves results over the long term, drives organisational strategy, supports planning and decision-making, and acts as an effective tool for communicating achieved results to stakeholders Within c4c, a methodological model was developed to identify a list of metrics and underlying data points to be suggested for adoption by c4c. The model considered metrics-specific issues, including: Terminology. Common practice and use of metrics—collected from examples of national networks and sponsors. Lean Management approach in clinical research (e.g. “time” as one of the key performance measures). Goal-Question-Metric Paradigm (defining goals behind the processes to be measured and using these to decide precisely what to measure). Multi-Criteria Decision Analysis (to aggregate several simple metrics into one meaningful combined metric). Target setting. A cross-work stream collaboration between c4c partners led to the selection of an initial core set of 13 metrics (Appendix B) from a list of 126 proposed metrics. The core set, prioritised by function and business case, is used to measure the performance of the studies used to define the network’s processes and to test its viability (so-called proof-of-viability studies), thereby testing the usefulness and actionability of this core set. Each metric has a target (value or range) and several attributes defined, including Name and Code, Process (mapped to Network or Clinical Trial processes), Definition, Data Points, as well as prioritisation for collection. The subset was chosen after a three-month consultation process across all c4c National Hubs and Industry partners of the consortium. The c4c Network Committee approved the metrics after a pilot phase of utilising with academic proof-of-viability studies. These metrics are critical to the c4c network and trial performance management framework and are continuously reviewed and evaluated. MICYRN—Maternal Infant Child and Youth Research Network In early 2019, MICYRN collaborated with I-ACT to learn about the PICTR initiative and metrics collected in the United States. Following the discussions with I-ACT, MICYRN engaged with its clinical trials consortium (CTC) comprised of scientific and operational representatives across 16 clinical trial units at MICYRN’s member research organizations to discuss the Quality Improvement (QI) and Performance Metrics initiative. Buy-in from the CTC was achieved and deemed important to the maternal-child health research community in Canada. The MICYRN leadership team conducted individual teleconferences with CTC sites to identify a list of meaningful indicators across the 3 domains of quality, efficiency, and timeliness; 11 interviews were completed. Using the interview data, an electronic survey was created with the compiled list of 14 indicators and disseminated to the 16 consortium sites for completion. Sites were asked to rank each indicator in order of importance to their site (1–14). 11/16 CTC sites completed the survey. The survey results were analysed, reducing the list to the top 6 indicators identified by the CTC sites. The 6 indicators were reviewed by the MICYRN leadership team and in terms of tangible action items that MICYRN could support and facilitate. The MICYRN Annual General Assembly brought together the CTC to collectively generate common data elements and definitions, inclusion/exclusion criteria, timeframe, methods of data collection, frequency of reporting, and unit analysis, further reducing the indicators to 5 (Appendix C). The CTC and MICYRN leadership team are currently working on metrics collection and action items for each of the 5 defined indicators. In summary, metric selection was driven by site quality improvement in iACT (11 metrics), by network performance in c4c (13 metrics), and by both in MICYRN (5 metrics). In 2018, PICTR worked closely with members of the site network to assess current paediatric clinical trial research operations. Sites completed surveys about their operations and met frequently to share gaps in their processes. Based on site feedback and subject matter experts (SME), a preliminary list of measurable goals and metrics was developed for improving the clinical trials process within sites. To ensure the program’s goals and metrics aligned across the industry, PICTR hosted an SME meeting in Chicago in 2019, bringing together key stakeholders to discuss the conduct of clinical trials including pharmaceutical companies, federal agencies, academia, research sites, other global paediatric networks, and patients and families. The meeting outcome was a draft set of six metrics used to identify gaps in the clinical research operations process at site level. Following the SME meeting, 14 sites participated in a pilot project collecting research operations metrics focused on the institutional review board and contracts process. The pilot aided in validating the program goals and identifying additional metrics after which, there was an ongoing collaboration with key stakeholders resulting in a final set of 11 core research operation metrics (Appendix A). Quality Improvement initiatives for sites were based on these metrics. C4c collects metrics to measure quality and performance of processes and network. Implementing a Performance Measurement System has a positive organisational effect, improves results over the long term, drives organisational strategy, supports planning and decision-making, and acts as an effective tool for communicating achieved results to stakeholders Within c4c, a methodological model was developed to identify a list of metrics and underlying data points to be suggested for adoption by c4c. The model considered metrics-specific issues, including: Terminology. Common practice and use of metrics—collected from examples of national networks and sponsors. Lean Management approach in clinical research (e.g. “time” as one of the key performance measures). Goal-Question-Metric Paradigm (defining goals behind the processes to be measured and using these to decide precisely what to measure). Multi-Criteria Decision Analysis (to aggregate several simple metrics into one meaningful combined metric). Target setting. A cross-work stream collaboration between c4c partners led to the selection of an initial core set of 13 metrics (Appendix B) from a list of 126 proposed metrics. The core set, prioritised by function and business case, is used to measure the performance of the studies used to define the network’s processes and to test its viability (so-called proof-of-viability studies), thereby testing the usefulness and actionability of this core set. Each metric has a target (value or range) and several attributes defined, including Name and Code, Process (mapped to Network or Clinical Trial processes), Definition, Data Points, as well as prioritisation for collection. The subset was chosen after a three-month consultation process across all c4c National Hubs and Industry partners of the consortium. The c4c Network Committee approved the metrics after a pilot phase of utilising with academic proof-of-viability studies. These metrics are critical to the c4c network and trial performance management framework and are continuously reviewed and evaluated. In early 2019, MICYRN collaborated with I-ACT to learn about the PICTR initiative and metrics collected in the United States. Following the discussions with I-ACT, MICYRN engaged with its clinical trials consortium (CTC) comprised of scientific and operational representatives across 16 clinical trial units at MICYRN’s member research organizations to discuss the Quality Improvement (QI) and Performance Metrics initiative. Buy-in from the CTC was achieved and deemed important to the maternal-child health research community in Canada. The MICYRN leadership team conducted individual teleconferences with CTC sites to identify a list of meaningful indicators across the 3 domains of quality, efficiency, and timeliness; 11 interviews were completed. Using the interview data, an electronic survey was created with the compiled list of 14 indicators and disseminated to the 16 consortium sites for completion. Sites were asked to rank each indicator in order of importance to their site (1–14). 11/16 CTC sites completed the survey. The survey results were analysed, reducing the list to the top 6 indicators identified by the CTC sites. The 6 indicators were reviewed by the MICYRN leadership team and in terms of tangible action items that MICYRN could support and facilitate. The MICYRN Annual General Assembly brought together the CTC to collectively generate common data elements and definitions, inclusion/exclusion criteria, timeframe, methods of data collection, frequency of reporting, and unit analysis, further reducing the indicators to 5 (Appendix C). The CTC and MICYRN leadership team are currently working on metrics collection and action items for each of the 5 defined indicators. In summary, metric selection was driven by site quality improvement in iACT (11 metrics), by network performance in c4c (13 metrics), and by both in MICYRN (5 metrics). Appendix A–D describe the metrics provided by the participating networks. The metrics developed are broadly at either trial level, or at site, and/or country level. They are related to individual services developed, and/or network/infrastructure Figure summarizes the commonalities of the approach to identifying and developing these metrics across the three networks. All networks used a staged evidence-based approach based on existing evidence and wide internal stakeholder consultation and co-creation, keeping in mind the expected implementation of metrics across sites and organizations. Appendix D summarises metrics related to each phase across each contributing network. The network driven by site quality improvement did not have indicators for capacity/capability or identification/feasibility (Table ). 15 metrics for trial start up and conduct were identified. Metrics related to approvals were found in all three networks. Topics relating to protocol review were only included by the network driven by site quality improvement. Topics relating to numbers of paediatric interventional clinical trials and investigators participating in these at country level were only included by the network focussing on country-wide approach. Site identification/feasibility indicators were only included by the network that was driven only by network management. The challenges faced when reviewing and identifying common metrics reported by the three networks were: Technical differences: c4c, I-ACT and MICYRN use (and source data from organisations that may use) different technical standards and systems, making it difficult to exchange data and information. Measurement and semantic differences: All three networks use different terminology, definitions for each data point and metrics, and coding systems, making it difficult to compare data across organizations. Each of the three networks used slightly different reference points and definitions to capture similar metrics. For example, specific definitions used for site “initiation”, “activation” and “ready for enrolment” timelines were different between networks, impacting how the dates for these steps were captured. The same was noted in recruitment dates related to patient screening, consent, or enrolment. The source of information also varies; c4c collects detailed information from sponsors, whereas I-ACT and MICYRN collect the information from sites. Organizational policies: Parent and partner organisations have different policies and regulations regarding data sharing and use; these need to be addressed to establish common guidelines for data exchange. These differences often arise because of the characteristics of health systems. Our findings and discussion suggest that metrics are identified, defined, and developed according to each stakeholder's goals and the processes they can measure or influence. By comparing the identified metrics across the networks, we found specific shared metrics measured across all three networks that can form the basis of comparators for the service/support that the networks provide across the trial lifecycle. Shared metrics could measure the effectiveness of interoperable networks. An example of a shared metric is shown in Table , illustrating challenges with terminologies and data point/measures alignment. The adoption of rigorous, harmonized operational metrics along with performance targets can support tracking, evaluating, benchmarking, and predicting performance . To our knowledge, this is the first report of international comparisons between international paediatric clinical research networks. c4c, I-ACT and MICYRN each have developed and implemented a well-defined set of metrics. Despite differences, common ground exists in the approaches, methods and sources of data collection for these metrics. The review and usage of these metrics are defined by each network’s internal goals. The main aim aligned across all three networks is to ensure the efficient conduct of clinical trials across the network sites. Adopting common metrics, standards and terminologies across organizations helps ensure data interoperability, identifying common trends, and allows the networks to work towards benchmarking. The networks have worked together to identify a core set of metrics which is comparable to other multi-site, multi-national clinical research organizations working towards efficiency in trial set-up, enrolment, and completion . Developing common metrics for multiple networks from the start of the networks would be ideal. Each network needs to establish its focus and identity before liaising with other networks. Several challenges exist concerning I-ACT, MICYRN and c4c using metrics inter-operably. Some of these challenges can be addressed by clearly identifying specific collaborative activities that address organisational, measurement and technical issues in a comprehensive and coherent manner. One challenge relates to the specific nature of what is being measured and how, which ultimately impacts the measurement properties, utility, and impact of selected metrics, pertinent to how the metrics and underlying data points were defined. The absence of a widely accepted standard for data nomenclature, exchange and interoperability means that theses aspects will need to be addressed within each network and then across the networks. For future inter-operability, all 3 networks will need to agree upon common sets of metrics and accompanying definitions. Another general limitation of metrics-driven network initiatives is the oversight and influence that each network has over their respective sites. Each network has been designed and established with its partner organisations with differences in communication channels, sponsor interactions activities, and structures, all of which impact the information collection and flow through the organisations. The networks cannot mandate sharing of data because different partner organizations have different cultures, governance, and ways of working, which can impact their willingness and ability to collaborate on interoperability efforts. In particular, there is a limitation in capturing and interpreting variations in some metrics that are beyond the control of the network. An example of this is seen in recruitment timelines. The recruitment metrics were mostly aggregated and do not consider potential reasons for efficiency or delays. This limitation makes it difficult to identify specific actions for standardization and improvement in the future. Furthermore, c4c and I-ACT both have specific roles and objectives for their organizational sites or country-level networks and they receive a mixture of private and public funds to support the initiative. On the other hand, the sites affiliated with MICYRN are academic member organizations that operate without dedicated funds to support their metrics collection initiative. As a result, MICYRN does not have the same level of influence and incentive for their sites, and the collection of certain metrics largely depends on the individual sites rather than the network. This poses additional challenges in gathering accurate and comprehensive metrics. Despite these challenges, the overarching goal of these networks is to improve the conduct of clinical research studies sponsored by industry and academia. Interoperable metrics will ensure that clinical research study operations across different networks can be reported on in a standardised manner, which makes it easier to compare data across different networks and countries. A more comprehensive, consistent and accurate view across sites and countries is possible, allowing better identification of issues and informed decisions to be made based on high-quality data. The above advantages will support tracking and performance management of network activities supporting clinical trials, collaborative decision making, and solution finding. Sharing of good practice and learnings across networks will result in efficiencies of trial set up and enrolment stages, thereby reducing costs and ultimately helping improve patient outcomes. Shared metrics and targets establish a common framework that will allow better identification of bottlenecks and hurdles in the trial delivery process, and development of quality improvement initiatives to support site and organizations, including adequate resourcing and process improvement. Interoperable metrics enable clinical research networks to collaborate and share data, which can lead to increased efficiency and the development of new treatments and therapies. Other existing clinical research networks around the world that include paediatric research activities collect clinical trial research performance metrics or benchmark data to assess the performance of their sites. However, the methodologies for collecting such data vary. Without a universal standard or methodology in place, networks cannot reference the same metrics across all domains of the trial lifecycle. For example, based on publically available data, the Paediatric Trials Network (PTN) addresses a reduction in time from the start of a study to completion and increased enrolment as part of its organizational improvements to increase efficiency and the Cystic Fibrosis Foundation (CF) Benchmarks use metrics focused on time to contract execution and time to first patient inclusion, whilst metrics from other networks such as the CTSA- Clinical Research Consortium and the Pediatric Emergency Care Applied Research Network (PECARN) do not address these areas of activity. Conversely, even when many networks focus on the same domains, methodological differences in the approach can be identified. That would be the case when contrasting metrics focused on time to IRB approval and recruitment from the three aforementioned research networks. These disparities are not dissimilar to the ones we identified, and can be at least partially justified by different contexts and purposes from each network, as well as constraints to data sources and data collection. Efforts are being made to establish more standardized approaches to data collection and measurement in clinical research . Regulatory bodies, professional organizations, and research institutions are working towards developing common frameworks and guidelines to facilitate the collection and reporting of data across different sites and networks. These initiatives aim to promote consistency and comparability in performance metrics, which can lead to better assessment and evaluation of clinical trial outcomes. It is important for clinical research networks and stakeholders to collaborate, share their experiences and knowledge to establish common methodologies and definitions to help advance the field and ensure the rigorous conduct of clinical trials, ultimately benefiting patients and advancing medical knowledge. It should be noted that this paper only focuses on a small sample of industry- and academia-facing large paediatric trial networks that are specialty agnostic with wide geographical coverage. Other networks focused on specific disciplines or covering other study designs may require a tailored approach to the selection of relevant operations metrics. In addition to these specific metric related limitations, interoperability efforts require significant funding and resources, as well as time and commitment to work together, which may not be available to all organisations. Recommendations and Next Steps The think-tank identified specific collaborative activities that are needed to develop and use interoperable metrics: Harmonization of processes for the collection of data related to metrics, including goals, data definitions, and measurement methods, across organizations can help ensure data comparability. Collaborative development of technology solutions that support interoperability, such as common data platforms and APIs. Provision of education and training to staff on the importance of data requirements, capture and integrity. Shared educational and training opportunities focusing on quality improvement may reduce burden on resources. Work with similar networks, e.g. those that may be academia-facing or not mandated to study new drugs with industry , on interoperable metrics and their implementation Addressing context at site and national level and regularly testing and evaluating the interoperability of data and systems across organizations to help identify and resolve any issues. Engaging stakeholders, including patients, healthcare providers, and regulatory agencies, in the interoperability efforts so that the solutions developed meet their needs and are widely adopted. Developing a multi-stakeholder engagement strategy for sites to be involved in metrics projects and across the three networks would further ensure interoperability. Establishing data sharing agreements to ensure the secure exchange of data and information. The think tank proposes the following next steps to utilise these metrics inter-operably: Define common or similar metrics/terminology that can be used within each network (intra network metrics) but be alike enough to allow interpretation as globally aligned networks. Define and align on data points that are collected to measure each metric. Review target values and/or comparators and/or benchmarks that may be used to drive performance. The think-tank identified specific collaborative activities that are needed to develop and use interoperable metrics: Harmonization of processes for the collection of data related to metrics, including goals, data definitions, and measurement methods, across organizations can help ensure data comparability. Collaborative development of technology solutions that support interoperability, such as common data platforms and APIs. Provision of education and training to staff on the importance of data requirements, capture and integrity. Shared educational and training opportunities focusing on quality improvement may reduce burden on resources. Work with similar networks, e.g. those that may be academia-facing or not mandated to study new drugs with industry , on interoperable metrics and their implementation Addressing context at site and national level and regularly testing and evaluating the interoperability of data and systems across organizations to help identify and resolve any issues. Engaging stakeholders, including patients, healthcare providers, and regulatory agencies, in the interoperability efforts so that the solutions developed meet their needs and are widely adopted. Developing a multi-stakeholder engagement strategy for sites to be involved in metrics projects and across the three networks would further ensure interoperability. Establishing data sharing agreements to ensure the secure exchange of data and information. The think tank proposes the following next steps to utilise these metrics inter-operably: Define common or similar metrics/terminology that can be used within each network (intra network metrics) but be alike enough to allow interpretation as globally aligned networks. Define and align on data points that are collected to measure each metric. Review target values and/or comparators and/or benchmarks that may be used to drive performance. This paper presents a review on the experience of three international paediatric clinical research networks to establish metrics for paediatric clinical trial support, demonstrating a disparity in methodology and common challenges in defining metrics. The adoption of rigorous, validated, and harmonized operational metrics, along with performance targets, can bring several advantages to international paediatric research networks. The recommended next steps will contribute to enabling international collaboration and benchmarking, thereby resulting in more efficient trial set-up, enrolment, and completion, reducing costs and improving patient outcomes. |
A comparison of patient appraisal of professional skills for GPs in training participating in differing education programs | 18991736-3dfb-453f-ad98-ee71140a9cd9 | 9462893 | Family Medicine[mh] | Healthcare systems internationally are under increasing pressure to recruit and retain physicians given worldwide demand that exceeds supply. The World Health Organization noted in 2006 that shortage of physicians was likely to be widespread in many countries by 2015 . Current predictions are that the USA, for example, will have a shortage of between 21,000 to 55,000 primary care physicians by 2033 . Also, there are predictions that there will be a shortage of over 9000 GPs by 2030 in Australia, representing almost a quarter of the workforce . Clearly, there is a need for suitably qualified GPs, adequately skilled to provide professional and empathic care in the context of the healthcare system in which they work. Internationally, there is increasing awareness that policies and strategies for increasing the number of international medical graduates (IMGs) and overseas trained doctors (OTDs, a term used to describe doctors who obtained their primary medical qualification in a country apart from Australia and New Zealand) in overburdened healthcare systems need to focus on enhancing pathways to allow such doctors to be registered and credentialed, so that they can practice effectively in their newly adopted country. Three policy issues under current debate include historical bias in the registration process, making it more difficult for IMGs and OTDs to qualify than locally trained doctors ; increased risk of complaints against IMGs and OTDs should pathways be eased; and racism and bias against IMGs and OTDs at both systemic and individual levels . There is also a perception that increased risk of complaints may be due not to lack of clinical skills but of professional or ‘soft skills’, such as interpersonal communication skills and empathy, where different cultural backgrounds can lead to different use of language and interactions with patients . Given the reliance on IMGs and OTDs for dealing with growing shortfalls in primary care, there needs to be more understanding of how the professional performance of IMGs and OTDs compares with their locally trained counterparts. Such understanding may identify improvements in aspects of IMG and OTD training as well as help these doctors to better understand the needs and expectations of their intended national healthcare system and its patients. Previous comparative studies on professional performance of IMGs and OTDs have tended to use outcomes such as patient survival and complaint rates , or simulated case studies . These outcomes, while important, do not focus on the skills that contribute to professional performance. In Australia, 29% of the current medical workforce consists of doctors who have trained overseas , and many of these doctors also identify difficulties that relate to their performance. For example, IMGs and OTDs have reported struggling when attempting the Australian Medical Council Examinations . There is also literature exploring the reasons doctors who have trained overseas may have difficulty working in Australia and it has been identified that the process of migration and adjustment has affected their performance . Further, a difference in personality traits of internationally trained doctors compared with Australian graduates has been demonstrated and may provide some insight into their professional attributes, and performance . Other factors identified have included difficulty with English, differences in medical education, length of time since medical school graduation, family and financial obligations, cultural approaches and beliefs, and the status and role of the physician . While these studies help build holistic understanding into the professional performance of IMGs and OTDs, and highlight some important factors to consider, they do not directly investigate the skills of these doctors transitioning into the Australian medical system. There has been very little attempt to identify the effect of different General Practice education and training programs on the ability of IMGs and OTDs, as well as locally trained doctors, to acquire the desirable professional skills deemed necessary for working effectively in the general practice sector. In particular, there appears to be no detailed study of how doctors gaining General Practice specialist registration through the different programs are perceived by patients in terms of their professional skills. Finally, there appears to have been no comparison between patient perception of General Practitioners in Training (GPiT) on the one hand, and patient perception of experienced practitioners on the other, to identify possible areas for enhancement of professional skills for both types of fellowship programs. This study seeks to understand the professional performance of doctors, as perceived by patients, with a particular focus on doctors who have gained their primary qualification overseas. Previous cultural factors can be expected to influence how this group communicates with patients . Many doctors undertaking the Royal Australian College of General Practice (RACGP) Practice Experience Program (PEP) obtained their primary qualifications overseas and are working in areas of workforce shortage under limited or no formal supervision . PEP is a self-directed education program delivered in partnership with training organisations to support doctors gain RACGP Fellowship, thereby allowing them to continue to practice in Australia in the primary care sector as GP specialists. The Australian General Practice Training (AGPT) program, on the other hand, prepares mainly Australian and New Zealand medical graduates for RACGP Fellowship and specialist registration by providing a three- or four-year educator-directed training program, including intensive supervision. Eligibility for the AGPT is more restricted than for the PEP with a subsequent competitive selection process. AGPT program training takes place in hospitals and general practices. The AGPT program currently is the most common pathway for Australian registrars to achieve General Practice Fellowship. Doctors on both programs are GPs in Training (GPiTs). Further details of the two programs and their differences using the TIDieR checklist as a guide can be found in Table , with the latest information on the demographics of doctors involved available via the RACGP website . The overall aim of our study is to compare patients’ experiences of the professionalism of GPiTs on the two different training programs given their distinctly different cohort demographics. Understanding any differences can lead to improvements in training programs, peer-dialogue, and reflection for the benefit of patients. To provide a benchmark against which both program groups can be measured, a third and large dataset of patient ratings for current (Fellowed) GPs undertaking patient feedback as part of their continuing professional development (CPD) for ongoing Medical Board of Australia registration was used. Since this third group will already have had several years’ experience in the primary care sector and the majority have achieved GP specialist registration, their professional performance as rated by patients can provide standards and a benchmark to which the two groups of trainee practitioners may wish to aspire. Patient data from these three groups are labelled Dataset A, Dataset B and Dataset C below.
Overview Patient feedback was obtained for doctors going through the Royal Australian College of General Practice (RACGP) Practice Experience Program (PEP) and the Australian General Practice Training Program (AGPT), with the former intended primarily for IMGs and OTDs, and the latter for local medical graduates including from New Zealand. Further details of the two programs and their differences using the TIDieR checklist as a guide can be found in Table : Comparison of education programs. This study deals with the patient feedback aspect of the two progams only. Patient feedback was also obtained for patients visiting experienced GPs for comparative purposes, resulting in data for three groups of doctors (two trainee, one experienced). Data The data consist of 57,745 anonymized patient responses to three groups of doctor (average 36–39 patients per doctor type) working in Australia, resulting in three datasets: Dataset A. 221 doctors who have trained primarily overseas and who are enrolled in the RACGP PEP; Dataset B. 355 General Practice registrars enrolled in the AGPT Program; and Dataset C. 923 Australian GPs who receive patient feedback as part of their CPD program (GP CPD). The patient questionnaire used in this study deals with the patient’s visit to their doctor and asks patients to rate their just completed consultation experience. PEP and AGPT patient questionnaires use 13 questions, and GP patient questionnaires use 12 of the 13 questions (more details below). All questions ask for responses using a five-point Likert scale with labels ‘poor’, ‘fair’, ‘good’, ‘very good’, and ‘excellent’. Additional file provides the full text of each question, with a shortened version as used in this report. Data collection A Human Research Ethics Committee approved this study (clearance number 2020000515 from the University of Queensland). The participants gave informed consent for their non-identifiable data to be used in research as part of the consent process to undertake feedback. The data were collected in the period between 1st January 2018 and 30th April 2020. A pack of 50 questionnaires per participating doctor was sent to practices, with written instructions provided to practice reception staff to hand out the questionnaire to consecutive patients so that convenience sampling based on willingness to participate was implemented. Patients were asked to rate their encounter according to their experience of that specific visit. To ensure patient confidentiality and anonymity, and to encourage honest feedback, completed questionnaires were placed in self-sealed envelopes and into ballot-style boxes by patients themselves before departure from the practice. No post-departure completion through email or internet took place. Further details concerning the content and format of patient questionnaire can be obtained by emailing the authors. Questionnaires were processed by Client Focused Evaluation Program (CFEP) Surveys in Brisbane, Australia. Paper questionnaires were scanned and verified electronically by an experienced data auditor. Data were imported to an in-house software system running on an enterprise database where they were further checked and verified. The patient datasets were exported as Microsoft Excel Spreadsheets to an SPSS database (SPSS for Windows Version 25) and cleaned and checked prior to data analysis. Statistical analysis On the basis that the intervals between the five Likert scale points are equal, item responses were converted into percentages (‘poor’ = 20%, ‘fair’ = 40%, ‘good’ = 60%, ‘very good’ = 80%, ‘excellent’ = 100%) to allow for parametric techniques based on means, standard deviations and variances. Conversion to percentages can aid intelligibility, allow benchmark comparison across different studies and groups as well as highlight differences without the need to represent results to four or five decimal places. Presenting percentages also provides consistency with previous doctor feedback results in the Australian professional performance framework that have been presented in percentages . Two levels of analysis were conducted: at the raw score rater- and item- level (irrespective of doctor rated), and at the aggregated doctor level where doctors received the mean item scores of all their raters. The sampling strategy detailed above has special characteristics that need to be accounted for, that is, the data are unbalanced because of variable numbers of raters per ratee, fully nested because all the ratees may be unique to that rater, and uncrossed because raters provide only one rating per ratee on one occasion. Cronbach’s alpha is reported below, but the alpha results should be interpreted with caution since some of the assumptions of its use (e.g., all raters are rating the same subject, object, or event) are not met in this study. Its use here is to check on the internal consistency of the questionnaire (questionnaire reliability). A signal-to-noise ratio (SNR) measure for dealing with unbalanced, uncrossed and fully nested data is also used to provide an estimate of data reliability . The content and construct validity of the original patient questionnaire were first established in 1999 as the Doctor’s Interpersonal Skill Questionnaire (DISQ) and its validity re-evaluated in 2010 when being assessed for use in the relicensure of doctors by the UK GMC . Its validity and reliability were reaffirmed in 2013 after minor edits were made to the wording of some items and the revised questionnaire applied in unbalanced, uncrossed and fully nested studies involving over 85,000 patients to over 2000 doctors . Analysis of variance (ANOVA) is a collection of statistical models used to analyze the differences among group means. The observed variance in a particular variable is partitioned into components attributable to different sources of variation. In its simplest form, ANOVA provides a statistical test of whether or not the means of several groups are equal. ANOVA is used to test for differences in item ratings and averages within and between PEP, AGPT and GP CPD data. Principal component analysis (PCA) is a data reduction technique for explaining variance in data using a smaller set of variables than the original variables or items. The Kaiser-Meyer-Olkin (KMO) test is a measure to determine sampling adequacy for each item. KMO values between 0.8 and 1.0 indicate that there are enough samples and sufficiently low variance for efficient identification of underlying components. Bartlett’s sphericity tests whether there are relations among variables suitable for structure detection, and PCA is used here to confirm the presence of components previously found when demonstrating criterion and construct validity . Single measures intraclass coefficients (ICCs) provide a relative measure of the variability in the sample of responses and is useful for estimating the agreement between raters on how to interpret the items. Values between 0.4 to 0.6 are considered ‘moderate agreement’, between 0.6 and 0.8 ‘good agreement’ and above 0.8 ‘very good agreement’ . One-way random ICCs are used in this study to check for reliability of the questionnaires given that all the raters are different. A t-test compares the means of two variables to determine whether there is a difference. Such a test can be used to estimate whether the responses given by two populations to a single set of items differ significantly. T-tests assume a normal distribution and the raw score data in this study are negatively skewed. However, its use is justified here because of the large sample sizes and the need to check whether item means differ between groups of doctors after aggregation by doctor (distribution of distributions). Linear regression is used to estimate and control for possible bias in ratings due to sociodemographic factors. These factors are entered first into the regression model against the dependent variable (average patient score) followed by entry of the independent variables (questionnaire items), with comparisons made concerning the amount of variance explained at each step. Psychometric network analysis provides graphical representations of relationships and interactions between variables such as questionnaire items . Nodes in the graph represent the items and links represent the strength of association between them. Inter-item mean score correlations, scaled between 0 and 1 are used, with width of links proportional to the strength of the association. The layout adopted is the force-directed ‘spring’, where variables with strongest associations and therefore of hypothesized strongest influence are placed closer together and at the centre of the graph . Summing the correlations for each item results in a node ‘strength’ measure that can be useful for assessing the influence of items and identifying possible points for future intervention, based on the assumption that changes in central items should have greatest impact on other items. Centrality scores are presented as standard scores (standard deviations above or below mean 0) to allow for comparison across the different doctor groups. All statistical analysis was carried out with SPSS v25 and network analysis through qgraph in R.
Patient feedback was obtained for doctors going through the Royal Australian College of General Practice (RACGP) Practice Experience Program (PEP) and the Australian General Practice Training Program (AGPT), with the former intended primarily for IMGs and OTDs, and the latter for local medical graduates including from New Zealand. Further details of the two programs and their differences using the TIDieR checklist as a guide can be found in Table : Comparison of education programs. This study deals with the patient feedback aspect of the two progams only. Patient feedback was also obtained for patients visiting experienced GPs for comparative purposes, resulting in data for three groups of doctors (two trainee, one experienced).
The data consist of 57,745 anonymized patient responses to three groups of doctor (average 36–39 patients per doctor type) working in Australia, resulting in three datasets: Dataset A. 221 doctors who have trained primarily overseas and who are enrolled in the RACGP PEP; Dataset B. 355 General Practice registrars enrolled in the AGPT Program; and Dataset C. 923 Australian GPs who receive patient feedback as part of their CPD program (GP CPD). The patient questionnaire used in this study deals with the patient’s visit to their doctor and asks patients to rate their just completed consultation experience. PEP and AGPT patient questionnaires use 13 questions, and GP patient questionnaires use 12 of the 13 questions (more details below). All questions ask for responses using a five-point Likert scale with labels ‘poor’, ‘fair’, ‘good’, ‘very good’, and ‘excellent’. Additional file provides the full text of each question, with a shortened version as used in this report.
A Human Research Ethics Committee approved this study (clearance number 2020000515 from the University of Queensland). The participants gave informed consent for their non-identifiable data to be used in research as part of the consent process to undertake feedback. The data were collected in the period between 1st January 2018 and 30th April 2020. A pack of 50 questionnaires per participating doctor was sent to practices, with written instructions provided to practice reception staff to hand out the questionnaire to consecutive patients so that convenience sampling based on willingness to participate was implemented. Patients were asked to rate their encounter according to their experience of that specific visit. To ensure patient confidentiality and anonymity, and to encourage honest feedback, completed questionnaires were placed in self-sealed envelopes and into ballot-style boxes by patients themselves before departure from the practice. No post-departure completion through email or internet took place. Further details concerning the content and format of patient questionnaire can be obtained by emailing the authors. Questionnaires were processed by Client Focused Evaluation Program (CFEP) Surveys in Brisbane, Australia. Paper questionnaires were scanned and verified electronically by an experienced data auditor. Data were imported to an in-house software system running on an enterprise database where they were further checked and verified. The patient datasets were exported as Microsoft Excel Spreadsheets to an SPSS database (SPSS for Windows Version 25) and cleaned and checked prior to data analysis.
On the basis that the intervals between the five Likert scale points are equal, item responses were converted into percentages (‘poor’ = 20%, ‘fair’ = 40%, ‘good’ = 60%, ‘very good’ = 80%, ‘excellent’ = 100%) to allow for parametric techniques based on means, standard deviations and variances. Conversion to percentages can aid intelligibility, allow benchmark comparison across different studies and groups as well as highlight differences without the need to represent results to four or five decimal places. Presenting percentages also provides consistency with previous doctor feedback results in the Australian professional performance framework that have been presented in percentages . Two levels of analysis were conducted: at the raw score rater- and item- level (irrespective of doctor rated), and at the aggregated doctor level where doctors received the mean item scores of all their raters. The sampling strategy detailed above has special characteristics that need to be accounted for, that is, the data are unbalanced because of variable numbers of raters per ratee, fully nested because all the ratees may be unique to that rater, and uncrossed because raters provide only one rating per ratee on one occasion. Cronbach’s alpha is reported below, but the alpha results should be interpreted with caution since some of the assumptions of its use (e.g., all raters are rating the same subject, object, or event) are not met in this study. Its use here is to check on the internal consistency of the questionnaire (questionnaire reliability). A signal-to-noise ratio (SNR) measure for dealing with unbalanced, uncrossed and fully nested data is also used to provide an estimate of data reliability . The content and construct validity of the original patient questionnaire were first established in 1999 as the Doctor’s Interpersonal Skill Questionnaire (DISQ) and its validity re-evaluated in 2010 when being assessed for use in the relicensure of doctors by the UK GMC . Its validity and reliability were reaffirmed in 2013 after minor edits were made to the wording of some items and the revised questionnaire applied in unbalanced, uncrossed and fully nested studies involving over 85,000 patients to over 2000 doctors . Analysis of variance (ANOVA) is a collection of statistical models used to analyze the differences among group means. The observed variance in a particular variable is partitioned into components attributable to different sources of variation. In its simplest form, ANOVA provides a statistical test of whether or not the means of several groups are equal. ANOVA is used to test for differences in item ratings and averages within and between PEP, AGPT and GP CPD data. Principal component analysis (PCA) is a data reduction technique for explaining variance in data using a smaller set of variables than the original variables or items. The Kaiser-Meyer-Olkin (KMO) test is a measure to determine sampling adequacy for each item. KMO values between 0.8 and 1.0 indicate that there are enough samples and sufficiently low variance for efficient identification of underlying components. Bartlett’s sphericity tests whether there are relations among variables suitable for structure detection, and PCA is used here to confirm the presence of components previously found when demonstrating criterion and construct validity . Single measures intraclass coefficients (ICCs) provide a relative measure of the variability in the sample of responses and is useful for estimating the agreement between raters on how to interpret the items. Values between 0.4 to 0.6 are considered ‘moderate agreement’, between 0.6 and 0.8 ‘good agreement’ and above 0.8 ‘very good agreement’ . One-way random ICCs are used in this study to check for reliability of the questionnaires given that all the raters are different. A t-test compares the means of two variables to determine whether there is a difference. Such a test can be used to estimate whether the responses given by two populations to a single set of items differ significantly. T-tests assume a normal distribution and the raw score data in this study are negatively skewed. However, its use is justified here because of the large sample sizes and the need to check whether item means differ between groups of doctors after aggregation by doctor (distribution of distributions). Linear regression is used to estimate and control for possible bias in ratings due to sociodemographic factors. These factors are entered first into the regression model against the dependent variable (average patient score) followed by entry of the independent variables (questionnaire items), with comparisons made concerning the amount of variance explained at each step. Psychometric network analysis provides graphical representations of relationships and interactions between variables such as questionnaire items . Nodes in the graph represent the items and links represent the strength of association between them. Inter-item mean score correlations, scaled between 0 and 1 are used, with width of links proportional to the strength of the association. The layout adopted is the force-directed ‘spring’, where variables with strongest associations and therefore of hypothesized strongest influence are placed closer together and at the centre of the graph . Summing the correlations for each item results in a node ‘strength’ measure that can be useful for assessing the influence of items and identifying possible points for future intervention, based on the assumption that changes in central items should have greatest impact on other items. Centrality scores are presented as standard scores (standard deviations above or below mean 0) to allow for comparison across the different doctor groups. All statistical analysis was carried out with SPSS v25 and network analysis through qgraph in R.
Overall, while scores fell in the ‘very good’ to ‘excellent’ range, there were small but statistically significant differences between patient scores for PEP doctors and AGPT registrars at both the raw score and aggregated levels, with PEP doctors scoring lower. GP CPD doctors received the highest scores, especially on items dealing with confidence in their ability and satisfaction with the visit. Patients aged 25 and below gave the lowest scores to all doctor groups. Patients seeing their regular GP gave higher scores than other patients. Internal consistency and reliability of the questionnaires and data were acceptably high. Confirmatory PCA identified the same underlying assessment components as previous, comparable studies. Network analysis revealed that ability to listen was central to patient perceptions of PEP doctors, whereas concern for patient was central to AGPT program doctors. More detailed analysis now follows. Patient data (raw scores) Table provides an overview of the patient data across the three datasets. There were 7907 patient responses to 221 PEP doctors (Dataset A), and 13,623 patient responses for 355 AGPT registrars (Dataset B). The average patient raw score (irrespective of doctor rated) on all 13 items was 90.25% for PEP (SD = 12.92) and 90.98% (SD = 12.08) for AGPT, indicating an overall patient response tending towards the higher end of the ‘very good’ to ‘excellent’ range. Post-hoc power analysis showed 98.1% power for these means, SDs and sample sizes for detecting differences at 0.05 significance level. For both PEP and AGPT patients, the highest scoring item was ‘Respect shown’ (92.24 and 93.15%, respectively), and the lowest ‘Reassurance’ (89.38 and 89.84%, respectively. The average rate of missing responses was very low for both PEP and AGPT patients (0.32, 0.31%), with the highest being for ‘Take care of myself’ (0.9, 0.62%) and the lowest ‘Warmth of greeting’ (0.1, 0.06%). Of the PEP patients, 13.7% were under 25 years of age ( n = 1086), 52.8% between 25 and 59 ( n = 4178) and 31.4% over 60 ( n = 2482). The corresponding AGPT patient figures were 17.1% ( n = 2336), 54.9% ( n = 7480) and 24.7% ( n = 2482), respectively (Fig. ). PEP patients under 25 gave a significantly lower average score (88.4%, p ≤ 0.05) than both patients aged 25–59 (90.7%) and patients over 60 (90.5%). AGPT patients between 25 and 59 gave significantly higher average scores (91.53%, p ≤ 0.05) than patients under 25 (90.99%) and over 60 (90.14%). With respect to gender, 59.4% of PEP patients and 63.7% of AGPT patients were female, while 38.1 and 32.7%, respectively were male. There were 2.5 and 3.6% of PEP and AGPT patients, respectively, who did not declare their gender. Male patients from each pathway gave a significantly lower average score (89.58% PEP, 89.54% AGPT) than their corresponding female patients (90.79% PEP, 91.89% AGPT, p ≤ 0.01). The majority of PEP patients (61.1%) reported that their visit was with their usual doctor, whereas only 28.1% of AGPT patients reported the same. Both groups of patients who saw their usual doctor gave an average score of just over 92%. However, for patients who reported not seeing their usual doctor, there as a marked difference in the average score, with patients seeing PEP participants giving 86.9% in comparison to 90.4% given by patients consulting AGPT registrars (Fig. ). The GP CPD data (Dataset C) consists of 36,215 patient responses to 923 GPs undertaking CPD programs. GP CPD patient questionnaires used the same items as the PEP and AGPT questionnaires with the exception of the item ‘Take care of yourself’. This item was removed since many GPs are typically located in large practice settings where a number of other practice staff (e.g., nurses, practice managers, and physiotherapists) are also involved in this patient aspect. The average patient score on the 12 GP CPD items was 91.39% (SD = 12.27). The highest scoring item was ‘Respect shown’ (92.93%) and the lowest ‘Reassurance’ (90.44%). The average rate of missing responses was 0.9%, with the highest (1.3%) being for ‘Time for visit’ and the lowest (0.4%) for ‘Warmth of greeting’. Analysis of demographic data showed that 8.3% of GP CPD patients were under 25 years of age ( n = 3010), 50% between 25 and 59 ( n = 18,119) and 39.3% over 60 ( n = 14,248). Similarly to the PEP patients, GP CPD patients under 25 gave a significantly lower average score (89.66%, p ≤ 0.01) than patients aged 25–59 (91.57%) and patients over 60 (91.65%). With respect to gender, 60% of GP CPD patients were female, 36.2% were male, and 3.8% did not declare their gender. Male patients gave a significantly lower average score (90.62%, p ≤ 0.01) than female patients (91.93%), as was seen with PEP and AGPT patients. The majority of GP CPD patients (80.1%) reported that their visit was with their usual GP, whereas 15.9% reported that it was not. Patients reporting seeing their usual GP gave a significantly higher score (92.26%) than those who did not see their usual GP (87.5%, p ≤ 0.01). One-way random ICC across all 12–13 items for each of the three patient datasets was 0.77, indicating good agreement among the different raters for interpreting the questionnaire items. Additionally, Cronbach’s alpha was a high 0.97, indicating high internal consistency of the questionnaire irrespective of the type of doctor being rated. The average inter-item correlation varied between r = 0.76 and r = 0.78 for all three datasets. SNR estimates were in the range 0.89 to 0.90 for all three datasets, indicating that 89–90% of the data was likely to be true data and the rest due to noise and error from interactions between raters, items, and ratees. Estimating the effect of patient demographics and item removal For all three doctor groups, patient age and gender contributed less than 0.5% (adjusted R 2 ≤ 0.005) of the variance in average patient scores. Patients seeing their usual doctor contributed 4% (adjusted R 2 = 0.039) to PEP patient average score, less than 1% (adjusted R 2 = 0.009) to AGPT patient average score, and less than 2% (adjusted R 2 = 0.019) for GP CPD patient average score. The 12–13 Likert items contributed the remaining 96 to 98% of the variance. Aggregation of raw score patient data at the doctor level was undertaken without adjustment for demographic factors (Section 3.4 below). The item ‘Take care of myself’, which is not part of the GP CPD items but is part of the PEP and AGPT items, contributed just 0.1% (adjusted R 2 = 0.001) of the variance to PEP and AGPT patient average score after taking into account the other 12 items. Patient GP CPD average scores are therefore unlikely to be impacted by its absence. Principal component analysis of patient data A Kaiser-Meyer-Olkin (KMO) sampling adequacy measure of 0.98 and a significant Bartlett’s test for sphericity ( p ≤ 0.001) indicated that PCA was appropriate. Confirmatory PCA using the varimax rotation method (to spread the highly loaded items across the components) revealed two previously identified primary dimensions known to belong to patient-doctor professional relationships, namely, interpersonal communication (component 1) and caring/empathy (component 2) , thereby establishing criterion (external) validity (Table ). ‘Satisfaction with the visit’ (item 1) was related to interpersonal communication in line with previous studies , thereby establishing construct validity. The amount of variance explained by the two components was over 80% for each group of doctors. PEP, AGPT and GP CPD doctors (mean scores) For PEP doctors there was an average of 35.78 patients per doctor (SD = 4.11, minimum 30, maximum 48, response rate 72%), with a mean PEP doctor score of 90.27 (SD = 6.32, range 60.69–99.16, n = 221). The floor effect based on bottom 15th percentile was 85 and the ceiling effect based on top 15th percentile was 95.71. The lower and upper quartiles were 87.98 and 94.16. For AGPT doctors, there was an average of 38.37 patients per doctor (SD = 8.04, minimum 30, maximum 96, response rate 77%), with a mean AGPT doctor score of 90.99 (SD = 4.87, range 65.6–98.13, n = 355). The floor and ceiling effects were 86.67 and 95.26, and lower and upper quartiles were 88.81 and 94.31, respectively. For GP CPD, the mean score was 91.43 (Average patients per doctor = 39.24, SD = 5.19, range 64.29–100, n = 923, response rate 78%), with floor and ceiling effects of 86.77 and 96.3, and lower and upper quartiles of 88.87 and 95.19, respectively. The average score difference between experienced GPs and trainees was 0.79, with the largest individual item differences being in ‘Confidence in ability’ and ‘Satisfaction with visit’ (Table ). Multiple t-tests showed that PEP doctors received significantly lower item scores than AGPT doctors and GP CPD doctors ( p ≤ 0.01). While there was a tendency for AGPT doctors to receive lower scores on some items than GP CPD doctors, this was not significant ( p = 0.13). Comparison by percentiles (Fig. ) showed that GP CPDs had significantly higher scores than each of the other two doctor groups across all 10 percentiles ( p ≤ 0.01). There was no significant difference in scores by percentile between PEP and AGPT doctors. Network analysis comparisons between doctor groups Psychometric network analysis using correlations between item mean scores revealed that, for PEP and AGPT doctors combined, the central associations were between ‘Concern for patient’, ‘Ability to listen’ and ‘Reassurance provided’ (Fig. , left). ‘Respect shown’ was also strongly associated with ‘Ability to listen’. For GP CPD, several other strong associations were apparent (Fig. , right). In particular, ‘Ability to listen’ was strongly associated with ‘Explanations’ and ‘Concern for patient’. ‘Time for visit’ lay at the periphery of both networks, and ‘Warmth of greeting’ was more peripheral for PEP/AGPT doctors than for GP CPD doctors. When PEP and AGPT doctors were separated, ‘Ability to listen’ was central in the PEP network with strong links to ‘Expressing concern’ and ‘Taking care of myself’ (Fig. , left). For the AGPT network, ‘Concern for patient’ was central with strong links to ‘Taking care of myself’, ‘Consideration’ and ‘Recommendation’. Both structures reveal strong links between ‘Confidence in ability’ and ‘Satisfaction with visit’. When PEP doctors with scores within the bottom 10th percentile (≤82.19%) were compared with PEP doctors with scores in the top 10th percentile (≥96.2%), caring/empathy items were located more centrally for the former group (Fig. , left) while interpersonal skills were more central for the latter (Fig. , right). In particular, ‘Reassurance’, ‘Ability to listen’, ‘Warmth of greeting’ and ‘Explanations’ formed a tight central cluster for the top PEP doctors. When AGPT doctors with scores in the bottom 10th percentile (≤85.15%) were compared with AGPT doctors with scores in the top 10th percentile (≥95.76%), interpersonal skills consisting of ‘Confidence in ability’, ‘Warmth of greeting’, ‘Reassurance’ and ‘Explanations’ formed a central core for the latter group (Fig. , right). ‘Ability to listen’, ‘Explanations’ and ‘Concern for patients’ formed a central core for the former group (Fig. , left). Standardized strength values are shown in Fig. for each of the doctor networks (GP CPD, PEP, AGPT) and indicate that ‘Ability to listen’ has strong connections in all three networks, followed by ‘Concern for patient’ and ‘Consideration’. Weakest nodes in terms of influence are ‘Warmth of greeting’ and ‘Time for visit’.
Table provides an overview of the patient data across the three datasets. There were 7907 patient responses to 221 PEP doctors (Dataset A), and 13,623 patient responses for 355 AGPT registrars (Dataset B). The average patient raw score (irrespective of doctor rated) on all 13 items was 90.25% for PEP (SD = 12.92) and 90.98% (SD = 12.08) for AGPT, indicating an overall patient response tending towards the higher end of the ‘very good’ to ‘excellent’ range. Post-hoc power analysis showed 98.1% power for these means, SDs and sample sizes for detecting differences at 0.05 significance level. For both PEP and AGPT patients, the highest scoring item was ‘Respect shown’ (92.24 and 93.15%, respectively), and the lowest ‘Reassurance’ (89.38 and 89.84%, respectively. The average rate of missing responses was very low for both PEP and AGPT patients (0.32, 0.31%), with the highest being for ‘Take care of myself’ (0.9, 0.62%) and the lowest ‘Warmth of greeting’ (0.1, 0.06%). Of the PEP patients, 13.7% were under 25 years of age ( n = 1086), 52.8% between 25 and 59 ( n = 4178) and 31.4% over 60 ( n = 2482). The corresponding AGPT patient figures were 17.1% ( n = 2336), 54.9% ( n = 7480) and 24.7% ( n = 2482), respectively (Fig. ). PEP patients under 25 gave a significantly lower average score (88.4%, p ≤ 0.05) than both patients aged 25–59 (90.7%) and patients over 60 (90.5%). AGPT patients between 25 and 59 gave significantly higher average scores (91.53%, p ≤ 0.05) than patients under 25 (90.99%) and over 60 (90.14%). With respect to gender, 59.4% of PEP patients and 63.7% of AGPT patients were female, while 38.1 and 32.7%, respectively were male. There were 2.5 and 3.6% of PEP and AGPT patients, respectively, who did not declare their gender. Male patients from each pathway gave a significantly lower average score (89.58% PEP, 89.54% AGPT) than their corresponding female patients (90.79% PEP, 91.89% AGPT, p ≤ 0.01). The majority of PEP patients (61.1%) reported that their visit was with their usual doctor, whereas only 28.1% of AGPT patients reported the same. Both groups of patients who saw their usual doctor gave an average score of just over 92%. However, for patients who reported not seeing their usual doctor, there as a marked difference in the average score, with patients seeing PEP participants giving 86.9% in comparison to 90.4% given by patients consulting AGPT registrars (Fig. ). The GP CPD data (Dataset C) consists of 36,215 patient responses to 923 GPs undertaking CPD programs. GP CPD patient questionnaires used the same items as the PEP and AGPT questionnaires with the exception of the item ‘Take care of yourself’. This item was removed since many GPs are typically located in large practice settings where a number of other practice staff (e.g., nurses, practice managers, and physiotherapists) are also involved in this patient aspect. The average patient score on the 12 GP CPD items was 91.39% (SD = 12.27). The highest scoring item was ‘Respect shown’ (92.93%) and the lowest ‘Reassurance’ (90.44%). The average rate of missing responses was 0.9%, with the highest (1.3%) being for ‘Time for visit’ and the lowest (0.4%) for ‘Warmth of greeting’. Analysis of demographic data showed that 8.3% of GP CPD patients were under 25 years of age ( n = 3010), 50% between 25 and 59 ( n = 18,119) and 39.3% over 60 ( n = 14,248). Similarly to the PEP patients, GP CPD patients under 25 gave a significantly lower average score (89.66%, p ≤ 0.01) than patients aged 25–59 (91.57%) and patients over 60 (91.65%). With respect to gender, 60% of GP CPD patients were female, 36.2% were male, and 3.8% did not declare their gender. Male patients gave a significantly lower average score (90.62%, p ≤ 0.01) than female patients (91.93%), as was seen with PEP and AGPT patients. The majority of GP CPD patients (80.1%) reported that their visit was with their usual GP, whereas 15.9% reported that it was not. Patients reporting seeing their usual GP gave a significantly higher score (92.26%) than those who did not see their usual GP (87.5%, p ≤ 0.01). One-way random ICC across all 12–13 items for each of the three patient datasets was 0.77, indicating good agreement among the different raters for interpreting the questionnaire items. Additionally, Cronbach’s alpha was a high 0.97, indicating high internal consistency of the questionnaire irrespective of the type of doctor being rated. The average inter-item correlation varied between r = 0.76 and r = 0.78 for all three datasets. SNR estimates were in the range 0.89 to 0.90 for all three datasets, indicating that 89–90% of the data was likely to be true data and the rest due to noise and error from interactions between raters, items, and ratees.
For all three doctor groups, patient age and gender contributed less than 0.5% (adjusted R 2 ≤ 0.005) of the variance in average patient scores. Patients seeing their usual doctor contributed 4% (adjusted R 2 = 0.039) to PEP patient average score, less than 1% (adjusted R 2 = 0.009) to AGPT patient average score, and less than 2% (adjusted R 2 = 0.019) for GP CPD patient average score. The 12–13 Likert items contributed the remaining 96 to 98% of the variance. Aggregation of raw score patient data at the doctor level was undertaken without adjustment for demographic factors (Section 3.4 below). The item ‘Take care of myself’, which is not part of the GP CPD items but is part of the PEP and AGPT items, contributed just 0.1% (adjusted R 2 = 0.001) of the variance to PEP and AGPT patient average score after taking into account the other 12 items. Patient GP CPD average scores are therefore unlikely to be impacted by its absence.
A Kaiser-Meyer-Olkin (KMO) sampling adequacy measure of 0.98 and a significant Bartlett’s test for sphericity ( p ≤ 0.001) indicated that PCA was appropriate. Confirmatory PCA using the varimax rotation method (to spread the highly loaded items across the components) revealed two previously identified primary dimensions known to belong to patient-doctor professional relationships, namely, interpersonal communication (component 1) and caring/empathy (component 2) , thereby establishing criterion (external) validity (Table ). ‘Satisfaction with the visit’ (item 1) was related to interpersonal communication in line with previous studies , thereby establishing construct validity. The amount of variance explained by the two components was over 80% for each group of doctors.
For PEP doctors there was an average of 35.78 patients per doctor (SD = 4.11, minimum 30, maximum 48, response rate 72%), with a mean PEP doctor score of 90.27 (SD = 6.32, range 60.69–99.16, n = 221). The floor effect based on bottom 15th percentile was 85 and the ceiling effect based on top 15th percentile was 95.71. The lower and upper quartiles were 87.98 and 94.16. For AGPT doctors, there was an average of 38.37 patients per doctor (SD = 8.04, minimum 30, maximum 96, response rate 77%), with a mean AGPT doctor score of 90.99 (SD = 4.87, range 65.6–98.13, n = 355). The floor and ceiling effects were 86.67 and 95.26, and lower and upper quartiles were 88.81 and 94.31, respectively. For GP CPD, the mean score was 91.43 (Average patients per doctor = 39.24, SD = 5.19, range 64.29–100, n = 923, response rate 78%), with floor and ceiling effects of 86.77 and 96.3, and lower and upper quartiles of 88.87 and 95.19, respectively. The average score difference between experienced GPs and trainees was 0.79, with the largest individual item differences being in ‘Confidence in ability’ and ‘Satisfaction with visit’ (Table ). Multiple t-tests showed that PEP doctors received significantly lower item scores than AGPT doctors and GP CPD doctors ( p ≤ 0.01). While there was a tendency for AGPT doctors to receive lower scores on some items than GP CPD doctors, this was not significant ( p = 0.13). Comparison by percentiles (Fig. ) showed that GP CPDs had significantly higher scores than each of the other two doctor groups across all 10 percentiles ( p ≤ 0.01). There was no significant difference in scores by percentile between PEP and AGPT doctors.
Psychometric network analysis using correlations between item mean scores revealed that, for PEP and AGPT doctors combined, the central associations were between ‘Concern for patient’, ‘Ability to listen’ and ‘Reassurance provided’ (Fig. , left). ‘Respect shown’ was also strongly associated with ‘Ability to listen’. For GP CPD, several other strong associations were apparent (Fig. , right). In particular, ‘Ability to listen’ was strongly associated with ‘Explanations’ and ‘Concern for patient’. ‘Time for visit’ lay at the periphery of both networks, and ‘Warmth of greeting’ was more peripheral for PEP/AGPT doctors than for GP CPD doctors. When PEP and AGPT doctors were separated, ‘Ability to listen’ was central in the PEP network with strong links to ‘Expressing concern’ and ‘Taking care of myself’ (Fig. , left). For the AGPT network, ‘Concern for patient’ was central with strong links to ‘Taking care of myself’, ‘Consideration’ and ‘Recommendation’. Both structures reveal strong links between ‘Confidence in ability’ and ‘Satisfaction with visit’. When PEP doctors with scores within the bottom 10th percentile (≤82.19%) were compared with PEP doctors with scores in the top 10th percentile (≥96.2%), caring/empathy items were located more centrally for the former group (Fig. , left) while interpersonal skills were more central for the latter (Fig. , right). In particular, ‘Reassurance’, ‘Ability to listen’, ‘Warmth of greeting’ and ‘Explanations’ formed a tight central cluster for the top PEP doctors. When AGPT doctors with scores in the bottom 10th percentile (≤85.15%) were compared with AGPT doctors with scores in the top 10th percentile (≥95.76%), interpersonal skills consisting of ‘Confidence in ability’, ‘Warmth of greeting’, ‘Reassurance’ and ‘Explanations’ formed a central core for the latter group (Fig. , right). ‘Ability to listen’, ‘Explanations’ and ‘Concern for patients’ formed a central core for the former group (Fig. , left). Standardized strength values are shown in Fig. for each of the doctor networks (GP CPD, PEP, AGPT) and indicate that ‘Ability to listen’ has strong connections in all three networks, followed by ‘Concern for patient’ and ‘Consideration’. Weakest nodes in terms of influence are ‘Warmth of greeting’ and ‘Time for visit’.
The results presented here further our understanding of the communication skills and professionalism of doctors undertaking GP training, as perceived by their patients. This is also the first comparison of these skills between doctors on the AGPT and PEP pathways to RACGP Fellowship. With over 21,000 patient responses to 576 doctors undertaking GP training, and over 36,000 patient responses to over 900 GPs, the results presented here would appear to have validity in terms of margin for error and representativeness. This is also partially supported by consistency analysis of the data, which shows good agreement among patients about how to interpret the questionnaire, as well as the power analysis. Patient response rates for questionnaires completed through convenience sampling on site vary from 72 to 78%. These compare favourably with postal response rates (typically 20 to 60%) and are in line with previous patient satisfaction studies as well as considered ‘high’ to ‘very high’ in the context of minimizing potential for non-response bias . One of the impacts of ‘big data’ is that small differences between groups tend to be identified as significant because of the large numbers involved. Differences that appear minor with limited sample sizes can become statistically significant as the quantity of data grows, enabling finer significant discriminations to be made . For instance, the difference between an average GP CPD score and an average PEP/AGPT score is only 0.8% (Table ). The discussion below needs to be interpreted in the context of all three doctor groups achieving scores in the very good to excellent range (averages over 90%). Nevertheless, the small differences that are statistically significant may be useful for identifying trends that have functional significance for training programs, as identified below. Patients were most satisfied with their experience with GP CPD doctors, followed by AGPT and PEP doctors (Table ). In particular, patients had greatest confidence in the ability of GP CPD doctors. Patients were more satisfied with AGPT doctors than PEP doctors on ‘Explanations’, ‘Time for visit’, ‘Express concerns’ and ‘Ability to listen’. Percentile analysis showed patients rated AGPT doctors higher than PEP doctors until the 80th percentile (Fig. ). Patients rated the very top PEP doctors (90th percentile) as better than the top AGPT doctors, with both still rated below GP CPD doctors. Patients rated AGPT doctors better than GP CPD doctors at the very lowest 10th percentile (Fig. , 85.15% vs 84.95). Female patients gave higher scores than male patients, and patients gave higher scores for visits to their usual doctor (Fig. ). These aspects could have benefitted PEP doctors due to greater proportion of such patients in comparison to AGPT doctors (Table ). PEP doctors are already working in General Practice on entry to their program, whilst AGPT registrars are placed into a practice on entry and so do not have an established patient load. Patients rated their doctors under two, equally balanced, previously identified components of interpersonal communication and caring/empathy (Table ). These components appear to be consistent across all three doctor groups studied here. Network analysis showed that all doctor groups had strong connections between ‘Concern for patient’ and ‘Consideration’ (Fig. ). ‘Reassurance’ and ‘Confidence in ability’ were also strongly linked, based on patient feedback, for PEP and AGPT doctors (Fig. , left). For GP CPD doctors, strong links were demonstrated between ‘Ability to listen’, ‘Explanations’ and ‘Reassurance’ (Fig. , right). ‘Ability to listen’ was also linked strongly with ‘Concern for patient’, ‘Consideration’, ‘Reassurance’ and ‘Express concerns’. When separate networks for PEP and AGPT doctors were compared (Fig. ), ‘Respect’ was central for PEP doctors, with strong links to ‘Concern for patient’, ‘Take care of myself’ and ‘Confidence in ability’. For APGT doctors, ‘Concern for patient’ was central, with strong links to ‘Take care of myself’, ‘Consideration’ and ‘Recommendation’. The lowest scoring PEP doctors were distinguished from the top scoring PEP doctors by the centrality of care/empathy items for the former group and interpersonal communication skills for the latter group (Fig. ). This pattern was repeated to some extent for AGPT doctors (Fig. ). GP CPD doctors were identified by ‘Ability to listen’ being central and strongly related to other items (Fig. , right). Future studies could usefully study the relationship between empathy and caring on the one hand, and communication and interpersonal skills on the other, to identify ways in which practitioners may be able to better communicate that they care so that patients gain more confidence in the diagnosis and advice provided. For instance, methods involving scheduled follow-up discussions either via email or in person, or requesting feedback from the patient on how the management regime is progressing, could be possible ways to demonstrate empathy and care through further communication. Studies focused on these particular aspects of care, confidence and communication could lead to the gap between the lowest scoring doctors and the highest scoring doctors reducing even further. One implication of these results is that the perception that IMGs and OTDs may lack the ‘soft skills’ to successfully practice in their new country will need revising. Our results show that IMGs and OTDs perform similarly with respect to communication and professionalism skills as their locally trained counterparts. However, our network analysis indicates that there may be issues of ‘connectedness’ and difference in priority between such skills that may need further exploration. In particular, the relationship between interpersonal communication and caring/empathy dimensions can vary according to background and training. This is consistent with the literature indicating that IMGs and OTDs can have difficulty adjusting to new cultures, communication styles, languages (including slang), health systems and health beliefs . A practical suggestion may be for training programs to be enhanced to integrate interpersonal skills and caring/empathy skills more fully, with the GP CPD network being used as a benchmark, to complement recent and similar recommendations for changes in medical undergraduate courses . There is also growing interest in the use of feedback for debriefing and development purposes . For instance, anonymous patient-doctor sessions could be recorded and PEP/AGPT doctors then asked to rate the patient’s experience using the patient questionnaire, with comparisons made against real patient data (all subject to ethical approval and permission of all parties concerned). Network analysis indicates that focusing on ability to listen and concern for patient (the two most central items) might be useful for enhancing the ability of PEP and AGPT doctors to appreciate the importance of communication and care/empathy for patient-centredness, which will likely benefit their registration pathway process. Given that the PEP program is oriented towards self-directed education with variable supervisory arrangements, as well as located predominantly outside major cities, a challenge for future program development may be to identify methods for enhancing mechanisms, such as greater contact with experienced GPs, for helping trainee GPs to enhance their communication and interpersonal skills as identified above. Interpersonal communication is now accepted as a fundamental clinical skill in medical practice , with good communication establishing trust between patient and doctor as well as leading to better exchange of information. Listening, explaining and empathizing can have a major effect on patient health status and satisfaction. The psychometric network for experienced GPs (Fig. , right), for instance, shows a tight and central clustering of interpersonal communication component items (‘Ability to listen’, ‘Explanation provided’) with empathy component items (‘Expressing concerns’, ‘Consideration’, ‘Concern for patient’). Experienced GPs also receive the highest satisfaction ratings. Networks for PEP and AGPT doctors show different relationships between items, leading to speculative hypotheses and interpretations concerning differences between doctor training groups in terms of possible clinical performance in comparison with experienced GPs. However, in the absence of other sources of data concerning clinical effectiveness of consultation, these networks only identify possible areas for changes in training programs and additional support for doctor groups, as discussed above. No conclusions can be drawn from these networks concerning the clinical effectiveness of consultations by any doctor or doctor group, or how these differences affect clinical treatment of patients and patient satisfaction. The aim of the research was to understand how doctors undertaking the PEP and AGPT pathways to GP Fellowship, and Fellowed GPs compare regarding communication skills and professionalism. This research demonstrates the high quality of patient care given by PEP and AGPT doctors, as well as Fellowed GPs, and highlights the interrelationships between professional skills, including which skills are focal or central to each doctor group. Overall, each group of doctors has excellent performance, and doctors on GP Fellowship pathways can aspire to consolidate their skills cohesively to further improve their performance, as seen with the experienced GPs. Given that the PEP program is oriented towards self-directed education with variable supervisory arrangements in geographically diverse practice locations, a challenge for future program development may be to identify methods for strengthening mechanisms, such as greater contact with experienced GPs, to help trainee GPs enhance their communication and interpersonal skills as identified above. This recommendation is in line with colleague feedback obtained for the same group of trainees, which showed that colleagues, while rating the clinical skills of PEP trainees highly, identified a gap in communication skills in comparison with AGPT GPiT .
Limitations of this study include the variable numbers of doctors used for each part of the analysis due to data being collected at different times for such a large-scale study. The later stage of data collection (i.e., early to mid-2020) was affected by COVID-19, leading to early termination of data collection. While this study reports on the quantitative aspects of the study, further work involving observations and qualitative analysis, including qualitative analysis of comments supplied by patients, is required to identify specific behavioural patterns of doctors that may affect ratings provided. Additionally, there is limited demographic data available for PEP doctors, and due to the eligibility criteria it has been assumed that the majority of doctors undertaking this pathway to RACGP Fellowship are IMGs and OTDs. While some aspects of patient demographics were taken into account in the analysis, no sociodemographic or ethnic aspects of individual doctors being rated were collected to ensure lack of personal identification. There may be bias against doctors based on sociodemographic and ethnic factors, although the small difference in average ratings between the two doctor groups would suggest that differences between patient groups were larger than differences between doctor groups. The possible effects of such bias on ratings are not measured in this study. Finally, while response rates through convenience sampling are high to very high, there is unknown potential bias in non-responses which can limit the generalizability of the results to other patient populations, such as questionnaires only being completed if patients were satisfied with their visit or, conversely, patients more likely to complete their questionnaire because they were unhappy with their visit. An assumption made in this study is that any patient bias is randomly distributed and contributes equally to all doctor ratings.
Additional file 1. Questionnaire Items (Long and Short Versions).
|
Through the Looking-Glass: Psychoneuroimmunology and the Microbiome-Gut-Brain Axis in the Modern Antiretroviral Therapy Era | 21be52df-a6ef-4c69-ad23-b68aebe5fbc8 | 9553251 | Physiology[mh] | The central nervous system, neuroendocrine signaling, and the immune system dynamically interact in human immunodeficiency virus (HIV) pathogenesis . Even before the advent of highly active antiretroviral therapy (HAART), psychoneuroimmunology (PNI) research uncovered neuroendocrine mechanisms whereby psychological factors influence clinical HIV progression . In the HAART era (i.e., 1997–2006), the associations of psychological factors with HIV disease markers and clinical HIV progression often persisted after adjusting for medication adherence . Although informative, HAART-era studies mostly examined neuroendocrine mechanisms linking psychological factors with HIV pathogenesis (i.e., mind-to-body pathways whereby mental health and stress influenced HIV disease markers). Relatively few studies characterized the pathways whereby HIV-associated pathophysiologic alterations amplify the risk of depression, substance use disorders, and other neuropsychiatric disorders (i.e., body-to-mind pathways) . This unidirectional approach diminished our ability to test integrated models to guide the development of biobehavioral treatments for these prevalent comorbidities among people with HIV (PWH) . During the past 15 years (2007–present), profound advances in the medical management of HIV ushered in the modern antiretroviral therapy (ART) era where current treatment guidelines recommend ART at any CD4+ T-cell count . Modern ART regimens are less burdensome such that many PWH take one pill daily with far fewer adverse effects and can achieve viral suppression (<200 copies/ml) at much lower levels of adherence. Consequently, rates of viral suppression have drastically increased, with 86% of those receiving HIV clinical care at eight sites across the United States were virally suppressed by 2015 . Although PWH who have depression and PWH who use stimulants such as methamphetamine may achieve viral suppression more slowly , these priority populations are displaying unprecedently high rates of viral suppression. This provides new opportunities to examine PNI pathways in PWH who are virally suppressed to mitigate (but not eliminate) the influence of ART adherence . Multilevel determinants are implicated in the etiology and maintenance of depressive disorders among PWH . Depression is a prevalent comorbidity, affecting as many as one-third of PWH (31%; 95% confidence interval = 28%–34%) . Greater risk of suicide has also been well characterized among PWH. Although suicide risk decreased somewhat with the advent of HAART , the suicide rate remains threefold higher among PWH than the general Swiss population . Recent evidence indicates that multiple indicators of suicide risk are elevated among PWH around the globe . Advancing our understanding of the complex etiology of depression and suicide risk in PWH will require more comprehensive efforts to understand the relevance of HIV-associated biological alterations in the context of key structural and social determinants. Substance use disorders are prevalent comorbidities among PWH that increase the risk of HIV seroconversion, complicate HIV disease management, and fuel faster clinical HIV progression . PWH who use stimulants such as methamphetamine display faster clinical HIV disease progression, even after adjusting for ART adherence and viral load . PNI studies to characterize the bidirectional, neuroimmune consequences of co-occurring stimulant use disorders and HIV are needed to catalyze the development of novel, biobehavioral treatments . Similarly, there may be clinically meaningful neuroimmune predictors and consequences of tobacco, alcohol, and other substance use that warrant further study in PWH . The overarching goal of this narrative review is to provide a framework to guide PNI research in the modern ART era with a focus on microbiome-gut-brain axis interactions. There are new opportunities to investigate the potentially bidirectional mechanisms linking depression, substance use disorders, and other neuropsychiatric comorbidities with HIV-associated pathophysiologic alterations that persist despite effective ART . To accomplish this goal, we first examine seminal PNI studies conducted through the HAART era. Many studies yielded provocative, mechanistic results that can guide PNI research examining the role of neuroendocrine signaling in the reciprocal interactions between gastrointestinal tract (and its resident microorganisms), immune system, and central nervous system, which will be referred to as the microbiome-gut-brain axis. We step “through the looking-glass” to envision a new generation of PNI studies that leverage our conceptual model of the microbiome-gut-brain axis in PWH receiving effective ART. This narrative review concludes with recommendations for a PNI research agenda to investigate the bidirectional pathways linking the microbiome-gut-brain axis and neuropsychiatric comorbidities in PWH. PNI studies conducted through the HAART era delineated neuroendocrine mechanisms whereby psychological factors are linked to clinical HIV progression. There was convergent evidence from multiple cohort studies that trauma, stress, and depression predict faster clinical HIV progression . On the other hand, positive psychological factors such as positive affect, finding meaning, and spirituality predicted slower clinical HIV progression . These effects are thought to be partially attributable to alterations in neuroendocrine signaling, with a primary focus on hypothalamic-pituitary-adrenal (HPA) axis function and autonomic nervous system activation. Clinical studies conducted during the HAART era provided support for the premise that alterations in the HPA axis and the autonomic nervous system mediate the effects of psychological factors on immune function in PWH. Randomized controlled trials indicated that decreases in urinary cortisol and norepinephrine (NE) output partially mediated the effects of a cognitive-behavioral stress management intervention on HIV disease markers such as subsets of naive CD4+ T cells . This was further supported by findings indicating that elevated urinary cortisol and NE predicted higher viral load after adjusting for HAART adherence , and greater autonomic nervous system activation at rest predicted poorer CD4+ T-cell recovery and decreased viral suppression in response to HAART . There was also evidence that higher urinary cortisol was associated with fewer subsets of naive CD4+ T cells in the HAART era , and more diurnal variation in salivary cortisol (thought to be a healthy pattern) was associated with less CD4+ T-cell and CD8+ T-cell activation in PWH who were ART naive . The relevance of the autonomic nervous system in HIV pathogenesis is further supported by a reactivity study examining a speech stressor task where PWH (compared with people without HIV) displayed greater stress-induced increases in activated CD8+ CD38+ T cells, smaller increases in natural killer cell number and cytotoxicity, and suppression of the T-cell lymphoproliferative response to phytohemagglutinin mitogen, which are each indicative of immune dysregulation . The associations of catecholamines such as epinephrine with these stress-induced immunologic changes varied by HIV status, suggesting functional alterations in the communication of the sympathetic nervous system with the immune system among PWH. There is compelling evidence that sympathetic nervous system innervation of lymphoid organs (e.g., lymph nodes) modulates HIV replication and potentially the HIV reservoir (i.e., immune cells, organs, and lymphoid tissue where HIV replication persists despite viral suppression). NE released from the sympathetic nervous system nerve terminals that juncture with lymphoid tissue binds with β 2 receptors on the lymphocyte membrane, thereby inducing cellular changes in the lymphocyte via the G protein–linked adenyl cyclase–cAMP–protein kinase A signaling cascade . In vitro data have shown that cellular changes of this nature are associated with decrements in interferon-γ and interleukin-10, which in turn, are linked to elevations in HIV viral load . Results from other in vitro studies indicate that similar pathways underlie sympathetic nervous system induced reactivation of human herpesvirus-8 and cytomegalovirus (CMV), two viruses from the herpesvirus family linked to clinical HIV progression . PNI studies are needed to determine if behavioral and psychological factors can influence the HIV reservoir as well as reactivate herpesvirus coinfections by modulating the autonomic nervous system. Simian immunodeficiency virus (SIV) studies support stress-related alterations in communication between the sympathetic nervous system and immune system. There is plasticity in the density of sympathetic nervous system catecholaminergic neural fibers that juncture with the lymph nodes, which is modified by social stress in rhesus macaques . In fact, SIV replication is increased by 3.9-fold in the vicinity of catecholaminergic varicosities with the lymph nodes . Interestingly, there is some indication of specificity for the effect of social stress as related to antiviral responses. Although methamphetamine administration amplified sympathetic nervous system activation, only social instability increased the density of catecholaminergic nerve fibers in lymph nodes while downregulating the expression of genes associated with innate antiviral responses and upregulating inflammatory genes in rhesus macaques . Another recent study of acute SIV in pigtailed macaques observed that singly housed (versus socially housed) animals displayed higher viral load in plasma and cerebrospinal fluid, greater CD4+ T-cell count declines, and more CD4+ and CD8+ T-cell activation . Taken together, these studies underscore the potentially central importance of social processes in HIV pathogenesis. Although social support was extensively examined in early PNI studies with PWH, findings were generally mixed and varied as a function of disease stage as well as mode of transmission . These well-characterized mechanisms will continue to guide the scientific premise of PNI research, and there are four important opportunities for building upon these seminal PNI studies in the modern ART era. First, substantial advances in our understanding of HIV-associated pathophysiologic alterations can guide efforts to characterize the complex, bidirectional connections between neuroendocrine signaling and the microbiome-gut-brain axis. Second, it is unclear to what extent neuroendocrine signaling will continue to influence HIV pathogenesis in the modern ART era. Third, findings from SIV studies underscore the need for more robust efforts in human studies to characterize the relevance of social adversity, which is only beginning to be examined in PNI research conducted in PWH. Fourth, many PNI studies used stringent inclusion criteria that increased scientific rigor at the expense of generalizability (e.g., conducted mostly in sexual minority men, excluded substance users). Studies are needed to characterize the relevance of PNI mechanisms in more representative samples of PWH in the modern ART era. In the sections hereinafter, we review evidence supporting a microbiome-gut-brain axis model relevant to depression, substance use disorders, and other neuropsychiatric disorders in PWH (Figure ). We begin by providing an overview of what is known about the composition of the gut microbiota in PWH and the relevance of modifiable factors like stimulant use on microbiota composition and function. This section also highlights the importance of direct vagal afferent communication between the gut microbiota and the brain in modifying oxytocin release, mood, and behaviors relevant to depression and substance use disorders. Next, we describe the importance of persistent gut-immune dysregulation in PWH with a focus on indirect , immune-mediated pathways that could account for microbiome-gut-brain axis interactions. Then, we present findings regarding the relevance of peripheral immune dysregulation for depression in PWH and how this may be amplified by experiences relevant to social adversity. Finally, we briefly review evidence suggesting that neuroendocrine signaling can modify the microbiome-gut-brain axis in the modern ART era. The Gut Microbiota in PWH The gut microbiota include trillions of bacteria, eukaryotes, and viruses . These microorganisms play a vital role in maintaining neuroimmune homeostasis and are associated with health outcomes, including HIV disease progression . Most research has focused on bacteria, which can influence neural, endocrine, and immune cells in the gut, periphery, and brain through the release of bacterial metabolites or via direct contact of bacteria (or bacterial products) with host cells . This can lead to chronic peripheral and neuroinflammation, dysregulated neurotransmitter synthesis, and altered afferent vagal nerve stimulation. In addition, gut bacteria can influence intestinal permeability, impact blood-brain-barrier integrity, modify nutrient absorption, alter the growth of pathogenic bacteria, and take part in drug metabolism, all of which are relevant to neuroimmune functioning among PWH . Many of these pathways whereby gut microbiota could potentiate depression, substance use disorders, and other neuropsychiatric comorbidities that compromise HIV disease management remain largely unexamined. PWH show differences in the composition and function of the gut microbiota compared with people without HIV . Microbiome alterations are more prominent among untreated PWH, but disruptions persist despite effective ART . The gut microbiome generally shows a pattern of decreased community diversity as well as enrichment of bacterial taxa linked to inflammation, microbial translocation, and poor CD4+ T-cell restoration. There is also evidence of depletion of bacterial taxa that are inversely associated with these markers . In addition to the role of HIV infection in altering the gut environment, the gut microbiome can be modified by a variety of behavioral and psychological factors, but studies in PWH are generally lacking or correlative. For example, stimulant use and alcohol use are associated with altered gut microbiota among PWH . There is also increasing interest in the mechanisms whereby microbiome-related alterations in reward processing influence the maintenance and severity of various neuropsychiatric disorders, including substance use disorders . Further research is needed to characterize the bidirectional microbiome-gut-brain axis mechanisms relevant to problematic patterns of alcohol or other substance use among PWH. Gut Microbiota and Vagal Nerve Stimulation The vagus nerve innervates the gut, providing bidirectional (80% afferent, 20% efferent) communication with the central nervous system. Vagal tone is essential for decreasing gut permeability so that microbial products like lipopolysaccharide (LPS) do not leak into the periphery to amplify residual immune dysregulation. A recent study of PWH who presented with gastrointestinal symptoms observed that vagal dysfunction was associated with bacterial outgrowth in the small intestines that was, in turn, linked to greater systemic inflammation . Efferent vagal pathways that are modulated by decreases in depressive symptoms have the potential to alter the gut microbiota and decrease microbial translocation, as evidenced by the effects of vagal stimulation . Prior clinical research in people without HIV observed that increased vagal tone is an important indicator of a favorable response to cognitive-behavioral treatments for depression . PWH may also experience autonomic neuropathy , which is linked to vagal dysfunction and microbiome-gut-brain axis alterations. Gut microbiota can also modulate mood and behaviors by altering stimulation of the vagal nerve. For example, gut bacteria release metabolites (or induce host cells to release metabolites) that stimulate the vagal nerve afferents that alter hypothalamic function to promote the release of oxytocin (a neuropeptide that influences social behavior, metabolism, and wound healing) in the brain. In fact, Lactobacillus reuteri have been shown to activate vagal afferents to potentiate oxytocin release from the paraventricular nucleus of the hypothalamus . This process promotes the coordinated activity of oxytocin and serotonin in the nucleus accumbens to increase the experience of social reward , which could have important implications for anhedonia, a common symptom of many neuropsychiatric disorders, including substance use disorders and depression . Although oxytocin has not been extensively examined among PWH, there is some evidence of nonlinear associations in PNI studies . Higher oxytocin levels have been associated with more severe depressive symptoms in PWH , potentially reflecting a compensatory response to depressive symptoms. In another cross-sectional study, the association of elevated stress with lower CD4+ T-cell count was observed only at higher levels of oxytocin among low-income ethnic minority women with HIV . There is a clear need for microbiome-gut-brain axis studies to understand the relevance of oxytocin in context of other important neuroimmune pathways among PWH. The Central Importance of Gut-Immune Dysfunction in HIV Pathogenesis Damage to the gut during acute HIV is thought to be a primary driver of alterations in gut microbiota , expansion of the enteric virome , depletion of CD4+ T cells , and increased gut permeability . Most notably, there is a preferential depletion of T-helper 17 cells (Th 17 ) cells that are essential for maintaining gut barrier integrity , and this is currently irreversible unless ART is initiated early in acute HIV infection . Further research is needed to understand how key alterations to microbiome-gut-brain axis pathways could amplify the risk of persistent neuropsychiatric comorbidities during acute HIV infection and after suppressive ART. HIV-induced damage to the gut results in the translocation of microbial products such as LPS and peptidoglycan (PGN) across the gut barrier. When these microbial products cross the damaged gut barrier and enter the periphery, they chronically stimulate immune cells, which causes ongoing immune activation and inflammation in PWH . Despite the prominent focus on microbial translocation of LPS, PGN is also linked to inflammation in the gut and brain . Fragments of PGN are released into circulation as byproducts of bacterial cell remodeling and are typically recognized by pattern recognition receptors and other PGN-specific recognition proteins on host immune cells . Although the role of PGN in HIV has not been well established, PGN has been implicated in autoimmune diseases such as Crohn disease , rheumatoid arthritis , and multiple sclerosis . The potential implications for neuroinflammation are supported by observations of high concentrations of PGN in brain lesions in people with multiple sclerosis . Given its role in neuroimmune modulation, studies are needed to better understand the relevance of PGN for depression, substance use disorders, and other neuropsychiatric disorders among PWH . There is growing evidence for interactions between stimulant use and microbiome-gut-brain axis pathways via the gut-immune dysregulation. One such pathway is through the impact of stimulant-induced microbial translocation and gut barrier dysfunction. Stimulants such as methamphetamine can decrease parasympathetic tone , which could increase intestinal permeability. Stimulants also directly damage gut barrier integrity in self-administering HIV-1 transgenic rats and were associated with depleted tryptophan (an essential amino acid precursor for serotonin) in PWH receiving HAART . Tryptophan catabolites could damage gut barrier integrity through several mechanisms described hereinafter . Bearing in mind that there are currently no treatments for stimulant use disorders approved by the US Food and Drug Administration, mechanistic studies characterizing these complex microbiome-gut-brain axis pathways could identify novel targets for pharmacologic interventions in PWH who have stimulant use disorders that serve as a key obstacle to sustained viral suppression. Another mechanism whereby the gut microbiota can alter immune surveillance is inducing the enzyme indoleamine 2,3-dioxygenase 1 (IDO-1) to catabolize tryptophan . Tryptophan catabolites influence immune function in the periphery and are neuroactive. PWH display increases in tryptophan catabolism that are linked to depressive symptoms and only partially normalized by ART . Tryptophan catabolism may also be exacerbated among those with protein-deficient diets, highlighting the importance of food insecurity in the intersection of biological and structural determinants relevant to depression among PWH . Examining the role of tryptophan catabolism in the microbiome-gut-brain axis in the modern ART era could help address the underlying HIV-associated biological alterations that modify depression and suicide risk to optimize mental health treatment for PWH. Because PWH have more gut bacterial taxa that produce a metabolite homologous to IDO-1 , HIV-associated alterations in gut microbiota community composition are accompanied greater tryptophan catabolism. Several gut bacteria produce tryptophan catabolites that exert local effects and translocate to have systemic effects . Although there is some evidence that Lactobacillus could reduce IDO-1 activity in SIV-infected macaques , further research is needed to elucidate the mechanisms whereby changes in the gut microbiota alter tryptophan catabolism in PWH . Furthermore, the translocation of microbial products across the gut barrier upregulates IDO-1 via immune activation, which can amplify tryptophan catabolism and its neuropsychiatric consequences. Dendritic cells are essential for presenting antigens to other immune cells and regulating polarization of T cells into functional subsets. Dendritic cells expressing IDO-1 are thought to promote the development of regulatory T cells that suppress the immune response . In untreated HIV, upregulation of IDO-1 in dendritic cells is observed in the lymph nodes and gut lymphoid tissues, and there is concomitant depletion of Th 17 cells that are essential for maintaining gut barrier integrity . The mechanistic importance of tryptophan catabolism is supported by in vitro experimental findings that 3-hydroxyanthranilic acid (a catabolite of tryptophan) decreases the ratio of Th 17 to regulatory T cells in peripheral blood mononuclear cells . Because tryptophan catabolites cross the blood-brain barrier , there is also biological plausibility for observed associations of tryptophan depletion in plasma with cognitive impairment, depressed mood, anhedonia, and impulsivity in PWH . The role of the gut microbiota in influencing gut-immune dysregulation may also amplify neuropsychiatric comorbidities beyond substance use and depression, such as cognitive impairment. PWH show a decrease in bacterial production of neuroprotective short-chain fatty acids and other neuroactive metabolites, as well as concomitant increases in tryptophan catabolism, immune activation, and inflammation secondary to microbial translocation as noted previously . These mechanisms can potentiate the production of neurotoxic metabolites, increase permeability of the blood-brain barrier, and stimulate immune cells to transport HIV across the blood-brain barrier . In PWH, markers of microbial translocation (e.g., LPS), monocyte activation, changes in gut microbiota composition (e.g., alteration in the ratio of Firmicutes to Bacteroides ), and heightened IDO-1 activity have been associated with cognitive impairment and morphological brain changes . An extensive review of microbiome-gut-brain axis mechanisms relevant to cognitive impairment among PWH has been published elsewhere . Residual Immune Dysregulation in PWH Even with long-term effective ART, PWH experience persistent residual immune dysregulation, leading to faster onset and greater severity of age-related diseases . Specifically, elevated immune activation, inflammation, and coagulation in those receiving effective ART are associated with a spectrum of comorbidities that are not related to AIDS, including cardiovascular disease , cognitive impairment , metabolic disorders , and cancer . The clinical relevance of residual immune dysregulation is supported by the fact that HIV-associated non-AIDS conditions occur at younger ages with more severe health implications . Residual immune dysregulation could also have important consequences for the development and maintenance of neuropsychiatric disorders such as depression. Several PNI studies examining the neuropsychiatric consequences of residual immune dysregulation have focused on depression in PWH. Findings from the Multicenter AIDS Cohort study (1984–2010) indicate that elevations in soluble markers of immune activation and inflammation predict greater odds of screening positive for depression in men, and interestingly, this association was more pronounced among those without HIV . Specifically, serum C-reactive protein (CRP) levels greater than 3 mg/L were associated with 2.3 times greater odds of screening positive for depression. Another study observed associations between elevated CRP and depressive symptom severity in PWH , particularly among men with HIV where elevated CRP levels were associated with 3.7 times greater odds of moderate to severe depressive symptoms. Results from another large cohort study of PWH also indicated that greater depressive symptom severity was associated with an elevated proinflammatory profile consisting of elevated D-dimer, IL-6, and CRP, albeit in only in men . Although the inflammatory profile covaries with depression in PWH, many studies conducted to date around the globe have measured different cytokines, focused on depressive symptoms and not clinical diagnoses, enrolled modest sample sizes, and did not adjust for multiple comparisons . Finally, there is also evidence to support a potentially unique role of innate immune activation. Results from the Veterans Aging Cohort Study (2005–2006) indicate that somatic symptoms of depression were associated with elevated markers of monocyte activation . The beneficial associations of selective serotonin reuptake inhibitor use with lower monocyte activation and reduced inflammation in this cohort study underscore the importance of understanding the peripheral mechanisms whereby pharmacological treatment of depression could modify the microbiome-gut-brain axis crosstalk in PWH receiving effective ART, particularly among those who present with elevated inflammation (e.g., CRP >3.0 mg/L). There is increasing evidence that CMV reactivation is another important driver of residual immune dysregulation in the modern ART era . CMV infection activates a variety of innate and adaptive cell types including monocytes, macrophages, and CD4+, and CD8+ T cells with high levels of viral replication in select mucosal sites including the gut . In the modern ART era, immunologic measures of CMV and Epstein-Barr virus reactivation are independently linked to HIV-associated non-AIDS conditions such as myocardial infarction , and CMV-induced immunologic alterations are thought to potentiate accelerated aging in PWH . Among PWH receiving effective ART, latently infected T cells maintain the HIV reservoir via several mechanisms including T-cell proliferation . Impaired ability to control CMV infection could be a primary driver of cytokine and chemokine overflow that fuel immune activation and immunosenescence of T cells to maintain the HIV reservoir . PNI studies conducted in the modern ART era are needed to examine the bidirectional pathways linking asymptomatic CMV reactivation, residual immune dysregulation, and psychological factors. Elucidating the Effects of Social Adversity PWH experience social adversity stemming from developmental vulnerabilities as well as intersectional stigma and discrimination that could amplify the risk of neuropsychiatric comorbidities via alterations in the microbiome-gut-brain axis. Many PWH experience early life stress that serves as a risk factor for developmental trajectories that increase vulnerability to HIV and could create fundamentally different set points in neuroendocrine signaling . Developmental vulnerabilities as well as intersectional stigma and discrimination related to HIV and other minoritized identities (e.g., racial/ethnic minority, sexual minority, or gender minority) may have important health consequences . The field of social genomics provides opportunities to examine the epigenetic mechanisms whereby the HPA axis and the autonomic nervous system could influence residual immune dysregulation in the modern ART era. The conserved transcriptional response to adversity (CTRA) is a validated profile of leukocyte gene expression reflecting upregulated expression of genes involved in inflammation as well as downregulated expression of genes involved with type I interferon (a cytokine involved in antiviral and antitumor responses) and antibody synthesis . Psychosocial challenges such as experiences of discrimination have been associated with a higher CTRA, irrespective of HIV status . This observation is consistent with findings where homophobic victimization was associated with a higher CTRA among sexual minority men without HIV . Interestingly, methamphetamine use and intimate partner violence are associated with upregulation of the inflammatory genes as well as upregulation of the type I interferon and antibody synthesis genes among sexual minority men with HIV . The upregulation of type I interferon and antibody synthesis genes may be driven in part by active viremia in this study. These findings are further supported by cross-sectional studies documenting the associations of sexual minority stress and recent stimulant use with alterations in gene expression patterns relevant to immune activation and inflammation in methamphetamine-using sexual minority men with HIV who were virally suppressed . More comprehensive, longitudinal studies are needed to understand how social adversity and stimulant use alter the transcriptional regulation of leukocytes among PWH. The gut microbiota include trillions of bacteria, eukaryotes, and viruses . These microorganisms play a vital role in maintaining neuroimmune homeostasis and are associated with health outcomes, including HIV disease progression . Most research has focused on bacteria, which can influence neural, endocrine, and immune cells in the gut, periphery, and brain through the release of bacterial metabolites or via direct contact of bacteria (or bacterial products) with host cells . This can lead to chronic peripheral and neuroinflammation, dysregulated neurotransmitter synthesis, and altered afferent vagal nerve stimulation. In addition, gut bacteria can influence intestinal permeability, impact blood-brain-barrier integrity, modify nutrient absorption, alter the growth of pathogenic bacteria, and take part in drug metabolism, all of which are relevant to neuroimmune functioning among PWH . Many of these pathways whereby gut microbiota could potentiate depression, substance use disorders, and other neuropsychiatric comorbidities that compromise HIV disease management remain largely unexamined. PWH show differences in the composition and function of the gut microbiota compared with people without HIV . Microbiome alterations are more prominent among untreated PWH, but disruptions persist despite effective ART . The gut microbiome generally shows a pattern of decreased community diversity as well as enrichment of bacterial taxa linked to inflammation, microbial translocation, and poor CD4+ T-cell restoration. There is also evidence of depletion of bacterial taxa that are inversely associated with these markers . In addition to the role of HIV infection in altering the gut environment, the gut microbiome can be modified by a variety of behavioral and psychological factors, but studies in PWH are generally lacking or correlative. For example, stimulant use and alcohol use are associated with altered gut microbiota among PWH . There is also increasing interest in the mechanisms whereby microbiome-related alterations in reward processing influence the maintenance and severity of various neuropsychiatric disorders, including substance use disorders . Further research is needed to characterize the bidirectional microbiome-gut-brain axis mechanisms relevant to problematic patterns of alcohol or other substance use among PWH. The vagus nerve innervates the gut, providing bidirectional (80% afferent, 20% efferent) communication with the central nervous system. Vagal tone is essential for decreasing gut permeability so that microbial products like lipopolysaccharide (LPS) do not leak into the periphery to amplify residual immune dysregulation. A recent study of PWH who presented with gastrointestinal symptoms observed that vagal dysfunction was associated with bacterial outgrowth in the small intestines that was, in turn, linked to greater systemic inflammation . Efferent vagal pathways that are modulated by decreases in depressive symptoms have the potential to alter the gut microbiota and decrease microbial translocation, as evidenced by the effects of vagal stimulation . Prior clinical research in people without HIV observed that increased vagal tone is an important indicator of a favorable response to cognitive-behavioral treatments for depression . PWH may also experience autonomic neuropathy , which is linked to vagal dysfunction and microbiome-gut-brain axis alterations. Gut microbiota can also modulate mood and behaviors by altering stimulation of the vagal nerve. For example, gut bacteria release metabolites (or induce host cells to release metabolites) that stimulate the vagal nerve afferents that alter hypothalamic function to promote the release of oxytocin (a neuropeptide that influences social behavior, metabolism, and wound healing) in the brain. In fact, Lactobacillus reuteri have been shown to activate vagal afferents to potentiate oxytocin release from the paraventricular nucleus of the hypothalamus . This process promotes the coordinated activity of oxytocin and serotonin in the nucleus accumbens to increase the experience of social reward , which could have important implications for anhedonia, a common symptom of many neuropsychiatric disorders, including substance use disorders and depression . Although oxytocin has not been extensively examined among PWH, there is some evidence of nonlinear associations in PNI studies . Higher oxytocin levels have been associated with more severe depressive symptoms in PWH , potentially reflecting a compensatory response to depressive symptoms. In another cross-sectional study, the association of elevated stress with lower CD4+ T-cell count was observed only at higher levels of oxytocin among low-income ethnic minority women with HIV . There is a clear need for microbiome-gut-brain axis studies to understand the relevance of oxytocin in context of other important neuroimmune pathways among PWH. Damage to the gut during acute HIV is thought to be a primary driver of alterations in gut microbiota , expansion of the enteric virome , depletion of CD4+ T cells , and increased gut permeability . Most notably, there is a preferential depletion of T-helper 17 cells (Th 17 ) cells that are essential for maintaining gut barrier integrity , and this is currently irreversible unless ART is initiated early in acute HIV infection . Further research is needed to understand how key alterations to microbiome-gut-brain axis pathways could amplify the risk of persistent neuropsychiatric comorbidities during acute HIV infection and after suppressive ART. HIV-induced damage to the gut results in the translocation of microbial products such as LPS and peptidoglycan (PGN) across the gut barrier. When these microbial products cross the damaged gut barrier and enter the periphery, they chronically stimulate immune cells, which causes ongoing immune activation and inflammation in PWH . Despite the prominent focus on microbial translocation of LPS, PGN is also linked to inflammation in the gut and brain . Fragments of PGN are released into circulation as byproducts of bacterial cell remodeling and are typically recognized by pattern recognition receptors and other PGN-specific recognition proteins on host immune cells . Although the role of PGN in HIV has not been well established, PGN has been implicated in autoimmune diseases such as Crohn disease , rheumatoid arthritis , and multiple sclerosis . The potential implications for neuroinflammation are supported by observations of high concentrations of PGN in brain lesions in people with multiple sclerosis . Given its role in neuroimmune modulation, studies are needed to better understand the relevance of PGN for depression, substance use disorders, and other neuropsychiatric disorders among PWH . There is growing evidence for interactions between stimulant use and microbiome-gut-brain axis pathways via the gut-immune dysregulation. One such pathway is through the impact of stimulant-induced microbial translocation and gut barrier dysfunction. Stimulants such as methamphetamine can decrease parasympathetic tone , which could increase intestinal permeability. Stimulants also directly damage gut barrier integrity in self-administering HIV-1 transgenic rats and were associated with depleted tryptophan (an essential amino acid precursor for serotonin) in PWH receiving HAART . Tryptophan catabolites could damage gut barrier integrity through several mechanisms described hereinafter . Bearing in mind that there are currently no treatments for stimulant use disorders approved by the US Food and Drug Administration, mechanistic studies characterizing these complex microbiome-gut-brain axis pathways could identify novel targets for pharmacologic interventions in PWH who have stimulant use disorders that serve as a key obstacle to sustained viral suppression. Another mechanism whereby the gut microbiota can alter immune surveillance is inducing the enzyme indoleamine 2,3-dioxygenase 1 (IDO-1) to catabolize tryptophan . Tryptophan catabolites influence immune function in the periphery and are neuroactive. PWH display increases in tryptophan catabolism that are linked to depressive symptoms and only partially normalized by ART . Tryptophan catabolism may also be exacerbated among those with protein-deficient diets, highlighting the importance of food insecurity in the intersection of biological and structural determinants relevant to depression among PWH . Examining the role of tryptophan catabolism in the microbiome-gut-brain axis in the modern ART era could help address the underlying HIV-associated biological alterations that modify depression and suicide risk to optimize mental health treatment for PWH. Because PWH have more gut bacterial taxa that produce a metabolite homologous to IDO-1 , HIV-associated alterations in gut microbiota community composition are accompanied greater tryptophan catabolism. Several gut bacteria produce tryptophan catabolites that exert local effects and translocate to have systemic effects . Although there is some evidence that Lactobacillus could reduce IDO-1 activity in SIV-infected macaques , further research is needed to elucidate the mechanisms whereby changes in the gut microbiota alter tryptophan catabolism in PWH . Furthermore, the translocation of microbial products across the gut barrier upregulates IDO-1 via immune activation, which can amplify tryptophan catabolism and its neuropsychiatric consequences. Dendritic cells are essential for presenting antigens to other immune cells and regulating polarization of T cells into functional subsets. Dendritic cells expressing IDO-1 are thought to promote the development of regulatory T cells that suppress the immune response . In untreated HIV, upregulation of IDO-1 in dendritic cells is observed in the lymph nodes and gut lymphoid tissues, and there is concomitant depletion of Th 17 cells that are essential for maintaining gut barrier integrity . The mechanistic importance of tryptophan catabolism is supported by in vitro experimental findings that 3-hydroxyanthranilic acid (a catabolite of tryptophan) decreases the ratio of Th 17 to regulatory T cells in peripheral blood mononuclear cells . Because tryptophan catabolites cross the blood-brain barrier , there is also biological plausibility for observed associations of tryptophan depletion in plasma with cognitive impairment, depressed mood, anhedonia, and impulsivity in PWH . The role of the gut microbiota in influencing gut-immune dysregulation may also amplify neuropsychiatric comorbidities beyond substance use and depression, such as cognitive impairment. PWH show a decrease in bacterial production of neuroprotective short-chain fatty acids and other neuroactive metabolites, as well as concomitant increases in tryptophan catabolism, immune activation, and inflammation secondary to microbial translocation as noted previously . These mechanisms can potentiate the production of neurotoxic metabolites, increase permeability of the blood-brain barrier, and stimulate immune cells to transport HIV across the blood-brain barrier . In PWH, markers of microbial translocation (e.g., LPS), monocyte activation, changes in gut microbiota composition (e.g., alteration in the ratio of Firmicutes to Bacteroides ), and heightened IDO-1 activity have been associated with cognitive impairment and morphological brain changes . An extensive review of microbiome-gut-brain axis mechanisms relevant to cognitive impairment among PWH has been published elsewhere . Even with long-term effective ART, PWH experience persistent residual immune dysregulation, leading to faster onset and greater severity of age-related diseases . Specifically, elevated immune activation, inflammation, and coagulation in those receiving effective ART are associated with a spectrum of comorbidities that are not related to AIDS, including cardiovascular disease , cognitive impairment , metabolic disorders , and cancer . The clinical relevance of residual immune dysregulation is supported by the fact that HIV-associated non-AIDS conditions occur at younger ages with more severe health implications . Residual immune dysregulation could also have important consequences for the development and maintenance of neuropsychiatric disorders such as depression. Several PNI studies examining the neuropsychiatric consequences of residual immune dysregulation have focused on depression in PWH. Findings from the Multicenter AIDS Cohort study (1984–2010) indicate that elevations in soluble markers of immune activation and inflammation predict greater odds of screening positive for depression in men, and interestingly, this association was more pronounced among those without HIV . Specifically, serum C-reactive protein (CRP) levels greater than 3 mg/L were associated with 2.3 times greater odds of screening positive for depression. Another study observed associations between elevated CRP and depressive symptom severity in PWH , particularly among men with HIV where elevated CRP levels were associated with 3.7 times greater odds of moderate to severe depressive symptoms. Results from another large cohort study of PWH also indicated that greater depressive symptom severity was associated with an elevated proinflammatory profile consisting of elevated D-dimer, IL-6, and CRP, albeit in only in men . Although the inflammatory profile covaries with depression in PWH, many studies conducted to date around the globe have measured different cytokines, focused on depressive symptoms and not clinical diagnoses, enrolled modest sample sizes, and did not adjust for multiple comparisons . Finally, there is also evidence to support a potentially unique role of innate immune activation. Results from the Veterans Aging Cohort Study (2005–2006) indicate that somatic symptoms of depression were associated with elevated markers of monocyte activation . The beneficial associations of selective serotonin reuptake inhibitor use with lower monocyte activation and reduced inflammation in this cohort study underscore the importance of understanding the peripheral mechanisms whereby pharmacological treatment of depression could modify the microbiome-gut-brain axis crosstalk in PWH receiving effective ART, particularly among those who present with elevated inflammation (e.g., CRP >3.0 mg/L). There is increasing evidence that CMV reactivation is another important driver of residual immune dysregulation in the modern ART era . CMV infection activates a variety of innate and adaptive cell types including monocytes, macrophages, and CD4+, and CD8+ T cells with high levels of viral replication in select mucosal sites including the gut . In the modern ART era, immunologic measures of CMV and Epstein-Barr virus reactivation are independently linked to HIV-associated non-AIDS conditions such as myocardial infarction , and CMV-induced immunologic alterations are thought to potentiate accelerated aging in PWH . Among PWH receiving effective ART, latently infected T cells maintain the HIV reservoir via several mechanisms including T-cell proliferation . Impaired ability to control CMV infection could be a primary driver of cytokine and chemokine overflow that fuel immune activation and immunosenescence of T cells to maintain the HIV reservoir . PNI studies conducted in the modern ART era are needed to examine the bidirectional pathways linking asymptomatic CMV reactivation, residual immune dysregulation, and psychological factors. PWH experience social adversity stemming from developmental vulnerabilities as well as intersectional stigma and discrimination that could amplify the risk of neuropsychiatric comorbidities via alterations in the microbiome-gut-brain axis. Many PWH experience early life stress that serves as a risk factor for developmental trajectories that increase vulnerability to HIV and could create fundamentally different set points in neuroendocrine signaling . Developmental vulnerabilities as well as intersectional stigma and discrimination related to HIV and other minoritized identities (e.g., racial/ethnic minority, sexual minority, or gender minority) may have important health consequences . The field of social genomics provides opportunities to examine the epigenetic mechanisms whereby the HPA axis and the autonomic nervous system could influence residual immune dysregulation in the modern ART era. The conserved transcriptional response to adversity (CTRA) is a validated profile of leukocyte gene expression reflecting upregulated expression of genes involved in inflammation as well as downregulated expression of genes involved with type I interferon (a cytokine involved in antiviral and antitumor responses) and antibody synthesis . Psychosocial challenges such as experiences of discrimination have been associated with a higher CTRA, irrespective of HIV status . This observation is consistent with findings where homophobic victimization was associated with a higher CTRA among sexual minority men without HIV . Interestingly, methamphetamine use and intimate partner violence are associated with upregulation of the inflammatory genes as well as upregulation of the type I interferon and antibody synthesis genes among sexual minority men with HIV . The upregulation of type I interferon and antibody synthesis genes may be driven in part by active viremia in this study. These findings are further supported by cross-sectional studies documenting the associations of sexual minority stress and recent stimulant use with alterations in gene expression patterns relevant to immune activation and inflammation in methamphetamine-using sexual minority men with HIV who were virally suppressed . More comprehensive, longitudinal studies are needed to understand how social adversity and stimulant use alter the transcriptional regulation of leukocytes among PWH. Seminal PNI studies provided support for the role of neuroendocrine signaling in altered immune function among PWH. In the modern ART era, PNI studies should expand our understanding of the relevance of the complex, bidirectional interactions of catecholamines, cortisol, and oxytocin with gut microbiota. These neuroendocrine signaling mechanisms for the biologically plausible effects of depression and other neuropsychiatric comorbidities on the microbiome-gut-brain axis in PWH are described briefly hereinafter. Experimental studies are needed to understand whether and how interventions to reduce depression and substance use influence the microbiome-gut-brain axis via alterations in these distinct neuroendocrine signaling pathways. There is evidence for the impact of neuroendocrine signaling on the gut and inflammation. Catecholamines directly affect the growth and virulence of several species of anaerobes, such as Enterobacteriaceae , in the gut . More recently, this group demonstrated that higher adrenergic baroreceptor sensitivity, an index of a more hyperadrenergic state, was associated with greater inflammation in PWH . Another study with 4000 PWH from the Strategies for Management of Antiretroviral Therapy study revealed cross-sectional inverse associations between heart rate variability with soluble markers of inflammation and coagulation . Bearing in mind that stimulants act primarily by altering the function of catecholamines, stimulants and HIV are independently associated with higher NE in cerebrospinal fluid and dysregulated systemic metabolism of tyrosine (an essential amino acid precursor for catecholamines) . Alterations in catecholamines could partially explain associations of stimulant use with greater immune activation, inflammation, HIV proviral DNA in immune cells, and immune exhaustion in PWH who are virally suppressed or undetectable . Interestingly, the associations of stimulant use with these outcomes often persist after adjusting for self-reported ART adherence . A group-based cognitive-behavioral intervention decreased urinary NE output in sexual minority men with HIV, which is consistent with the notion that psychological interventions can influence the autonomic nervous system in PWH . Further research is needed to determine if there are distinct effects of vagal tone versus systemic release catecholamines from the HPA axis on the microbiome-gut-brain axis in PWH. Although changes in HPA axis functioning resulting in elevated cortisol and catecholamines are well characterized, further research is needed with PWH to understand the mechanistic role of the gut microbiota in influencing the neuroendocrine stress responses. There is some evidence that gut microbiota metabolize glucocorticoids with relevance for hypertension and androgen production . There are also dynamic, bidirectional interactions where glucocorticoids are thought to influence the composition of gut microbiota , which have been shown to modify glucocorticoid receptor expression in mice . Translational research with humans is clearly needed to understand the role of the HPA axis in modifying microbiome-gut-brain axis responses, particularly among PWH where adrenal insufficiency secondary to infections such as a CMV is more common . There is also increasing evidence that oxytocin release in the brain can modify microbiome-gut-brain axis function . For example, one recent experimental study with rats observed that intracisternal injection of oxytocin decreased microbial translocation because of LPS in a dose-dependent manner. Interestingly, these effects of oxytocin on decreased microbial translocation were blocked by vagotomy, which underscores the central importance of efferent vagal communication with the gut . There is also experimental evidence in rats that oxytocin could mitigate the effects of stress on gastric emptying and motility . Finally, findings from human studies indicate that oxytocin may play an important role in the enteric nervous system of the gut. This is partially supported by findings from biopsies of the gastrointestinal tract where oxytocin was expressed in the myenteric and submucous ganglia and nerve fibers . Further research in PWH is needed to determine whether and how increases in oxytocin could modify the microbiome-gut-brain axis. Modern PNI studies are needed to determine how several, interrelated HIV-associated pathophysiologic alterations could amplify the risk of neuropsychiatric comorbidities among PWH receiving effective ART. Drawing upon our conceptual model (Figure ), delineating the mechanisms whereby alterations in gut microbiota influence neural signaling (e.g., the vagus nerve), alter tryptophan catabolism and other neuroactive metabolites, and indirectly impact immune dysregulation in treated HIV is essential to catalyze the development of tailored pharmacologic treatments for depression, substance use disorders, and other neuropsychiatric disorders among PWH. On the other hand, neuropsychiatric disorders such as depression could also potentiate alterations in the microbiome-gut-brain axis via the HPA axis, autonomic nervous system, and oxytocin release. PNI studies are needed to examine how alterations in these neuroendocrine signaling pathways could modify microbiome-gut-brain axis pathways relevant to neuropsychiatric disorders and health outcomes in PWH. More robust efforts are needed to embrace experimental rigor by testing the mechanisms whereby biomedical (e.g., vagal stimulation, probiotics) or behavioral interventions (e.g., cognitive-behavioral therapy) modify microbiome-gut-brain axis pathways relevant to neuropsychiatric disorders. Other experimental, laboratory-based paradigms such as reactivity studies using either pharmacologic (e.g., hydrocortisone, intranasal oxytocin), psychosocial (e.g., Trier Social Stress Test), or physical (e.g., cold pressor task) probes could also yield important mechanistic insights into acute changes in key biological processes relevant to the microbiome-gut-brain axis. Bearing in mind that the experience of social evaluative threat is more closely linked to the HPA axis response , further research is needed to advance our understanding of the biobehavioral implications of social adversity across the life course for the microbiome-gut-brain axis among PWH. Many PWH experience prominent structural vulnerabilities such as housing instability and food insecurity that should receive greater attention in biobehavioral research . These structural vulnerabilities are often embedded within distinct residential environments, and a burgeoning literature has focused on demonstrating the associations of neighborhood-level factors with health outcomes such as viral suppression . Advancing our understanding of multilevel determinants of microbiome-gut-brain axis alterations will guide the development of more comprehensive intervention approaches to address prevalent neuropsychiatric comorbidities and optimize health outcomes among PWH. There is also a clear need for further research examining the relevance of sex and gender in microbiome-gut-brain axis crosstalk. Much of the early PNI research in PWH focused on sexual minority men and the extent to which this will generalize to cisgender women is unknown. There are divergent associations of inflammation with depression as a function of sex at birth across studies. Most notably, studies with cisgender women are needed to understand the distinct mechanisms that may diminish the associations of inflammation with depressive symptoms and poorer quality of life. There is also a clear need for studies with gender minority populations to understand whether and how gender-affirming biobehavioral treatments influence the microbiome-gut-brain axis. In summary, the modern ART era provides several opportunities to build a robust portfolio of research examining the role of microbiome-gut-brain axis in neuropsychiatric comorbidities. The central importance of HIV-induced damage to the gut and alterations in gut microbiota have enduring consequences for the mental and physical health of PWH. Interrogating these microbiome-gut-brain axis mechanisms could assist with identifying novel targets for biobehavioral interventions to address prevalent neuropsychiatric comorbidities among PWH such as depressive disorders and stimulant use disorders. This is essential to alleviate human suffering related to neuropsychiatric disorders and optimize the ability of many PWH to derive maximum benefits from ART. In Lewis Carol’s Through the Looking-Glass , Alice steps once again into a fantastical world, much like the possibilities of PNI and microbiome-gut-brain axis research in the modern ART era. |
Enhanced Osteoporosis Detection Using Artificial Intelligence: A Deep Learning Approach to Panoramic Radiographs with an Emphasis on the Mental Foramen | 06dda300-c1ca-490c-90e6-3c429e9ab641 | 11417815 | Dentistry[mh] | Osteoporosis, a systemic skeletal disorder marked by reduced bone mass, is anticipated to affect 60% of women over the age of 50 years by 2040 . The current gold standard for osteoporosis detection is a dual-energy X-ray absorptiometry (DXA) scan, which is only available in specialized osteoporosis centers and typically employed post-fracture . In addition to DXA measurements, the diagnosis of osteoporosis takes a history of fragility fractures and clinical risk factors such as age, gender, and family history into account . Given that osteoporosis-related fractures impose a financial burden on the health-care system amounting to double-digit billions of dollars worldwide, there is a critical demand for tools that are capable of early detection . However, age-related changes in bone density necessitate a tailored interpretation of DXA results . As individuals age, their bone density naturally decreases because of factors such as hormonal changes and decreased physical activity. DXA scans provide valuable insights into the longitudinal alterations in bone density, thereby enabling proactive interventions aimed at attenuating bone demineralization and diminishing the likelihood of fractures in aging cohorts . Consequently, the interpretation of cross-sectional investigations and individual longitudinal diagnoses requires meticulous consideration of each case’s unique circumstances. Generalizations made within mathematical models are subject to clinical controversy and warrant thorough discussion . A few studies have been published on the alternative detection option for osteoporosis, including magnetic resonance imaging (MRI), optical coherence tomography (oCT), and quantitative computed tomography (qCT), which appear to be viable and may offer cost and time savings . Nevertheless, given the significant impact of osteoporosis-related fractures on the health-care system caused by late detection, there is an urgent need for tools that enable early detection . The solutions discussed previously do not meet the requirements for effective early detection screening, underscoring a significant medical need for the development of such technologies. Panoramic radiography (PR) is a routine component of dental evaluations and gives a 2D overview of the dentition, the jaw, and the hard surrounding tissue. This may be conducted annually and presents a viable medium for such early detection efforts . Numerous studies have applied PR for osteoporosis detection, supporting its potential as an effective screening tool . In this context, specific measures that are derivable from panoramic radiographs (PRs) have been developed to assist in osteoporosis identification. These measures include three primary indices: the Mandibular Cortical Index (MCI), which assesses the shape of the mandibular cortex; the mandibular cortical width (MCW), which evaluates the ratio of the mandibular cortical width to the cortical height; and the Panoramic Mandibular Index (PMI), which is calculated as the ratio of the thickness of the mandibular cortex to the distance between the mental foramen and the inferior mandibular cortex . Notably, the MCI, also referred to as the Klemetti Index (KI), was introduced by Klemetti in 1993 and has since become the most frequently utilized indices for this purpose . Nevertheless, these indices are rarely used in the clinical workflow, since they must be manually calculated, are prone to errors, and are time-consuming. Comparing it to a gold standard technique such as DXA has proven it to be unreliable . Various studies of osteoporosis detection in PRs using a deep convolutional neural network (CNN) have been published recently, with promising outcomes . However, upon closer examination of the methodology and design of these studies, factors such as age, overlying structures, gender, disease severity, and radiographic technique may have influenced the radiographs’ ability to accurately detect osteoporosis-related changes, consequently affecting the algorithms’ performance . These variables were not adequately addressed in those studies, highlighting significant flaws in the application of PRs for osteoporosis detection and emphasizing the necessity for cautious interpretation. Furthermore, the methods of recent studies were not transparent. The aim of this study is to rectify these shortcomings of past studies and develop a reliable artificial intelligence (AI) application for the accurate detection of osteoporosis in PRs.
This study was conducted according to the guidelines of the Medical World Association (Declaration of Helsinki). Ethical approval was granted by the Institutional Review Board of the Charité ethics committee (EA2/089/22). The checklist for AI research in dentistry of the ITU/WHO focus group “Artificial Intelligence for Health (AI4H)” was consulted for the reporting in this study . Data This study retrospectively included patients who were diagnosed with osteoporosis at Danube Private University. Each patient underwent a DXA scan to assess their bone mineral density (BMD). Patients with a BMD T-score above −1 were classified as healthy, while those with a BMD T-score below −2.5 were diagnosed with osteoporosis (group A). PRs were retrieved from the medical records of each patient within 24 h following the DXA scan. After the radiographs of the osteoporosis patient group were collected, a second group of patients, related to the first group’s age and gender, was used as the control group (group B). Another healthy group of a non-correlating age (young patients) and gender distribution was also collected for validation purposes. Data Preparation and Model Training A total of 250 PRs with osteoporosis and a control group of 250 PRs without osteoporosis were automatically cropped to the region of interest of the mental foramen after the annotation of 60 images in the CvatAI annotator ( https://www.cvat.ai/ ) using the YoloV8 code. The region of interest, ROI, of the mental foramen was chosen based on the KI forecasting the most promising results . Subsequently, all images were visually inspected by an experienced investigator to identify any missing or improperly cropped images in the designated area. In total, a dataset of 500 grayscale images (comprising left and right views of the foramen and its surrounding area) was prepared meticulously for each group. The mental foramen was selected as the region of interest based on the aforementioned indices, which incorporate this anatomical structure. The cropped PRs with osteoporosis were randomly divided into three splits (70% train, 15% test, 15% validation). The control PRs without osteoporosis were sampled in accordance with the gender and age distribution of the training set (age = 54 ± 16 a, female/male ratio = 3/1). The validation split was used to select an optimal model performance during training and hyperparameter selection, while the held-out test split was used to evaluate the model’s performance after training and hyperparameter selection. For data preparation, the cropped PR images were augmented using random rotation with 10° and color-jitter (brightness = 0.2, hue = 0.1, contrast = 0.3, saturation = 0.3) and cropped and vertically flipped at random. Afterward, the data were scaled to 224 × 224 and normalized. The model optimization employed the SGD optimizer with an initial learning rate of 0.001 and a batch size of 16. To mitigate overfitting, an L2 lambda regularization strength of 1 × 10 −6 was applied. The model architecture was based on the pretrained Densenet201 network, which was adapted for binary classification output. After adjusting hyperparameters, the model underwent training for 150 epochs. Implementation was carried out using the PyTorch code, and training was performed on a single V100 GPU (NVIDIA). Furthermore, 500 cropped PRs of young patients under the age of 30 years without osteoporosis were used for the same model, performed with the same hyperparameter (age = 24 ± 6 a, female/male ratio = 3/1); see . Statistical Analysis The accuracy of osteoporosis detection in the foramen regions involved the processing of the cropped region using its classifier. Based on these predictions, performance metrics were computed, including precision on the validation set, as well as a confusion matrix for the comparison of the osteoporosis group versus the control group of the same age and the comparison of the osteoporosis group versus the young control group.
The detection accuracy and area under the curve (AUC) were calculated for the datasets. In and , the confusion matrix illustrates the detection results of osteoporosis on PR. The model achieved a precision of 73.6% and an F1 score of 0.74 for the validation accuracy in the group comparing osteoporosis patients with the age-matched control group . The AUC was 0.84. In the group comparing osteoporosis patients with the young control cohort, the model demonstrated an accuracy of 97.8% and an AUC of 0.98, with an F1 score of 0.97 . The correlating confusion matrices are shown in and .
This study demonstrated the capability of a deep learning algorithm to detect osteoporosis indicators in dental radiographs. It demonstrated the model’s effective performance in distinguishing osteoporosis patients from control groups of varying ages. The findings underscore its robustness in diagnostic accuracy and discrimination ability, emphasizing its potential utility in clinical practice for diverse demographic profiles. Within these models, significant differences in AI-assisted osteoporosis diagnosis outcomes emerged, primarily stemming from discrepancies in data selection and annotation methodologies . For instance, Sukegawa et al. , utilizing a meticulously curated dataset with comprehensive annotations, achieved commendable accuracy rates in identifying osteoporotic markers from dental radiographs. Conversely, Lee et al. , using inadequately annotated data with limited diversity, produced inferior results, highlighting the significant impact of data quality on AI performance. In this case, the inclusion of young patients in the dataset falsely improves the algorithm’s performance, as our study demonstrated through our calculations. Comparing the model to 224 images of young patients, the algorithm performed with a 97.81% accuracy. The performance of AI systems is significantly affected not only by data acquisition but also by factors such as the network architecture and hyperparameters during training and validation. This complexity poses challenges in assessing the reliability of reported results. In the field of automated data detection, the importance of data selection and annotation cannot be overstated. They form the basis upon which AI models are constructed, directly influencing their performance and outcomes. Numerous studies have emphasized the pivotal role of data quality in AI applications . The choice of data, their relevance, and the accuracy of annotation are pivotal factors that influence the efficacy of AI algorithms. In medical imaging, where precision and reliability are paramount, the impact of data selection and annotation is particularly pronounced. This investigation also reveals several limitations. The applications of PRs are not considered the gold standard for osteoporosis detection. The small sample size increases the risk of overfitting, potentially affecting the model’s performance when applied to new datasets. The use of only one type of radiograph further limits the model’s generalizability, as it may not perform as well with other imaging modalities. Additionally, the dataset is derived from a single institution, which may introduce biases that are specific to that population and reduce the external validity of the findings. As a result, the model’s applicability to broader clinical settings and diverse patient populations remains uncertain. Future studies should aim to validate these findings using larger, multi-institutional datasets and a variety of radiographic types to improve generalizability and robustness. Enhancing the data volume through multi-center collaborative efforts could elevate the accuracy and generalizability of the model’s diagnostic classification. Another limitation concerns the specific types of models evaluated. This research assessed the performance of YOLOv8 at various depths. Discovering an architecture with fewer parameters that maintains or improves performance could enhance its applicability by reducing computational costs. YOLOv8 is an object detection model that has not been specifically developed on medical images. Future research should focus on identifying an architecture that is optimal for different image qualities and patient demographic variables. Lastly, the manual cropping of images to position the mental foramen within the mandible based on the common indices introduces a potential bias in preoperative preparation. This underscores the need for refined image preparation techniques in future studies. In the future, it will be essential to evaluate this model alongside medical professionals to assess whether their diagnostic accuracy improves when they utilize regions that are highlighted by deep learning techniques. Conducting such comparisons will aid in enhancing the development and application of deep learning methodologies. Additional studies are required to ascertain the specific areas within a PR that the model identifies as significant, considering that the image encompasses various anatomical structures. Moreover, it is crucial to investigate the model’s adaptability and performance in terms of generalizability using datasets from different institutions.
This research showcased a proof-of-concept algorithm that highlights the potential of deep learning in identifying osteoporosis indicators in dental radiographs. Furthermore, our thorough examination of existing algorithms revealed that not all optimistic outcomes hold credibility under scrutiny of methodological integrity. Despite promising results, several limitations must be considered. The reliance on a small, single-institution dataset increases the risk of overfitting and limits generalizability. Furthermore, the exclusive use of PRs, which are not the gold standard for osteoporosis detection, and the need for manual image preparation introduce potential biases. Future research should focus on validating these findings with larger, multi-institutional datasets and exploring models that are better suited for medical imaging. Additionally, integrating deep learning techniques into clinical practice will require further evaluation alongside expert radiologists to enhance the diagnostic accuracy and practical applicability.
|
Expert consensus on orthodontic treatment of protrusive facial deformities | 7a67f421-38c8-4ddd-a782-51ea3ef6b62c | 11785726 | Dentistry[mh] | The protrusive facial deformity typically refers to the malocclusion where the upper and lower lips are protruded relative to the facial profile. – It includes skeletal Class I protrusion, characterized by the protruding upper and lower incisors, with or without forward-positioned jaws, and a mostly neutral molar relationship. , The more common facial deformity in China is skeletal Class II protrusion, which refers to a malocclusion characterized by a discrepancy in the three-dimensional relationship between the upper and lower jaws, accompanied by dental compensation, with distal or neutral molar relationships. – Among various misalignments in the sagittal, transverse, and vertical dimensions, the thickness of soft tissues can also impact the morphology of hard tissues. Compensatory interactions among the perioral muscles, teeth, and jaws may present completely different soft tissue profiles. Failure to consider the comprehensive coordination of teeth, jaws, and soft tissues, as well as the complicated mechanisms of malocclusion, during the formulation of orthodontic plans can often lead to incorrect assessments by orthodontists regarding treatment goals, difficulty levels, and outcomes. Therefore, a multidimensional analysis and judgment of the etiologic mechanisms of protrusive facial deformity are necessary to develop correct treatment plans. , The formulation of orthodontic treatment plans normally involves consideration of multiple dimensions and comprehensive elements. Multiple dimensions refer to the traditional three-dimensional concept, including the sagittal, vertical, and transverse dimensions and maturity of growth and development. Multi-elements include teeth alignment, jaw relationship, facial contour, periodontal condition, upper airway patency, temporomandibular joints, and perioral muscle balance. – The treatment plan, based on multidimensional analysis, differs from traditional orthodontic approaches that primarily focus on diagnosis and treatment concepts limited to teeth, jaws, and facial profiles. This consensus underscores the utilization of various orthodontic strategies, such as mandible advancement, retraction of anterior teeth, and maxillofacial vertical control. , These strategies aim to reduce anterior teeth and lip protrusion, increase chin prominence, harmonize nasolabial and chin-lip relationships, and improve the facial profile of patients with protrusive facial deformities. In cases of severe skeletal protrusive facial deformities, a combination of orthodontic and orthognathic treatments may be recommended. – Protrusive facial deformities are characterized by the forward position of the lips relative to the facial profile, assessed through the position of three critical anatomical landmarks, including the glabella, subnasale, and pogonion. Diagnosed by focusing on soft tissue morphology, these deformities include a variety of complex maxillofacial abnormalities with compromised soft-tissue contours, closely related to the upper airway, temporomandibular joint, and perioral muscle balance. The “functional matrix theory” of growth posits that facial growth is a response to functional needs and neurotrophic influences. , Effective lip competence and nasal breathing are essential for the synchronized development of maxillofacial elements, as normal respiratory airflow during these activities stimulates the development of relevant anatomical areas, thereby enhancing craniofacial structures. Protrusive facial deformities are often accompanied by decreased muscle tone, which affects the morphology of the underlying bone structures and prompts compensatory adjustment among the lips, teeth, and jaws, ultimately resulting in an imbalanced soft tissue contour. These deformities can also be caused by obstructive airway conditions exacerbating structural defects manifested by narrowed dental arches, elevated palates, hypoplasia of the face and nose, and excessive divergent skeletal Class II deformities known as “adenoid facies”. , The morphological mechanisms and clinical manifestations of protrusive facial deformities are diverse and complex. The simple form typically presents proclination of the upper and lower incisors, and a Class I molar relationship, which might be accompanied by maxillary and mandibular protrusion. Severe protrusive facial deformities may be accompanied by sagittal discrepancy between the jaws, excessive vertical growth, and abnormal molar relationships, which may result in dentofacial dysfunction. The complex etiology and pathogenesis of these deformities are diverse, resulting in a wide range of craniofacial morphologies, that need to be classified according to specific skeletal sagittal and vertical determinants (Fig. ). Epidemiological data reveal a high prevalence of protrusive facial deformities in China, typically characterized by normal maxilla and insufficient mandible. Given the multifactorial etiology and complex clinical manifestation of these deformities, a holistic assessment is essential to establish an accurate diagnosis and treatment plan, which should be followed by the multidimensional and total-element analysis strategies, to achieve the esthetic, functional and stable goals of orthodontic treatment. Given the intricate etiology and varied clinical manifestations, refining the diagnostic strategy for protrusive facial deformities characterized by protrusion is crucial for improving both orthodontic outcomes and patient satisfaction. Typically, orthodontic strategies include multidimensional assessments and total-element considerations. Multidimensional analysis refers to the comprehensive evaluation of sagittal, vertical and horizontal dimensions as well as growth potential in diagnosing and decision-making process of orthodontic cases. Complex protrusive facial deformities often involve abnormalities in multiple dimensions. Patients with hyperdivergent malocclusion exhibit excessive vertical facial growth and rotation of the mandible, resulting in an increased mandibular plane angle. The lower lip appears more prominent relative to the chin due to the clockwise rotation of the mandible in these patients, indicating that excessive retraction of incisors is not acceptable, resulting in excessive flattening of the lips. A vertical control strategy may be used to reduce the vertical height and alleviate the sagittal discrepancy. Due to favorable chin, hypodivergent skeletal Class II patients can obtain good profile by retraction of incisors to reduce lip prominence. Therefore, the extraction plan should be carefully selected to avoid excessive retraction of the incisors to form a concave profile. The normal width of the upper arch is the key to guide the sagittal growth of mandible. In children and adolescents, the narrow upper arch will limit the forward growth of the mandible, which causes the mandible to maintain the retrusive position, resulting in skeletal Class II malocclusions. The transverse discrepancy is closely related to facial esthetics. For patients with normal to large facial width, over-intruding the posterior teeth and counterclockwise rotation of the mandible are not recommended to avoid the deterioration of the facial width-to-length ratio. Based on the three-dimensional analysis in sagittal, vertical and horizontal dimensions, we emphasize the fourth dimension-growth and development. As a classic concept in orthodontics, growth and development is limited to children and adolescents by many orthodontists, which is not comprehensive. The growth and development we emphasize refer to the changes of dentofacial hard and soft tissues throughout the life cycle and the decisive role of genetic factors on growth patterns. The etiology of malocclusion is multifaceted, including teeth alignment, jaw relationship, facial contour, periodontal condition, upper airway patency, temporomandibular joints, and muscle balance. Failure to consider the clinical manifestations and the malocclusion formation mechanism may lead to incorrect prediction of the goal, difficulty index, strategy, and efficacy of orthodontic treatment. Therefore, the optimal treatment plan should be formulated based on the total-element diagnosis and comprehensive analysis (Fig. ). Tooth alignment and occlusion Tooth alignment involves a variety of elements, including mesiodistal angulation, labiolingual angulation, rotation, crowding, labiolingual malposition, and arch form. The goal of orthodontic treatment is to achieve a combination of the above elements with good occlusal function, and the roots should be centered within the alveolar bone to maintain periodontal health. In patients treated with camouflage orthodontics, the root in the alveolar bone can be moderately displaced from the center and exhibit compensatory inclination within a safe range, achieving ideal tooth alignment and occlusion. In addition to the above elements, individual patients’ soft tissue characteristics should also be considered. For different vertical skeletal patterns and different face shapes, the arrangement of teeth should be considered individually. The dental arch should match the face shape. The appropriate width of the dental arch contributes to a coordinated buccal corridor. Otherwise, it can negatively impact smile esthetics. Skeletal relationship The skeletal relationship includes sagittal and vertical skeletal patterns. In addition to considering the relative position of the maxilla and mandible, the absolute sagittal jaw positions relative to the cranial base should also be considered in the diagnosis of a patient’s sagittal skeletal pattern, which is crucial for determining the target positions in orthodontic treatment planning. The impact of the vertical skeletal pattern on the sagittal pattern should also be considered in diagnosis and treatment plan design. Hyperdivergent skeletal pattern aggravates skeletal Class II malocclusion, while hypodivergent skeletal pattern aggravates skeletal Class III malocclusion. The jaw positional relationship is an important part of the dentofacial complex system, which should be integrated with soft tissue analysis to formulate the most appropriate treatment plans for patients presenting with protrusive facial deformities. Frontal and lateral profile Both the goals and limitations of modern orthodontic and orthognathic treatment are determined not only by the teeth and bones but also by the soft tissues of the face. The “soft tissue paradigm” advocated by modern orthodontics is to calculate the target position of incisors according to the soft tissue esthetics, so as to determine the details of treatment design such as the required space, the pattern of extraction, and the anchorage design. When making treatment plans, orthodontists should pay more attention to the patient’s frontal and lateral profile, and strive to make the treatment plan consistent with or close to the patient’s subjective esthetic expectations. Attention should also be paid to the adverse changes in profile caused by orthodontic treatment, particularly in patients with high zygomatic bone, sunken cheeks, etc. , Other elements The upper airway is closely related to health and life, and it is one of the key elements to be considered in the diagnosis and treatment of malocclusion. Following the retraction of incisors, the velopharyngeal, glossopharyngeal, and hypopharyngeal airway may become narrower. The target position of incisors should not only meet the needs of facial esthetics but also take into account the effect of incisor retraction on the size of the upper airway. For patients with upper airway stenosis, the amount of incisor retraction should be strictly controlled to maintain the inherent oral space and normal nasal respiratory function. If necessary, bimaxillary advancement surgery should be combined to correct the protrusion and increase the airway volume. Oral habits, such as sucking habit, abnormal tongue position, and tongue thrust swallowing, can break the balance of the internal and external strength of the jaw and arch, leading to malocclusion. Within the stomatognathic system, muscles often play a dominant role over bones. For patients exhibiting oral habits, orthodontists should ensure that changes in teeth, arches, and jaws are coordinated with muscle function during treatment. Only by removing oral habits and achieving normal perioral muscle function can ensure the long-term stability of orthodontic treatment. Temporomandibular disorders (TMD) are primarily characterized by joint pain, joint noise, and mandibular movement disorders. It is necessary to recognize the complexity of the etiology and pathophysiological mechanism of TMD and its impact on the stability of mandibular position and occlusion. , A high prevalence of TMD in skeletal class II patients referred for orthognathic surgery, especially in those with a pronounced overjet and high mandibular plane angle. A solid cusp-fossa relationship of the teeth should be established during orthodontic treatment, which is an important factor for the long-term stability of tooth alignment and occlusion. The periodontal status, including gingival texture, periodontal pocket depth, tooth mobility, gingival recession, and alveolar bone level, should be evaluated before and during the orthodontic treatment. CBCT can also be used to evaluate the alveolar ridge height, alveolar bone thickness, alveolar ridge integrity (bone dehiscence and bone fenestration), and the relationship between root and bone. For patients with periodontal disease, the orthodontic treatment plan should be adjusted, the range of tooth movement should be reduced, and the communication between orthodontists and periodontists during the whole treatment process should be emphasized. Tooth alignment involves a variety of elements, including mesiodistal angulation, labiolingual angulation, rotation, crowding, labiolingual malposition, and arch form. The goal of orthodontic treatment is to achieve a combination of the above elements with good occlusal function, and the roots should be centered within the alveolar bone to maintain periodontal health. In patients treated with camouflage orthodontics, the root in the alveolar bone can be moderately displaced from the center and exhibit compensatory inclination within a safe range, achieving ideal tooth alignment and occlusion. In addition to the above elements, individual patients’ soft tissue characteristics should also be considered. For different vertical skeletal patterns and different face shapes, the arrangement of teeth should be considered individually. The dental arch should match the face shape. The appropriate width of the dental arch contributes to a coordinated buccal corridor. Otherwise, it can negatively impact smile esthetics. The skeletal relationship includes sagittal and vertical skeletal patterns. In addition to considering the relative position of the maxilla and mandible, the absolute sagittal jaw positions relative to the cranial base should also be considered in the diagnosis of a patient’s sagittal skeletal pattern, which is crucial for determining the target positions in orthodontic treatment planning. The impact of the vertical skeletal pattern on the sagittal pattern should also be considered in diagnosis and treatment plan design. Hyperdivergent skeletal pattern aggravates skeletal Class II malocclusion, while hypodivergent skeletal pattern aggravates skeletal Class III malocclusion. The jaw positional relationship is an important part of the dentofacial complex system, which should be integrated with soft tissue analysis to formulate the most appropriate treatment plans for patients presenting with protrusive facial deformities. Both the goals and limitations of modern orthodontic and orthognathic treatment are determined not only by the teeth and bones but also by the soft tissues of the face. The “soft tissue paradigm” advocated by modern orthodontics is to calculate the target position of incisors according to the soft tissue esthetics, so as to determine the details of treatment design such as the required space, the pattern of extraction, and the anchorage design. When making treatment plans, orthodontists should pay more attention to the patient’s frontal and lateral profile, and strive to make the treatment plan consistent with or close to the patient’s subjective esthetic expectations. Attention should also be paid to the adverse changes in profile caused by orthodontic treatment, particularly in patients with high zygomatic bone, sunken cheeks, etc. , The upper airway is closely related to health and life, and it is one of the key elements to be considered in the diagnosis and treatment of malocclusion. Following the retraction of incisors, the velopharyngeal, glossopharyngeal, and hypopharyngeal airway may become narrower. The target position of incisors should not only meet the needs of facial esthetics but also take into account the effect of incisor retraction on the size of the upper airway. For patients with upper airway stenosis, the amount of incisor retraction should be strictly controlled to maintain the inherent oral space and normal nasal respiratory function. If necessary, bimaxillary advancement surgery should be combined to correct the protrusion and increase the airway volume. Oral habits, such as sucking habit, abnormal tongue position, and tongue thrust swallowing, can break the balance of the internal and external strength of the jaw and arch, leading to malocclusion. Within the stomatognathic system, muscles often play a dominant role over bones. For patients exhibiting oral habits, orthodontists should ensure that changes in teeth, arches, and jaws are coordinated with muscle function during treatment. Only by removing oral habits and achieving normal perioral muscle function can ensure the long-term stability of orthodontic treatment. Temporomandibular disorders (TMD) are primarily characterized by joint pain, joint noise, and mandibular movement disorders. It is necessary to recognize the complexity of the etiology and pathophysiological mechanism of TMD and its impact on the stability of mandibular position and occlusion. , A high prevalence of TMD in skeletal class II patients referred for orthognathic surgery, especially in those with a pronounced overjet and high mandibular plane angle. A solid cusp-fossa relationship of the teeth should be established during orthodontic treatment, which is an important factor for the long-term stability of tooth alignment and occlusion. The periodontal status, including gingival texture, periodontal pocket depth, tooth mobility, gingival recession, and alveolar bone level, should be evaluated before and during the orthodontic treatment. CBCT can also be used to evaluate the alveolar ridge height, alveolar bone thickness, alveolar ridge integrity (bone dehiscence and bone fenestration), and the relationship between root and bone. For patients with periodontal disease, the orthodontic treatment plan should be adjusted, the range of tooth movement should be reduced, and the communication between orthodontists and periodontists during the whole treatment process should be emphasized. Jaw growth modification for protrusive facial deformities Mandibular functional treatment is a process in where orthodontists reposition the mandible after comprehensive evaluation of the stomatognathic system. Based on the growth potential of adolescents, the direction and extent of jaw growth should be effectively guided so as to alleviate the severity of skeletal malocclusions and improve the soft tissue contours. Orthodontic functional treatment should be initiated during the mixed dentition and early permanent dentition stages to take advantage of growth potential. The growth guidance of the mandible should be three-dimensional, involving the sagittal advancement accompanied by vertical and transverse adjustment. , In particular, vertical control could be emphasized in this process of mandibular advancement. , Adolescents with a low mandibular plane angle Adolescents with a low mandibular plane angle generally exhibit well-developed mandibular symphysis but may present with reduced prominence of soft tissue chin due to mandibular deficiency. Mandibular advancement devices (MADs), such as the activator and the twin-block can be used for mandibular advancement treatment. During the process, the eruption and reconstruction of the posterior teeth should be guided, to improve the growth pattern by increasing the facial height of the lower one-third and relieve the deep overbite of anterior teeth, which is beneficial for individuals with a short face pattern. Adolescents with a normal mandibular plane angle The treatment approach for adolescents with normal-angle dentofacial profiles is similar to that for those with low-angle deformities. Adolescents with significant sagittal jaw discrepancy can be treated with MADs combined with extraction orthodontic treatment. However, in the case of normal-angle adolescents, it is essential to maintain the lower facial height during mandibular advancement to prevent the elongation of molars and a clockwise rotation of the mandible, both of which could diminish chin prominence. Adolescents with a high mandibular plane angle The normal counterclockwise growth pattern of the mandible is transitioned to a clockwise trend in the hyperdivergent mandibular retrusion. Orthodontic treatment for adolescents exhibiting a long-face growth pattern is challenging. Due to the forward and inferior inclination of the occlusal plane, traditional MADs promote the saggital mandibular growth while adversely increasing the vertical dimension, resulting in unsatisfactory advancement of the mandible. The treatment of hyperdivergent skeletal Class II malocclusion is controversial. , Extraction treatment can retract the anterior teeth and improve lip prominence to a certain extent but has little effect on improving chin projection. Some scholars believe that Herbst or Twin-Block appliances with a vertical control strategy are recommended for hyperdivergent skeletal Class II malocclusion, with attention to leveling of the upper anterior occlusal plane. – The above approaches can avoid posterior teeth over-eruption and help to maintain the lower facial height during treatment. It should be emphasized that functional orthodontic treatment is not suitable for adolescents with a high mandibular plane angle, large lower facial height, and severe mentalis strain during lip closure. Anterior teeth retraction for protrusive facial deformities The pathogenesis of protrusive facial deformities manifests in two main forms: dental protrusion and Class II skeletal malocclusion. In all these cases, soft tissue features typically include upper lip protrusion and a compromised nasolabial relationship. The sagittal position and torque of the upper and lower anterior teeth influence the prominence of the upper and lower lip. Adjusting nasolabial and mentolabial appearance can enhance lateral appearance, contributing to overall facial harmony. Routine teeth extraction and anterior teeth retraction For cases of dental-facial protrusive deformities, teeth extraction is an effective treatment. For adolescent and adult skeletal malocclusion patients who have poor functional correction, and do not opt for orthognathic surgery, their options are limited to compensatory measures involving tooth and alveolar bone adjustments to address sagittal jaw discrepancies. This typically involves orthodontic extraction camouflage treatment, which can often compromise both appearance and stability. Assessment of soft tissue profile Maximizing the use of extraction space and precisely controlling the sagittal position and torque of the anterior teeth are crucial for ensuring treatment efficacy. To achieve a favorable retraction effect, ample space is essential for the anterior teeth to retract adequately. This is because changes in lip and tooth protrusion are disproportionate. Typically, a 1 mm retraction of incisors results in an ~0.6 mm decrease in lip protrusion. , Effective control of root retraction of the upper incisors is pivotal for reducing maxillary basal bone prominence. Teeth extraction mode For the treatment of protrusive facial deformities, choosing the right extraction mode depends on the severity of the protrusion. Extraction of anterior teeth helps alleviate anterior crowding and protrusion, and posterior tooth extraction helps to alleviate posterior crowding and control vertical dimensions. For adolescents and adults who undergo extraction, a common approach is to extract four first or second premolars or the extraction of maxillary first premolars and mandibular second premolars. , Additionally, in cases of open bite, extracting second premolars can help correct occlusal issues and facilitate better vertical control. Dentition distalization without premolar tooth extraction Orthodontists should be cautious when choosing to extract teeth. For example, patients with periodontitis and mild protrusion may be treated without tooth extraction. Interproximal enamel reduction, when necessary, can alleviate mild crowding and minimize the occurrence of “black triangles”. Orthodontic appliances like implant anchorage, the Pendulum appliance, the Frog appliance, and the extraoral arch can be used to achieve comprehensive distal movement of the teeth, correcting protrusion and deep overjet. – Implant anchorage, strategically placed in the subzygomatic ridge area, enables distal movement of the upper teeth without causing root interference. Vertical control strategies for protrusive facial deformities Previous studies and clinical observations have established that orthodontic treatment lacking vertical control could lead to tooth elongation and an increase in facial height. , , While this may benefit individuals with low-angle facial deformities, it often worsens the facial profile in individuals with high-angle facial protrusion. Vertical control can result in a counterclockwise rotation of the mandible, but this is still a contentious issue. , Our perspective is that vertical control strategies include the maintenance type and the mandibular counterclockwise rotation type, with the MP-SN angle being a key index. The maintenance type of vertical control alone typically does not suffice to improve facial profile. However, when combined with tooth extraction treatment, it can significantly improve the profile of individuals with protrusive facial deformities. In contrast, the other type of vertical control involves mandibular counterclockwise rotation achieved by reducing the height of the dental arches using various techniques. Importantly, the mandibular counterclockwise rotation type of vertical control represents a potentially independent and effective approach to ameliorating protrusion deformities. Orthodontic mechanisms of vertical control through counterclockwise rotation of the mandible The mechanisms of vertical control in adolescents Adolescents with facial protrusion are typically accompanied by a retrusive mandible. When employing orthodontic appliances, such as Twin-Block, Activator, Frankle II, and inclined guide plates in adolescents with high-angle protrusion, the potential risk of increasing vertical height must be carefully considered. In these cases, supplementary techniques such as J-hook appliances, auxiliary archwires, and implant anchorage can be utilized to control the vertical facial height and level the occlusal plane. This approach aims to induce counterclockwise rotation effect in mandibular growth and promote development of a Class I skeletal facial type. – The mechanisms involved in mandibular counterclockwise rotation in adults In adults, regardless of whether the patient presents with an anterior open bite or deep overbite, any occlusal contact during closure impedes the counterclockwise rotation of the mandible. The center of resistance for the mandible’s counterclockwise rotation is located in the condylar region. Through “compressing” the vertical height of the upper and lower tooth-alveolar bone complex, the mandible rotates forward and upward, driven by the action of the jaw-closing muscles. Generally, the vertical control technique involving mandibular counterclockwise rotation primarily focuses on intruding the teeth, reducing vertical dimension, and creating space for mandible to rotate, which contributes to the overall improvement of facial esthetics and occlusal function. Indications for mandibular counterclockwise rotation type of vertical control Personality factors of patients Tooth intrusion poses a risk of root resorption, necessitates active patient cooperation and understanding of potential risks. Patients with neurotic personality traits, such as anxiety and distrust, require strengthened doctor-patient communication and careful consideration of treatment plans. Consideration of the vertical height of the upper and lower arches The vertical height of the upper and lower arches varies among individuals with high-angle facial deformities. Parameters, such as U1-PP (vertical height of upper anterior teeth), U6-PP (vertical height of upper posterior teeth), L1-MP (vertical height of lower anterior teeth), and L6-MP (vertical height of lower posterior teeth), should be evaluated clinically and using imaging. These values inform the selection of the intrusion site and the design mode of mandibular counterclockwise rotation type of vertical control. Soft-tissue contour factors The patient’s soft tissue profile influences the selection of the intrusion site. If a patient’s smile shows insufficient teeth exposure due to excessive soft tissue length of the upper lip, methods involving intrusion of the upper anterior teeth may not be suitable. Vertical control strategies for counterclockwise rotation of the mandible The strategy involving mandibular counterclockwise rotation in individuals with high-angle facial deformities requires careful consideration of individual variations in facial contour, as well as clinical and radiographic parameters of the dental arches. Based on these factors, a personalized combination of intrusions can be selected for different manifestations of high-angle facial deformities. It should be noted that the counterclockwise rotation of the mandible is not based on the premise of the counterclockwise rotation of the occlusal plane. The following is a classification (Fig. ): Gummy smile with vertical overdevelopment of upper arch: Intrusion of upper anterior and posterior teeth, upward displacement and high-probability counterclockwise rotation of occlusal plane, and counterclockwise rotation of mandible. Gummy smile with overdeveloped upper anterior and lower posterior teeth: Intrusion of upper anterior and lower posterior teeth, counterclockwise rotation of occlusal plane and mandible. No gummy smile with open bite anterior teeth and vertically overdeveloped upper and lower posterior teeth: Intrusion of upper and lower posterior teeth, clockwise rotation of occlusal plane, and counterclockwise rotation of mandible. Gummy smile with open bite anterior teeth and vertical overdevelopment of upper and lower posterior teeth: Intrusion of upper anterior, upper and lower posterior teeth, upward movement of occlusal plane, and counterclockwise rotation of mandible. Gummy smile with vertically overdeveloped upper and lower teeth: Intrusion of upper and lower anterior and posterior teeth, upward and high-probability counterclockwise rotation of occlusal plane, and counterclockwise rotation of mandible. No gummy smile with normal upper arch and vertically overdeveloped lower arch: Intrusion of lower anterior and posterior teeth, unchanged occlusal plane, and counterclockwise rotation of mandible. Each combination targets specific characteristics of facial deformity, ensuring a tailored treatment approach to achieve optimal outcomes. Mandibular functional treatment is a process in where orthodontists reposition the mandible after comprehensive evaluation of the stomatognathic system. Based on the growth potential of adolescents, the direction and extent of jaw growth should be effectively guided so as to alleviate the severity of skeletal malocclusions and improve the soft tissue contours. Orthodontic functional treatment should be initiated during the mixed dentition and early permanent dentition stages to take advantage of growth potential. The growth guidance of the mandible should be three-dimensional, involving the sagittal advancement accompanied by vertical and transverse adjustment. , In particular, vertical control could be emphasized in this process of mandibular advancement. , Adolescents with a low mandibular plane angle Adolescents with a low mandibular plane angle generally exhibit well-developed mandibular symphysis but may present with reduced prominence of soft tissue chin due to mandibular deficiency. Mandibular advancement devices (MADs), such as the activator and the twin-block can be used for mandibular advancement treatment. During the process, the eruption and reconstruction of the posterior teeth should be guided, to improve the growth pattern by increasing the facial height of the lower one-third and relieve the deep overbite of anterior teeth, which is beneficial for individuals with a short face pattern. Adolescents with a normal mandibular plane angle The treatment approach for adolescents with normal-angle dentofacial profiles is similar to that for those with low-angle deformities. Adolescents with significant sagittal jaw discrepancy can be treated with MADs combined with extraction orthodontic treatment. However, in the case of normal-angle adolescents, it is essential to maintain the lower facial height during mandibular advancement to prevent the elongation of molars and a clockwise rotation of the mandible, both of which could diminish chin prominence. Adolescents with a high mandibular plane angle The normal counterclockwise growth pattern of the mandible is transitioned to a clockwise trend in the hyperdivergent mandibular retrusion. Orthodontic treatment for adolescents exhibiting a long-face growth pattern is challenging. Due to the forward and inferior inclination of the occlusal plane, traditional MADs promote the saggital mandibular growth while adversely increasing the vertical dimension, resulting in unsatisfactory advancement of the mandible. The treatment of hyperdivergent skeletal Class II malocclusion is controversial. , Extraction treatment can retract the anterior teeth and improve lip prominence to a certain extent but has little effect on improving chin projection. Some scholars believe that Herbst or Twin-Block appliances with a vertical control strategy are recommended for hyperdivergent skeletal Class II malocclusion, with attention to leveling of the upper anterior occlusal plane. – The above approaches can avoid posterior teeth over-eruption and help to maintain the lower facial height during treatment. It should be emphasized that functional orthodontic treatment is not suitable for adolescents with a high mandibular plane angle, large lower facial height, and severe mentalis strain during lip closure. Adolescents with a low mandibular plane angle generally exhibit well-developed mandibular symphysis but may present with reduced prominence of soft tissue chin due to mandibular deficiency. Mandibular advancement devices (MADs), such as the activator and the twin-block can be used for mandibular advancement treatment. During the process, the eruption and reconstruction of the posterior teeth should be guided, to improve the growth pattern by increasing the facial height of the lower one-third and relieve the deep overbite of anterior teeth, which is beneficial for individuals with a short face pattern. The treatment approach for adolescents with normal-angle dentofacial profiles is similar to that for those with low-angle deformities. Adolescents with significant sagittal jaw discrepancy can be treated with MADs combined with extraction orthodontic treatment. However, in the case of normal-angle adolescents, it is essential to maintain the lower facial height during mandibular advancement to prevent the elongation of molars and a clockwise rotation of the mandible, both of which could diminish chin prominence. The normal counterclockwise growth pattern of the mandible is transitioned to a clockwise trend in the hyperdivergent mandibular retrusion. Orthodontic treatment for adolescents exhibiting a long-face growth pattern is challenging. Due to the forward and inferior inclination of the occlusal plane, traditional MADs promote the saggital mandibular growth while adversely increasing the vertical dimension, resulting in unsatisfactory advancement of the mandible. The treatment of hyperdivergent skeletal Class II malocclusion is controversial. , Extraction treatment can retract the anterior teeth and improve lip prominence to a certain extent but has little effect on improving chin projection. Some scholars believe that Herbst or Twin-Block appliances with a vertical control strategy are recommended for hyperdivergent skeletal Class II malocclusion, with attention to leveling of the upper anterior occlusal plane. – The above approaches can avoid posterior teeth over-eruption and help to maintain the lower facial height during treatment. It should be emphasized that functional orthodontic treatment is not suitable for adolescents with a high mandibular plane angle, large lower facial height, and severe mentalis strain during lip closure. The pathogenesis of protrusive facial deformities manifests in two main forms: dental protrusion and Class II skeletal malocclusion. In all these cases, soft tissue features typically include upper lip protrusion and a compromised nasolabial relationship. The sagittal position and torque of the upper and lower anterior teeth influence the prominence of the upper and lower lip. Adjusting nasolabial and mentolabial appearance can enhance lateral appearance, contributing to overall facial harmony. Routine teeth extraction and anterior teeth retraction For cases of dental-facial protrusive deformities, teeth extraction is an effective treatment. For adolescent and adult skeletal malocclusion patients who have poor functional correction, and do not opt for orthognathic surgery, their options are limited to compensatory measures involving tooth and alveolar bone adjustments to address sagittal jaw discrepancies. This typically involves orthodontic extraction camouflage treatment, which can often compromise both appearance and stability. Assessment of soft tissue profile Maximizing the use of extraction space and precisely controlling the sagittal position and torque of the anterior teeth are crucial for ensuring treatment efficacy. To achieve a favorable retraction effect, ample space is essential for the anterior teeth to retract adequately. This is because changes in lip and tooth protrusion are disproportionate. Typically, a 1 mm retraction of incisors results in an ~0.6 mm decrease in lip protrusion. , Effective control of root retraction of the upper incisors is pivotal for reducing maxillary basal bone prominence. Teeth extraction mode For the treatment of protrusive facial deformities, choosing the right extraction mode depends on the severity of the protrusion. Extraction of anterior teeth helps alleviate anterior crowding and protrusion, and posterior tooth extraction helps to alleviate posterior crowding and control vertical dimensions. For adolescents and adults who undergo extraction, a common approach is to extract four first or second premolars or the extraction of maxillary first premolars and mandibular second premolars. , Additionally, in cases of open bite, extracting second premolars can help correct occlusal issues and facilitate better vertical control. Dentition distalization without premolar tooth extraction Orthodontists should be cautious when choosing to extract teeth. For example, patients with periodontitis and mild protrusion may be treated without tooth extraction. Interproximal enamel reduction, when necessary, can alleviate mild crowding and minimize the occurrence of “black triangles”. Orthodontic appliances like implant anchorage, the Pendulum appliance, the Frog appliance, and the extraoral arch can be used to achieve comprehensive distal movement of the teeth, correcting protrusion and deep overjet. – Implant anchorage, strategically placed in the subzygomatic ridge area, enables distal movement of the upper teeth without causing root interference. For cases of dental-facial protrusive deformities, teeth extraction is an effective treatment. For adolescent and adult skeletal malocclusion patients who have poor functional correction, and do not opt for orthognathic surgery, their options are limited to compensatory measures involving tooth and alveolar bone adjustments to address sagittal jaw discrepancies. This typically involves orthodontic extraction camouflage treatment, which can often compromise both appearance and stability. Assessment of soft tissue profile Maximizing the use of extraction space and precisely controlling the sagittal position and torque of the anterior teeth are crucial for ensuring treatment efficacy. To achieve a favorable retraction effect, ample space is essential for the anterior teeth to retract adequately. This is because changes in lip and tooth protrusion are disproportionate. Typically, a 1 mm retraction of incisors results in an ~0.6 mm decrease in lip protrusion. , Effective control of root retraction of the upper incisors is pivotal for reducing maxillary basal bone prominence. Teeth extraction mode For the treatment of protrusive facial deformities, choosing the right extraction mode depends on the severity of the protrusion. Extraction of anterior teeth helps alleviate anterior crowding and protrusion, and posterior tooth extraction helps to alleviate posterior crowding and control vertical dimensions. For adolescents and adults who undergo extraction, a common approach is to extract four first or second premolars or the extraction of maxillary first premolars and mandibular second premolars. , Additionally, in cases of open bite, extracting second premolars can help correct occlusal issues and facilitate better vertical control. Maximizing the use of extraction space and precisely controlling the sagittal position and torque of the anterior teeth are crucial for ensuring treatment efficacy. To achieve a favorable retraction effect, ample space is essential for the anterior teeth to retract adequately. This is because changes in lip and tooth protrusion are disproportionate. Typically, a 1 mm retraction of incisors results in an ~0.6 mm decrease in lip protrusion. , Effective control of root retraction of the upper incisors is pivotal for reducing maxillary basal bone prominence. For the treatment of protrusive facial deformities, choosing the right extraction mode depends on the severity of the protrusion. Extraction of anterior teeth helps alleviate anterior crowding and protrusion, and posterior tooth extraction helps to alleviate posterior crowding and control vertical dimensions. For adolescents and adults who undergo extraction, a common approach is to extract four first or second premolars or the extraction of maxillary first premolars and mandibular second premolars. , Additionally, in cases of open bite, extracting second premolars can help correct occlusal issues and facilitate better vertical control. Orthodontists should be cautious when choosing to extract teeth. For example, patients with periodontitis and mild protrusion may be treated without tooth extraction. Interproximal enamel reduction, when necessary, can alleviate mild crowding and minimize the occurrence of “black triangles”. Orthodontic appliances like implant anchorage, the Pendulum appliance, the Frog appliance, and the extraoral arch can be used to achieve comprehensive distal movement of the teeth, correcting protrusion and deep overjet. – Implant anchorage, strategically placed in the subzygomatic ridge area, enables distal movement of the upper teeth without causing root interference. Previous studies and clinical observations have established that orthodontic treatment lacking vertical control could lead to tooth elongation and an increase in facial height. , , While this may benefit individuals with low-angle facial deformities, it often worsens the facial profile in individuals with high-angle facial protrusion. Vertical control can result in a counterclockwise rotation of the mandible, but this is still a contentious issue. , Our perspective is that vertical control strategies include the maintenance type and the mandibular counterclockwise rotation type, with the MP-SN angle being a key index. The maintenance type of vertical control alone typically does not suffice to improve facial profile. However, when combined with tooth extraction treatment, it can significantly improve the profile of individuals with protrusive facial deformities. In contrast, the other type of vertical control involves mandibular counterclockwise rotation achieved by reducing the height of the dental arches using various techniques. Importantly, the mandibular counterclockwise rotation type of vertical control represents a potentially independent and effective approach to ameliorating protrusion deformities. Orthodontic mechanisms of vertical control through counterclockwise rotation of the mandible The mechanisms of vertical control in adolescents Adolescents with facial protrusion are typically accompanied by a retrusive mandible. When employing orthodontic appliances, such as Twin-Block, Activator, Frankle II, and inclined guide plates in adolescents with high-angle protrusion, the potential risk of increasing vertical height must be carefully considered. In these cases, supplementary techniques such as J-hook appliances, auxiliary archwires, and implant anchorage can be utilized to control the vertical facial height and level the occlusal plane. This approach aims to induce counterclockwise rotation effect in mandibular growth and promote development of a Class I skeletal facial type. – The mechanisms involved in mandibular counterclockwise rotation in adults In adults, regardless of whether the patient presents with an anterior open bite or deep overbite, any occlusal contact during closure impedes the counterclockwise rotation of the mandible. The center of resistance for the mandible’s counterclockwise rotation is located in the condylar region. Through “compressing” the vertical height of the upper and lower tooth-alveolar bone complex, the mandible rotates forward and upward, driven by the action of the jaw-closing muscles. Generally, the vertical control technique involving mandibular counterclockwise rotation primarily focuses on intruding the teeth, reducing vertical dimension, and creating space for mandible to rotate, which contributes to the overall improvement of facial esthetics and occlusal function. Indications for mandibular counterclockwise rotation type of vertical control Personality factors of patients Tooth intrusion poses a risk of root resorption, necessitates active patient cooperation and understanding of potential risks. Patients with neurotic personality traits, such as anxiety and distrust, require strengthened doctor-patient communication and careful consideration of treatment plans. Consideration of the vertical height of the upper and lower arches The vertical height of the upper and lower arches varies among individuals with high-angle facial deformities. Parameters, such as U1-PP (vertical height of upper anterior teeth), U6-PP (vertical height of upper posterior teeth), L1-MP (vertical height of lower anterior teeth), and L6-MP (vertical height of lower posterior teeth), should be evaluated clinically and using imaging. These values inform the selection of the intrusion site and the design mode of mandibular counterclockwise rotation type of vertical control. Soft-tissue contour factors The patient’s soft tissue profile influences the selection of the intrusion site. If a patient’s smile shows insufficient teeth exposure due to excessive soft tissue length of the upper lip, methods involving intrusion of the upper anterior teeth may not be suitable. Vertical control strategies for counterclockwise rotation of the mandible The strategy involving mandibular counterclockwise rotation in individuals with high-angle facial deformities requires careful consideration of individual variations in facial contour, as well as clinical and radiographic parameters of the dental arches. Based on these factors, a personalized combination of intrusions can be selected for different manifestations of high-angle facial deformities. It should be noted that the counterclockwise rotation of the mandible is not based on the premise of the counterclockwise rotation of the occlusal plane. The following is a classification (Fig. ): Gummy smile with vertical overdevelopment of upper arch: Intrusion of upper anterior and posterior teeth, upward displacement and high-probability counterclockwise rotation of occlusal plane, and counterclockwise rotation of mandible. Gummy smile with overdeveloped upper anterior and lower posterior teeth: Intrusion of upper anterior and lower posterior teeth, counterclockwise rotation of occlusal plane and mandible. No gummy smile with open bite anterior teeth and vertically overdeveloped upper and lower posterior teeth: Intrusion of upper and lower posterior teeth, clockwise rotation of occlusal plane, and counterclockwise rotation of mandible. Gummy smile with open bite anterior teeth and vertical overdevelopment of upper and lower posterior teeth: Intrusion of upper anterior, upper and lower posterior teeth, upward movement of occlusal plane, and counterclockwise rotation of mandible. Gummy smile with vertically overdeveloped upper and lower teeth: Intrusion of upper and lower anterior and posterior teeth, upward and high-probability counterclockwise rotation of occlusal plane, and counterclockwise rotation of mandible. No gummy smile with normal upper arch and vertically overdeveloped lower arch: Intrusion of lower anterior and posterior teeth, unchanged occlusal plane, and counterclockwise rotation of mandible. Each combination targets specific characteristics of facial deformity, ensuring a tailored treatment approach to achieve optimal outcomes. The mechanisms of vertical control in adolescents Adolescents with facial protrusion are typically accompanied by a retrusive mandible. When employing orthodontic appliances, such as Twin-Block, Activator, Frankle II, and inclined guide plates in adolescents with high-angle protrusion, the potential risk of increasing vertical height must be carefully considered. In these cases, supplementary techniques such as J-hook appliances, auxiliary archwires, and implant anchorage can be utilized to control the vertical facial height and level the occlusal plane. This approach aims to induce counterclockwise rotation effect in mandibular growth and promote development of a Class I skeletal facial type. – The mechanisms involved in mandibular counterclockwise rotation in adults In adults, regardless of whether the patient presents with an anterior open bite or deep overbite, any occlusal contact during closure impedes the counterclockwise rotation of the mandible. The center of resistance for the mandible’s counterclockwise rotation is located in the condylar region. Through “compressing” the vertical height of the upper and lower tooth-alveolar bone complex, the mandible rotates forward and upward, driven by the action of the jaw-closing muscles. Generally, the vertical control technique involving mandibular counterclockwise rotation primarily focuses on intruding the teeth, reducing vertical dimension, and creating space for mandible to rotate, which contributes to the overall improvement of facial esthetics and occlusal function. Adolescents with facial protrusion are typically accompanied by a retrusive mandible. When employing orthodontic appliances, such as Twin-Block, Activator, Frankle II, and inclined guide plates in adolescents with high-angle protrusion, the potential risk of increasing vertical height must be carefully considered. In these cases, supplementary techniques such as J-hook appliances, auxiliary archwires, and implant anchorage can be utilized to control the vertical facial height and level the occlusal plane. This approach aims to induce counterclockwise rotation effect in mandibular growth and promote development of a Class I skeletal facial type. – In adults, regardless of whether the patient presents with an anterior open bite or deep overbite, any occlusal contact during closure impedes the counterclockwise rotation of the mandible. The center of resistance for the mandible’s counterclockwise rotation is located in the condylar region. Through “compressing” the vertical height of the upper and lower tooth-alveolar bone complex, the mandible rotates forward and upward, driven by the action of the jaw-closing muscles. Generally, the vertical control technique involving mandibular counterclockwise rotation primarily focuses on intruding the teeth, reducing vertical dimension, and creating space for mandible to rotate, which contributes to the overall improvement of facial esthetics and occlusal function. Personality factors of patients Tooth intrusion poses a risk of root resorption, necessitates active patient cooperation and understanding of potential risks. Patients with neurotic personality traits, such as anxiety and distrust, require strengthened doctor-patient communication and careful consideration of treatment plans. Consideration of the vertical height of the upper and lower arches The vertical height of the upper and lower arches varies among individuals with high-angle facial deformities. Parameters, such as U1-PP (vertical height of upper anterior teeth), U6-PP (vertical height of upper posterior teeth), L1-MP (vertical height of lower anterior teeth), and L6-MP (vertical height of lower posterior teeth), should be evaluated clinically and using imaging. These values inform the selection of the intrusion site and the design mode of mandibular counterclockwise rotation type of vertical control. Soft-tissue contour factors The patient’s soft tissue profile influences the selection of the intrusion site. If a patient’s smile shows insufficient teeth exposure due to excessive soft tissue length of the upper lip, methods involving intrusion of the upper anterior teeth may not be suitable. Tooth intrusion poses a risk of root resorption, necessitates active patient cooperation and understanding of potential risks. Patients with neurotic personality traits, such as anxiety and distrust, require strengthened doctor-patient communication and careful consideration of treatment plans. The vertical height of the upper and lower arches varies among individuals with high-angle facial deformities. Parameters, such as U1-PP (vertical height of upper anterior teeth), U6-PP (vertical height of upper posterior teeth), L1-MP (vertical height of lower anterior teeth), and L6-MP (vertical height of lower posterior teeth), should be evaluated clinically and using imaging. These values inform the selection of the intrusion site and the design mode of mandibular counterclockwise rotation type of vertical control. The patient’s soft tissue profile influences the selection of the intrusion site. If a patient’s smile shows insufficient teeth exposure due to excessive soft tissue length of the upper lip, methods involving intrusion of the upper anterior teeth may not be suitable. The strategy involving mandibular counterclockwise rotation in individuals with high-angle facial deformities requires careful consideration of individual variations in facial contour, as well as clinical and radiographic parameters of the dental arches. Based on these factors, a personalized combination of intrusions can be selected for different manifestations of high-angle facial deformities. It should be noted that the counterclockwise rotation of the mandible is not based on the premise of the counterclockwise rotation of the occlusal plane. The following is a classification (Fig. ): Gummy smile with vertical overdevelopment of upper arch: Intrusion of upper anterior and posterior teeth, upward displacement and high-probability counterclockwise rotation of occlusal plane, and counterclockwise rotation of mandible. Gummy smile with overdeveloped upper anterior and lower posterior teeth: Intrusion of upper anterior and lower posterior teeth, counterclockwise rotation of occlusal plane and mandible. No gummy smile with open bite anterior teeth and vertically overdeveloped upper and lower posterior teeth: Intrusion of upper and lower posterior teeth, clockwise rotation of occlusal plane, and counterclockwise rotation of mandible. Gummy smile with open bite anterior teeth and vertical overdevelopment of upper and lower posterior teeth: Intrusion of upper anterior, upper and lower posterior teeth, upward movement of occlusal plane, and counterclockwise rotation of mandible. Gummy smile with vertically overdeveloped upper and lower teeth: Intrusion of upper and lower anterior and posterior teeth, upward and high-probability counterclockwise rotation of occlusal plane, and counterclockwise rotation of mandible. No gummy smile with normal upper arch and vertically overdeveloped lower arch: Intrusion of lower anterior and posterior teeth, unchanged occlusal plane, and counterclockwise rotation of mandible. Each combination targets specific characteristics of facial deformity, ensuring a tailored treatment approach to achieve optimal outcomes. In cases of skeletal malocclusions where patients have reached skeletal maturity or require treatment beyond orthodontic compensation, a combined orthognathic and orthodontic approach is frequently recommended. This comprehensive method integrates orthognathic surgery to correct jaw positions for optimal soft tissue esthetics, alongside pre- and postoperative orthodontic interventions. These interventions play a pivotal role in reconstructing the occlusal relationship between the upper and lower dental arches and reinstating the functionality of the stomatognathic system. This coordinated approach ensures not only esthetic improvement but also functional restoration, resulting in more favorable treatment outcomes. Surgical design Crafting a surgical plan for orthognathic surgery requires a thorough analysis of the morphological mechanisms and the severity of the malocclusion. In cases of severe protrusive facial deformities, it’s crucial to conduct a thorough assessment encompassing a precise understanding of the abnormal relationships and degrees of malformation within the dental arches and jaw positions in various dimensions and the dimensions of the upper airway. Such a systematic approach enables optimal orthodontic tooth movement and jaw displacement during orthognathic surgery, achieving ideal functional and esthetic outcomes. , For patients with protrusive facial deformities, it’s important to be cautious of intraoperative mandibular advancement and traction on the masseter and suprahyoid muscle groups, as these maneuvers may heighten the risk of postoperative relapse. Research indicates that combining maxillary setback with mandibular advancement surgery yields superior postoperative stability compared to mandibular advancement surgery alone. Recognizing the constraints of orthognathic surgery is vital in treatment planning. It’s essential to consider the anatomical limitations of bone block mobility and the restrictions imposed by soft tissues. If refining local skeletal contours cannot be accomplished in a single procedure, additional surgeries may be necessary for further contouring. Additionally, there’s a growing trend in integrating orthognathic surgery with facial plastic surgery in craniofacial esthetic procedures. This integration offers the opportunity to incorporate interventions that modify nasal structure and facial soft tissue contours, further enhancing overall facial harmony and esthetic outcomes. Maxillary osteotomy Le Fort I osteotomy The Le Fort I osteotomy (Fig. ) is a surgical procedure characterized by a horizontal bone cut positioned above the inferior margin of the pterygoid process, the anterior wall of the maxillary sinus, the zygomaticomaxillary buttress, and above the maxillary tuberosity. , This technique mimics the anatomical course of a classic Le Fort Type I fracture. Its objective is to address maxillary hyperplasia by detaching and globally mobilizing the maxillary bone, encompassing the entire dental arch. Anterior maxillary osteotomy (AMO) The AMO procedure (Fig. ) entails two key steps: initially, a horizontal bone cut is made above the inferior margin of the pterygoid process and the anterior wall of the maxillary sinus. This is followed by a vertical bone cut through the gap left post-extraction of the anterior molars. This sequential process results in the detachment of the bone segment in the anterior maxillary region. By mobilizing this bone segment, which includes the anterior nasal spine and the anterior floor of the nose, AMO effectively corrects protrusive facial deformities present in the anterior maxillary teeth and alveolar bone. Mandibular osteotomy Sagittal split ramus osteotomy (SSRO) The SSRO procedure (Fig. ) involves a horizontal split of the mandible along its anatomical structure, , , as outlined in various references. This division divides the mandible into two segments: a proximal segment comprising the condyle, coronoid process, and mandibular angle, and a distal segment encompassing the body of the mandible, the entire mandibular dentition, and the neurovascular bundle of the inferior alveolar nerve. Through mobilization of the distal segment, SSRO effectively addresses mandibular protrusive facial deformities. Anterior mandibular subapical osteotomy (AMSO) The AMSO procedure (Fig. ) entails a sequential process. Initially, a horizontal bone cut is made at least 5 mm below the apices of the mandibular anterior teeth on the labial side of the mandible. Subsequently, a vertical bone cut is performed through the gap left post-extraction of the anterior molars. This sequential action results in the detachment of the bone segment in the mandibular anterior region. By mobilizing this bone segment, which includes the mandibular anterior teeth, AMSO effectively corrects protrusive facial deformities present in the anterior mandibular teeth and alveolar bone. Genioplasty Genioplasty (Fig. ) is a surgical procedure that involves a horizontal cut below the apices of the mandibular anterior teeth and beneath the mental foramen. This incision allows for the mobilization of the chin bone segment. Through this process, genioplasty effectively corrects protrusive facial deformities in chin development. , Surgery-first orthognathic approach The surgery-first approach (SFA) involves performing orthognathic surgery before initiating orthodontic treatment. The advantages of SFA include the immediate improvement of facial esthetics and a reduction in the duration of orthodontic treatment. The latter is related to a more physiological position of the teeth, the arrangement of the dental arch through surgery, and the regional acceleratory phenomenon after surgery. Cases with protrusive deformities that do not need too much preoperative orthodontic alignment and decompensation are regarded as indications of SFA. However, as with any immature technique, consensus regarding indications and surgical planning, as well as the evidence of long-term stability, is still lacking. – Preoperative and postoperative orthodontic treatment in orthognathic surgery Preoperative orthodontics plays a critical role in preparing patients for orthognathic surgery by addressing several key objectives. Its primary focus includes eliminating compensatory tooth tilting, harmonizing the morphology and size of the maxillary and mandibular dental arches, facilitating jaw displacement, occlusal alignment, and the establishment of a stable occlusion post-surgery. , Postoperative orthodontics further enhances treatment outcomes by refining tooth alignment and optimizing occlusal relationships, significantly contributing to the long-term stability of combined orthodontic and orthognathic treatments. , Close coordination between orthodontic and orthognathic surgery specialist teams is paramount for achieving optimal results, with orthodontic treatment being an integral component of the overall therapeutic approach. Crafting a surgical plan for orthognathic surgery requires a thorough analysis of the morphological mechanisms and the severity of the malocclusion. In cases of severe protrusive facial deformities, it’s crucial to conduct a thorough assessment encompassing a precise understanding of the abnormal relationships and degrees of malformation within the dental arches and jaw positions in various dimensions and the dimensions of the upper airway. Such a systematic approach enables optimal orthodontic tooth movement and jaw displacement during orthognathic surgery, achieving ideal functional and esthetic outcomes. , For patients with protrusive facial deformities, it’s important to be cautious of intraoperative mandibular advancement and traction on the masseter and suprahyoid muscle groups, as these maneuvers may heighten the risk of postoperative relapse. Research indicates that combining maxillary setback with mandibular advancement surgery yields superior postoperative stability compared to mandibular advancement surgery alone. Recognizing the constraints of orthognathic surgery is vital in treatment planning. It’s essential to consider the anatomical limitations of bone block mobility and the restrictions imposed by soft tissues. If refining local skeletal contours cannot be accomplished in a single procedure, additional surgeries may be necessary for further contouring. Additionally, there’s a growing trend in integrating orthognathic surgery with facial plastic surgery in craniofacial esthetic procedures. This integration offers the opportunity to incorporate interventions that modify nasal structure and facial soft tissue contours, further enhancing overall facial harmony and esthetic outcomes. Le Fort I osteotomy The Le Fort I osteotomy (Fig. ) is a surgical procedure characterized by a horizontal bone cut positioned above the inferior margin of the pterygoid process, the anterior wall of the maxillary sinus, the zygomaticomaxillary buttress, and above the maxillary tuberosity. , This technique mimics the anatomical course of a classic Le Fort Type I fracture. Its objective is to address maxillary hyperplasia by detaching and globally mobilizing the maxillary bone, encompassing the entire dental arch. Anterior maxillary osteotomy (AMO) The AMO procedure (Fig. ) entails two key steps: initially, a horizontal bone cut is made above the inferior margin of the pterygoid process and the anterior wall of the maxillary sinus. This is followed by a vertical bone cut through the gap left post-extraction of the anterior molars. This sequential process results in the detachment of the bone segment in the anterior maxillary region. By mobilizing this bone segment, which includes the anterior nasal spine and the anterior floor of the nose, AMO effectively corrects protrusive facial deformities present in the anterior maxillary teeth and alveolar bone. The Le Fort I osteotomy (Fig. ) is a surgical procedure characterized by a horizontal bone cut positioned above the inferior margin of the pterygoid process, the anterior wall of the maxillary sinus, the zygomaticomaxillary buttress, and above the maxillary tuberosity. , This technique mimics the anatomical course of a classic Le Fort Type I fracture. Its objective is to address maxillary hyperplasia by detaching and globally mobilizing the maxillary bone, encompassing the entire dental arch. The AMO procedure (Fig. ) entails two key steps: initially, a horizontal bone cut is made above the inferior margin of the pterygoid process and the anterior wall of the maxillary sinus. This is followed by a vertical bone cut through the gap left post-extraction of the anterior molars. This sequential process results in the detachment of the bone segment in the anterior maxillary region. By mobilizing this bone segment, which includes the anterior nasal spine and the anterior floor of the nose, AMO effectively corrects protrusive facial deformities present in the anterior maxillary teeth and alveolar bone. Sagittal split ramus osteotomy (SSRO) The SSRO procedure (Fig. ) involves a horizontal split of the mandible along its anatomical structure, , , as outlined in various references. This division divides the mandible into two segments: a proximal segment comprising the condyle, coronoid process, and mandibular angle, and a distal segment encompassing the body of the mandible, the entire mandibular dentition, and the neurovascular bundle of the inferior alveolar nerve. Through mobilization of the distal segment, SSRO effectively addresses mandibular protrusive facial deformities. Anterior mandibular subapical osteotomy (AMSO) The AMSO procedure (Fig. ) entails a sequential process. Initially, a horizontal bone cut is made at least 5 mm below the apices of the mandibular anterior teeth on the labial side of the mandible. Subsequently, a vertical bone cut is performed through the gap left post-extraction of the anterior molars. This sequential action results in the detachment of the bone segment in the mandibular anterior region. By mobilizing this bone segment, which includes the mandibular anterior teeth, AMSO effectively corrects protrusive facial deformities present in the anterior mandibular teeth and alveolar bone. Genioplasty Genioplasty (Fig. ) is a surgical procedure that involves a horizontal cut below the apices of the mandibular anterior teeth and beneath the mental foramen. This incision allows for the mobilization of the chin bone segment. Through this process, genioplasty effectively corrects protrusive facial deformities in chin development. , The SSRO procedure (Fig. ) involves a horizontal split of the mandible along its anatomical structure, , , as outlined in various references. This division divides the mandible into two segments: a proximal segment comprising the condyle, coronoid process, and mandibular angle, and a distal segment encompassing the body of the mandible, the entire mandibular dentition, and the neurovascular bundle of the inferior alveolar nerve. Through mobilization of the distal segment, SSRO effectively addresses mandibular protrusive facial deformities. The AMSO procedure (Fig. ) entails a sequential process. Initially, a horizontal bone cut is made at least 5 mm below the apices of the mandibular anterior teeth on the labial side of the mandible. Subsequently, a vertical bone cut is performed through the gap left post-extraction of the anterior molars. This sequential action results in the detachment of the bone segment in the mandibular anterior region. By mobilizing this bone segment, which includes the mandibular anterior teeth, AMSO effectively corrects protrusive facial deformities present in the anterior mandibular teeth and alveolar bone. Genioplasty (Fig. ) is a surgical procedure that involves a horizontal cut below the apices of the mandibular anterior teeth and beneath the mental foramen. This incision allows for the mobilization of the chin bone segment. Through this process, genioplasty effectively corrects protrusive facial deformities in chin development. , The surgery-first approach (SFA) involves performing orthognathic surgery before initiating orthodontic treatment. The advantages of SFA include the immediate improvement of facial esthetics and a reduction in the duration of orthodontic treatment. The latter is related to a more physiological position of the teeth, the arrangement of the dental arch through surgery, and the regional acceleratory phenomenon after surgery. Cases with protrusive deformities that do not need too much preoperative orthodontic alignment and decompensation are regarded as indications of SFA. However, as with any immature technique, consensus regarding indications and surgical planning, as well as the evidence of long-term stability, is still lacking. – Preoperative orthodontics plays a critical role in preparing patients for orthognathic surgery by addressing several key objectives. Its primary focus includes eliminating compensatory tooth tilting, harmonizing the morphology and size of the maxillary and mandibular dental arches, facilitating jaw displacement, occlusal alignment, and the establishment of a stable occlusion post-surgery. , Postoperative orthodontics further enhances treatment outcomes by refining tooth alignment and optimizing occlusal relationships, significantly contributing to the long-term stability of combined orthodontic and orthognathic treatments. , Close coordination between orthodontic and orthognathic surgery specialist teams is paramount for achieving optimal results, with orthodontic treatment being an integral component of the overall therapeutic approach. The protrusive deformity is one of the main causes affecting facial esthetics. Due to the complex etiology and diverse manifestations of protrusive deformities, orthodontic diagnosis and treatment strategies require consideration of multiple dimensions and comprehensive factors. Multidimensionality refers to the addition of a temporal dimension to the traditional three-dimensional concept, including sagittal, vertical, horizontal, and growth and development dimensions. Comprehensive factors encompass seven aspects, including teeth alignment, jaw relationship, facial contour, periodontal condition, upper airway patency, temporomandibular joints, and muscle balance. This consensus also provides a detailed discussion on the indications, intrusion strategies, and risk control associated with vertical control techniques for protrusive facial deformities. While numerous clinical studies on the treatment of protrusive facial deformities exist, there is a future need for large-sample, multi-center randomized controlled clinical trials. There is a particular lack of prospective research on vertical control techniques, and evidence-based thinking needs to be integrated into the evaluation of treatment efficacy and postoperative stability analysis for protrusive deformities. With the continuous innovation of non-bracket invisible orthodontic technology and its combined application with other orthodontic techniques, the range of orthodontic appliance options for protrusion deformities is continuously expanding, benefiting more patients. In recent years, artificial intelligence has been gradually popularized in the medical field, and its powerful data analysis and processing capabilities have brought about significant changes in the diagnosis, treatment, and prognosis prediction of protrusive deformities. In conclusion, the diagnosis and treatment of protrusive facial deformities are a systematic endeavor. As our understanding of the pathogenesis of protrusive facial deformities deepens, as clinical research continues to evolve, and as better research methods are applied in clinical practice, it will undoubtedly bring about more optimized results for the treatment and long-term stability of patients with protrusive facial deformities. |
Ethnobotanical, Phytochemistry, and Pharmacological Activity of | 44e2ba32-8d84-4a29-b01f-d79e4340bb87 | 9783306 | Pharmacology[mh] | The genus Onosma comprises more than 230 species across the globe. The Asian continent has the highest share in terms of Onosma species existence , most of which are represented in Turkey by 88 species followed by countries such as Iran and China by 58 and 29 species , respectively. Iraqi Kurdistan represents 32 species of the genus Onosma based on the latest botanical studies . However, recent investigations have revealed seven new species of Onosma in Asian countries, particularly Iran . Continuous exploration on the ethnobotanical and plant taxonomy studies led to the discovery of several new Onosma species across our continent . Some Onosma species have been well studied pharmacologically than others, and the most common ones are shown in . The ethnobotanical and in vitro studies have revealed that most of this Onosma species has many medicinal capabilities such as sedatives , antioxidant , anti-inflammatory , gastric disorders , antithrombotic , wound healing , Alzheimer , enzyme inhibitory , anti-tumor , anti-viral , antifungal , and COVID-19 curatives . The phytochemical studies on the genus Onosma have reported several chemical compounds as their main active ingredients, including naphthaquinones (5,8-dihydroxy-2-(4-methylpent-3-enyl) naphthalene-1,4-dione) , Phenolics (ferulic acid, vanillic acid), flavonoids (apigenin, luteolin) , alkannin, and shikonines (deoxyshikonin, isobutyrylshikonin, α-methylbutyrylshikonin, acetylshikonin) . The past decades have showed numerous new records, phytochemical, and pharmacological studies of the new Onosma species, and the published two reviews were found lacking integrity as they contained incoherent data with skipping of some biological activities of the genus Onosma . Therefore, in order to provide theoretical reference for further research and to comprehensively understand the medicinal applications of this genus, this article systematically reviewed traditional uses, chemical constituents, pharmacological activities, and clinical applications of the Onosma species based on the published literature. The authors independently extracted systematic literature data search from seven electronic databases: Google Scholar, PubMed, Science Direct, Sci-Finder, Wiley Online Library, Web of Science, and Baidu Scholar. The scientific name “ Onosma ” was searched to cover all relevant information from April 1800–2022, including folkloric uses, phytochemical contents, and pharmacological potentials (antimicrobial, anti-inflammatory, anticancer, antioxidant, enzyme inhibitory, and antidiabetic) of the Onosma species are presented in this review. More than 1000 articles were detected with keyword Onosma , about 132 articles were found with keyword Onosma phytochemical, 187 articles were found with keyword Onosma pharmacology, biological activity, medicinal uses, pharmacology, and toxicology. Out of these, 125 articles were published detailing the isolation and properties of different phytochemical contents of the Onosma genus, and a total of 95 papers were selected based on the quality, specificity, and the procedure of the investigation of Onosma extracts and its isolated compounds. The folkloric names of most Onosma species in most Middle east countries is Gaozaban, an Urdu word. It was first referred to O. bracteatum Wall, and among Arabic populations it is known as “Lisan-al-Thawr” or “Saqil ul-Hammam”. Furthermore, its English popular name is “Vipers Bugloss”, while, in Hindi language, the Onosma species known as “Ratanjot” as first referred to O. echioides L. . Distribution map of some Onosma species collected from different regions of Iran, Iraq, and Turkey is shown in . The point inputs to the models developed in this study were collected from their habitats of Iran such as Fars, Lorestan, Khuzestan, Kermanshah, Hamedan, Markazi, Ilam, Kohgiluyeh, Kerman, and Boyer-Ahmad provinces, and their habitats of Iraq include mainly some areas of Kurdistan, Sulaimani, Hawraman, Rwanduz, and Amedia districts. Meanwhile, their habitats of Turkey includes Sirt, Hakary, Anatolia, and Van . Onosma Taxonomy Onosma species belongs to the Boraginaceae family. The Boraginaceae family contains more than 100 genera and over 200 species, which are classified into five subfamilies: Boraginoideae , Cordioideae , Ehretioideae , Hydrophylloideae , and Lennooideae . The folkloric use of many Onosma species as medicinal plants for different health problems by local ethnic groups in several countries such as Iraq, Turkey, Iran, China, and India roots back to hundreds of years ago. Almost all plant parts, such as leaves, roots, underground parts, flowers, and the whole plant of this genus species are reported to have a broad range of therapeutic potentials . Species such as Onosma alborosea have traditionally been utilized by Iraqi Kurdistan populations as a remedy for sedative, heart diseases, and kidney disorders through ingesting its aerial part extracts prepared by aqueous extraction methods . The aerial parts of Onosma orientalis has been macerated with hot water for treating sedatives by Kurdish nations living in Iraqi Kurdistan . Furthermore, the O. armeniacum K. has been used as Turkish folkloric medicine for healing wounds, peptic ulcers, burns, dyspnea, hoarseness, hemorrhoids, and abdominal pains through methods of cooking and filtration of its roots with butter . The extracts (oil and aqueous extracts) of O. argentatum and O. chlorotricum has been traditionally utilized in Turkey and Iran (Lorestan province) for the treatment of wounds and cutaneous injures . Furthermore, the root extracts of O. hispidum Wall. have been used traditionally by the Iranian nation as curatives for headache, wounds, insect stings, bits, and inflammatory diseases, while its flowers have been ingested for cardiovascular problems . Moreover, same species has been used as a dye and as a substitute for alkanet . The O. bracteatum Wall. extracts have been reported as traditional herbal medicine as a tonic agent for improving the body’s immune system with enhancing regulation of urine output . The O. bracteatum Wall. also has been used as remedy for asthma, respiratory problems, tonic, alterative, demulcent, diuretic, spasmolytic, rheumatoid arthritis, diuretic, and antileprotic in India, Nepal, Kashmir, and northwestern Himalayas countries . The root extracts of O. sericeum have been traditionally used in cream preparations for skin injuries and burn scar treatments in Adıyaman, Turkey . The O. microcarpum has traditional medicine record for the healing of wounds and burn scars by rural residents of Il’yca district, Erzurum, Turkey . The leaf aqueous extracts of O. echioides DC. are prepared for children suffering from constipation and metabolic disorders. Meanwhile, its flowers are reported as a cordial and as a stimulant for orthopedic and cardiac problems . The dried roots of O. paniculata have a traditional medicinal record in Chinese herbal medicine for curing several human diseases including tumors . The O. aucheriana is another species with traditional medicinal usage for itchiness, leucoderma, bronchitis, abdominal pain, strangury, fever, wounds, burns, and urinary calculi. Meanwhile, its flowers have been highlighted as stimulants and cardio-tonics, and its leaf extracts have been ingested as laxatives, purgatives, and as wound curatives . Out of more than 230 species of Onosma , only 12 species were reported in traditional medicines as herbal medicine until now. This could be due to the large geographical distribution of the Onosma species and lack of scientific interest in the past, but this number is expected to increase in upcoming years as the researchers extensively search and investigate for other Onosma species after discovering some interesting phytochemical and pharmacological potentials of this genus in recent years. The traditional names, country, ingested parts, and medicinal purposes of the genus Onosma are listed in . Onosma Species The current systematic review of the phytochemical contents of Onosma species presents major identified organic classes such as naphthoquinone (33), flavonoids (30), hydrocarbon (23), phenolic (22), ester (17), alkaloids (20), terpenoids (10), carboxylic acid (11), fatty acids (9), aromatics (12), and liganin (5) compounds as shown in . In addition, miscellaneous chemicals such as 24,25-Dihydroxycholecalciferol, 5-hydroxymethyl-furoic acid, and uplandicine also enrich the diversity of the phytochemistry in Onosma plants. Segregation of phytochemical contents in different classes is challenging and not always a clear and easy task. According to the current search, a total of 198 compounds are detected in the Onosma species as detailed in this review , and this will open up new future study opportunities to explore pharmacological potentials of those phytochemicals. Most common Onosma compounds reported were rosmarinic acid, apigenin, ferulic acid, protocatechuic acid, chlorogenic acid, caffeic acid, p-coumaric acid, vanillic acid, luteolin, hyperoside, hesperidin, apigenin-7-Glucoside, luteolin-7, glucoside, isovalerylshikonin, acetylshikonin, pinoresinol, deoxyshikonin, 4, hydroxybenzoic acid, β,β-dimethylacryl, isovalerylshikonin, 2,5-Dihydroxybenzoic, and 3-Hydroxybenzoic acid as presented in . The bioactive structures of identified and characterized representative compounds, which are based on the repetition across published studies are shown in , in addition to . Onosma Species 8.1. Toxicity In Vivo Experiment The chloroform and ethanolic extracts of O. aucheranum , O. isauricum O. sericeum , O. tauricum , and O. tauricum were safe in the administered doses from 100 mg/kg to 200 mg/kg based on the assessment of acute toxicity in the carrageenan-induced paw edema experiment as no abnormality in the morbidity nor mortality was recorded after 24 hours post treatment . Furthermore, the 100, 200, 300, and 600 mg/kg of the MeOH of O. mutabilis administration to rats showed no changes in the appearance, behavior, and feed intake of the rats in a 7-day experiment . Moreover, by the tarsal toxicity test, researchers have shown the acaricidal activity of the root extracts of O. visianii experimented against Tetranychus urticae mites in bean plants ( P. vulgaris var. Carmen) after 24 h (considered as acute toxicity), which caused significant mortality of T. urticae adults with lethal doses 83.2 and 112.6 μg·cm causing 50% (LD 50 ) and 90% (LD 90 ) inhibition of oviposition, respectively. However, at 5 days (considered as chronic toxicity) from the start of the test, the lethal dose LD 50 was more than 30 times lower (2.6 μg·cm −2 ) as a function of time used in the LD 50 calculation . Over the last two decades, several Onosma species have been tested for their toxicity to laboratory animal models. A study on toxicity of the bark extracts of O. echioides roots to Sprague Dawley rats (140 ± 10 g body weight) was performed and reported significant improvement in the body weight, food consumption, water intake, serum glucose, hematology, and biochemistry of rats with no adverse effect at a fixed dose . 8.2. Genotoxicity and Mutagenicity Through the Allium-test, significant genotoxic effect from aqueous extracts of O. stellulata roots and aerial parts were observed in mitosis at meristematic cells of onion. Although the aerial parts showed significant genotoxicity after 4-h treatment (mitotic index was 2, 79%, vs. 9, 18% for control), but the root aqueous extracts had higher genotoxic effects. Genotoxic effects included changes in the structure of chromosomes (conglutination, spirality), and cytotoxic reaction and certain differentiation in the cell cycle, which were found to be in correlation with duration of treatment and solution concentration . A genotoxic study by Allium anaphase–telophase assay reported that the safety of the ethanolic extract of O. aucheriana aerial parts at lower dose (62.5 mg/mL) had no toxic or genotoxic effects, while the higher dose (500 mg/mL) showed significantly the highest genotoxic effect including chromosomal aberrations, cells with multipolarity, cell bridges, and vagrant chromosomes (24.4%), cell fragments, and mitosis entrance . In vivo genotoxic study of methanolic extracts of O. sericea and O. stenoloba at different doses (25, 50, 100, 200, and 400 μg/mL) against EMS-induced DNA damage in the flies and larvae of the wild-type strain of Drosophila melanogaster showed the absence of genotoxic effect of O. sericea and O. stenoloba at concentration 80 mg/mL. Furthermore, significant antigenotoxic effects reported after dual treatment with 80 mg/mL of both plant extracts plus EMS (ethyl methane sulfonate) caused significant decrease in DNA damage (with over 80% reduction) . By using Ames assay, the antimutagenic potential of ethanolic extract of O. bracteatum has been reported against sodium azide and 2-aminofluorene mutagenicity in Salmonella typhimurium in TA100 strain (-S9 mix) as it displayed significant inhibition rate (82.30% at 250 mg/0.1 mL/plate), showing strong modulation of genotoxicity of base-pair substitution mutagen sodium azide when compared to NPD (frameshift mutagen) in TA98 tester strain. The O. bracteatum extracts showed significant antimutagenicity activity for preincubation mode than in co-incubation approach without -S9 in both TA100 and TA98 . The chloroform and ethanolic extracts of O. aucheranum , O. isauricum O. sericeum , O. tauricum , and O. tauricum were safe in the administered doses from 100 mg/kg to 200 mg/kg based on the assessment of acute toxicity in the carrageenan-induced paw edema experiment as no abnormality in the morbidity nor mortality was recorded after 24 hours post treatment . Furthermore, the 100, 200, 300, and 600 mg/kg of the MeOH of O. mutabilis administration to rats showed no changes in the appearance, behavior, and feed intake of the rats in a 7-day experiment . Moreover, by the tarsal toxicity test, researchers have shown the acaricidal activity of the root extracts of O. visianii experimented against Tetranychus urticae mites in bean plants ( P. vulgaris var. Carmen) after 24 h (considered as acute toxicity), which caused significant mortality of T. urticae adults with lethal doses 83.2 and 112.6 μg·cm causing 50% (LD 50 ) and 90% (LD 90 ) inhibition of oviposition, respectively. However, at 5 days (considered as chronic toxicity) from the start of the test, the lethal dose LD 50 was more than 30 times lower (2.6 μg·cm −2 ) as a function of time used in the LD 50 calculation . Over the last two decades, several Onosma species have been tested for their toxicity to laboratory animal models. A study on toxicity of the bark extracts of O. echioides roots to Sprague Dawley rats (140 ± 10 g body weight) was performed and reported significant improvement in the body weight, food consumption, water intake, serum glucose, hematology, and biochemistry of rats with no adverse effect at a fixed dose . Through the Allium-test, significant genotoxic effect from aqueous extracts of O. stellulata roots and aerial parts were observed in mitosis at meristematic cells of onion. Although the aerial parts showed significant genotoxicity after 4-h treatment (mitotic index was 2, 79%, vs. 9, 18% for control), but the root aqueous extracts had higher genotoxic effects. Genotoxic effects included changes in the structure of chromosomes (conglutination, spirality), and cytotoxic reaction and certain differentiation in the cell cycle, which were found to be in correlation with duration of treatment and solution concentration . A genotoxic study by Allium anaphase–telophase assay reported that the safety of the ethanolic extract of O. aucheriana aerial parts at lower dose (62.5 mg/mL) had no toxic or genotoxic effects, while the higher dose (500 mg/mL) showed significantly the highest genotoxic effect including chromosomal aberrations, cells with multipolarity, cell bridges, and vagrant chromosomes (24.4%), cell fragments, and mitosis entrance . In vivo genotoxic study of methanolic extracts of O. sericea and O. stenoloba at different doses (25, 50, 100, 200, and 400 μg/mL) against EMS-induced DNA damage in the flies and larvae of the wild-type strain of Drosophila melanogaster showed the absence of genotoxic effect of O. sericea and O. stenoloba at concentration 80 mg/mL. Furthermore, significant antigenotoxic effects reported after dual treatment with 80 mg/mL of both plant extracts plus EMS (ethyl methane sulfonate) caused significant decrease in DNA damage (with over 80% reduction) . By using Ames assay, the antimutagenic potential of ethanolic extract of O. bracteatum has been reported against sodium azide and 2-aminofluorene mutagenicity in Salmonella typhimurium in TA100 strain (-S9 mix) as it displayed significant inhibition rate (82.30% at 250 mg/0.1 mL/plate), showing strong modulation of genotoxicity of base-pair substitution mutagen sodium azide when compared to NPD (frameshift mutagen) in TA98 tester strain. The O. bracteatum extracts showed significant antimutagenicity activity for preincubation mode than in co-incubation approach without -S9 in both TA100 and TA98 . Onosma Species 9.1. Antibacterial Activity The essential oils isolated from roots of O. sieheana showed appreciable antibacterial activity against gram negative bacteria ( Escherichia coli (MIC: 125 μg/mL) and Pseudomonas aeruginosa (MIC: 125 μg/mL) and gram positive bacteria ( Staphylococcus aureus (MIC: 125 μg/mL) and Bacillus subtilis (MIC: 250 μg/mL)) . The n-hexane–dichloromethane mixture extracts of O. argentatum roots showed antibacterial activity against Bacillus subtilis , Escherichia coli , and Staphylococcus aureus with MIC values 28, 13, and 32 μg/mL, respectively . The chloroform fraction of O. khyberianum whole plant parts showed significant antibacterial activity against Salmonella typhi , Shigella dysenteriae , and Vibrio cholera inhibition zone 28, 26, 26 mm, respectively. Ethanol fraction of O. khyberianum demonstrated significant antiradical activity against Shigella dysenteriae (21 mm) and Vibrio cholera (20 mm), while the least active fraction of O. khyberianum n-hexane showed activity against Vibrio cholera , S. aureus , and Shigella dysenteriae (inhibition zone: 12, 9, 8 mm, respectively) but completely inactive against Salmonella and E. coli . The crude ethanolic extracts of O. hispidum roots showed significant antibacterial activity against several gram positive and gram negative bacteria ( Corynebacterium diphtheria , C. diphtheriticum , Micrococcus lysodiecticus , S. aureus , S. epidermidis , S. saprophyticus , Enterococcus faecalis , E. faecalis 2400 , E. faecium , Streptococcus pneumonia , and S. pyogenes ) with inhibition zone range between 18–20 mm . The isolated naphtshoquinones (deoxyshikonin, isobutyrylshikonin, α- methylbutyrylshikonin, acetylshikonin, β-hydroxyisovalerylshikonin, 5,8- O -dimethyl isobutyrylshikonin, and 5,8- O -dimethyl deoxyshikonin) from O. visanii roots showed significant antibacterial activity against gram negative bacteria ( Citrobacter koseri , Hafnia alvei , maltophilia , Yersinia intermedia , Ps. proteolytica , and Stenotrophomonas ) and gram positive bacteria ( Bacillus megaterium , Enterococcus faecalis , S. epidermidis , Microbacterium arborescens , and Micrococcus luteus ) with MIC 50 and MIC 90 values between range 4.27–68.27 μg/mL and 4.77–76.20 μg/mL, respectively . The antibacterial activity (MIC values) from methanol extract of aerial parts of O. sericea and O. stenoloba were between 2.5−10 mg/mL. Both Onosma extracts had moderate antibacterial activity only on a few strains, namely A. chroococcum and E. coli with MIC values 2.5 and 5 mg/L, respectively. O. sericea extract exhibited low activity on gram positive strain M. lysodeikticus with MIC 10 mg/mL, while O. stenoloba extract showed notable antibacterial action on E. faecalis and A. tumefaciens with MIC values 5 and 10 mg/mL, respectively . 9.2. Antifungal Activity Antifungal activity of methanolic extracts of O. sericea and O. stenoloba aerial parts against fungal strains Phialophare fastigiata and Fusarium oxysporum has been reported as 2.5 and 5 μg/mLof MIC, respectively. Furthermore, the methanol extracts of O. sericea exhibited moderate activity (MIC range of 2.5−5 μg/mL) on Penicillium canescens FSB 24 and P. cyclopium FSB 23, while O. stenoloba had antifungal activity only against P. cyclopium (MIC 10 μg/mL). Moreover, the same study showed antifungal potentials (MIC 10 μg/mL) of O. sericea against Trichoderma longibrachiatum FSB 13 and Trichoderma harzianum FSB 12. Meanwhile, increased concentration (10 μg/mL) of Onosma extracts showed inactivity against Aspergillus niger FSB 31, Aspergillus glaucus FSB 32, Doratomyces stemonitis FSB 41, Phialophora fastigiata FSB 81, Alternaria alternata FSB 51, and Fusarium oxysporum FSB 91 . The methanol extracts from aerial parts of O. griffithii exhibit antifungal activity against Aspergillus flavus (55%) and Fusarium solani (40%). Meanwhile, the chloroformic extracts showed better antifungal activity against A. flavus (59%) and Fusarium solani (60%) . The antifungal activity of O. kheberianum against three fungal strains, Fusarium oxysporum , Alternaria alternate , and A. flavus were reported as 18, 13, and 7 mm, respectively, for ethanol fractions and 17, 11, and 9 mm, respectively, for chloroform fractions . A previous study also showed a lack of antifungal activity of n-hexane–dichloromethane extracts of O. argentatum roots against Trichophyton tonsurans , Trichophyton interdigitale , Microphyton gypseum , and Candida albicans . The essential oils from O. sieheana Hayek roots showed significant antifungal activity against yeast strains Candida glabrata and C. albicans , and the authors linked this activity with their phytoconstituents, namely Monoterpenes, such as cymene and thymol . The essential oils from O. chlorotricum roots exhibit higher antifungal activity (21 and 19.3 mean of inhibition zones (mm) against C. albicans and C. glaberata , respectively) than that of essential oils from O. microcarpum roots . The O. paniculatum cells showed strong response to fungal elicitors from Aspergillus sp., in an attempt to accelerate shikonin derivative formation and inversely arrest plant cell growth, which resulted in a slight change in shikonin contents . 9.3. Antioxidant Activity Onosma species have been comprehensively studied and researchers have revealed that they are a promising resources of antioxidants using various types of extraction and solvent methods . The O. ambigens aerial part extracts exhibited notable antioxidant action in the phosphomolybdenum, CUPRAC, FRAP, DPPH, and ABTS assays with values of 1.65, 0.95, 0.52, 1.86, and 1.45 mg/mL, respectively . The antioxidant activity of O. gigantea were significant in phosphomolybdenum (134.31 μmol trolox (TEs)/g air dry matter (adm)), chelating effect (32.97 μmol (EDTAEs)/g adm), on DPPH (32.14 μmol TEs/g adm) and ABTS (58.68 μmol TEs/g adm)), and reducing power (CUPRAC (50.23 μmol TEs/g adm) and FRAP (40.96 μmol TEs/g adm)) assays . The water extract of the aerial part of O. pulchra showed significant antioxidant actions in DPPH, ABTS, CUPRAC, and ferrous ion chelating tests (3.90, 2.55, 2.20, and 1.23 mg/mL, respectively). Meanwhile, the Phosphomolybdenum and FRAP assays showed superiority of MeOH extract (1.98 and 1.02 mg/mL, respectively) . The ethanol extract of aerial parts of O. bracteatum showed significant radical quenching activity in superoxide radical scavenging (EC 50 : 115.14 μg/mL) and lipid peroxidation (EC 50 : 199.33 μg/mL) assays . The methanol extracts of O. mutabilis showed higher antioxidant activity than that of water and ethyl acetate fractions, respectively, in which the antioxidant values for methanol extracts were 1.45 ± 0.05, 3.54 ± 0.064, 2.33 ± 0.045, 1.12 ± 0.023, and 1.62 ± 0.079 mg/mL in phosphomolybdenum, DPPH scavenging, ABTS, FRAP, and CUPRAC reducing, respectively . The methanol extract of aerial parts of O. frutescens showed significantly higher antioxidant activity in DPPH (1.14 mg/mL), ABTS (1.04 mg/mL), CUPRAC (0.53 mg/mL), FRAP (0.35 mg/mL), and phosphomolybdenum (1.18 mg/mL) tests than that (1.75,1.50, 0.87, 0.55, 1.97 mg/mL) and (2.18, 1.87, 0.99, 0.63, 1.92 mg/mL) for O. sericea and O. aucheriana , respectively. The ferrous ion chelating assays showed superiority of O. aucheriana (IC 50 : 2.57 mg/mL) over O. frutescens (4.68 mg/mL) and O. sericea (6.18 mg/mL) . The aqueous extract of O. aucheriana roots showed significant antioxidant activity in radical quenching activity (ABTS, DPPH) with IC50 values as 9.89 and 17.73 μg/mL. Additionally, the same species showed notable lipid peroxidation inhibition, and hydroxyl radical scavenging actions with IC50 values, 23.41 and 31.09 μg/mL, respectively . The methanolic extracts of O. trapezuntea aerial parts showed stronger antioxidant activity (IC 50 : 3.05 mg/mL in DPPH and 7.19 mg/mL in ABTS) than that (IC 50 : 2.63 mg/mL in DPPH and 5.23 mg/mL in ABTS) of O. rigidum . The O. argentatum root extracts (0.1% concentration) by n-hexane–dichloromethane mixture (1:1) showed significant 98% antioxidant activity (IC 50 : 0.0076% w / v ) by thiobarbituric acid (TBA) . The methanol extract of aerial parts of O. lycaonica Hub. -Mor. exhibited stronger antioxidant activity in 1,1-diphenyl-2-picrylhydrazyl scavenging activity (2.69 ± 0.10 mg/mL), cupric reducing antioxidant power (1.10 ± 0.01), ferric reducing antioxidant power (0.69 ± 0.01 mg/mL), and ferrous ion chelating activity (2.32 ± 0.16 mg/mL) than that of O. papillosa . However, the O. papillosa showed lower IC 50 or EC 50 values for phosphomolybdenum (1.90 ± 0.07 mg/mL) when compared to O. lycaonica (2.05 ± 0.07 mg/mL), which could be related to their phytochemical contents as O. lycaonica had higher phenolic contents, with (43.5 ± 1.5 mg (gallic acid equivalent)/g extracts), whereas O. papillosa was higher in flavonoids (32.9 ± 0.3 mg (quercetin equivalent)/g extracts) . The aerial part ethanol extracts of O. hookeri showed the same 2,2-diphenyl-1-picrylhydrazyl (77.77 ± 1.44 μg/mL) scavenging activity as butylated hydroxy toluene (72.70 ± 1.04 μg/mL), but slightly weaker 2,2′-azino-bis-3-ethylbenzthiazoline-6-sulphonic acid (553.56 ± 2.78 μg/mL) scavenging activity and total antioxidant capacity than that of BHT (51.44 ± 1.37 μg/mL), while the ethyl acetate fraction of O. hookeri showed better ABTS scavenger, with IC 50 value of 84.83 ± 1.37 μg/mL . The aerial part MeOH extracts of O. sericea significant antioxidant activity in DPPH scavenging (130.23 ± 5.31 mg TE/g extract), ABTS scavenging (235.53 ± 4.62 mg TE/g extract), FRAP (215.65 ± 2.51 mg TE/g extract), CUPRAC (359.63 ± 14.83 mg TE/g extract), total antioxidant capacity (2.46 ± 0.35 mmol TE/g extract), metal chelating activity (24.65 ± 2.21 mgEDTAE/g extract), while O. stenoloba stronger activity with values 53.96 ± 0.78, 95.60 ± 2.30, 76.48 ± 3.26, 142.88 ± 1.49 mg TE/g, 1.16 ± 0.05 mmol TE/g, and 5.51 ± 0.81 mg EDTAE/g in the same essays, respectively . The aerial part extract of O. isauricum exhibited significant antioxidant actions with superiority of its methanol extracts in DPPH (34.75 mg/mL) and CUPRAC (0.643 mg/mL), ferric reducing powers (0.211 mg/mL), ABTS (188.68 mgTE/g extract), superoxide radical scavenging ability (97.50 mgTE/g extract), and total antioxidant ability (86.02 mgAAE/g extract) than that (31.44 mg/mL, 0.471 mg/mL, 0.237 mg/mL, 130.91 mgTE/g, 159.92 mgTE/g, 55.36 mgAAE/g) and (4.69 mg/mL, 0.078 mg/mL, 0.021 mg/mL, 131.94 mgTE/g, 103.23 mgTE/g, 31.17 mgAAE/g extract) for water and ethyl extracts, respectively . The results of antioxidant investigations of O. mollis showed significant radical scavenging actions phosphomolybdenum, DPPH, and ABTS, (2.01, 3.33, 2.30 mg/mL, respectively) while reducing power activity, CUPRAC and FRAP, were found as 1.48 and 0.79 mg/mL, respectively . 9.4. Cytotoxicity Activity For the past decades, several studies have confirmed the traditional usage of the Onosma species as cytotoxic agents, and mammalian cancer cell division was inhibited by its extracts and isolated compounds . The methanol extract of O. mutabilis aerial parts indicated significant anticancer activity against prostate (DU-145), mammary (MCF-7), and cervical cancer (Hep2c) cells with IC 50 values as 35.67 ± 0.15, 28.79 ± 0.23, and 41.83 ± 0.21 μg/mL, respectively . The crude extracts of O. aucheriana showed significant cytotoxicity activity against human rhabdomyosarcoma, human cervix carcinoma Hep2c, and from murine fibroblast (L2OB) cell lines with IC 50 values range between 25.54 to 50.57 μg/mL . The isolated compounds acetylshikonin, dimethylacrylshikonin, α-methylbutyrylshikonin, and isovalerylshikonin from the roots of O. paniculata showed appreciable anticancer activity against human CCRF-CEM leukemia, MDA-MB-231 breast cancer, human U251 glioblastoma, HCT 116 colon cancer, and human melanoma (SBcl2, WM35, WM9, WM164) cell lines with IC 50 values ranging between 600 nM to 70 μM . The isolated naphtshoquinones α-methylbutyrylshikonin and acetylshikonin compounds from O. visanii roots demonstrated stronger cytotoxic activity against MDAMB-231 cells (IC 50 : 86.0 μg/mL and 80.2 μg/mL, respectively) than that of 118.9, 204.6, 424.7, 391.6, and 411.5 μg/mL of Deoxyshikonin, β-Hydroxyisovalerylshikonin, Isobutyrylshikonin, 5,8- O -Dimethyl deoxyshikonin, and 5,8- O -Dimethyl isobutyrylshikonin, respectively. Additionally, all compounds except 5,8- O -Dimethyl deoxyshikonin, and 5,8- O -Dimethyl isobutyrylshikonin reduced viability of MDA-MB-231 cells after 48 h of incubation. Furthermore, α-methylbutyrylshikonin demonstrated the higher anticancer activity against HCT116 cells (IC 50 : 15.2 μg/mL) than that 97.8 μg/mL, 24.6 μg/mL and 30.9 μg/mL of Deoxyshikonin, Acetylshikonin, and β-Hydroxyisovalerylshikonin, respectively . The effect of Onosma bracteatum has been studied against different cancer cell lines and the results showed that various concentrations (0.055, 0.11, 0.22, 0.44, 0.88, 1.7, and 3.52 µg/mL) of O. bracteatum decreased viability of cells in a time- and dose-dependent protocol . Furthermore, the hydrochloric root extracts of O. dichroanthum Boiss. roots have shown significant anticancer actions against gastric cancer cells . Moreover, O. paniculata has shown notable cytotoxicity activity against a number of cancer lines and linked their action with its ability to accelerate apoptosis . The 50 µg/mL ethanolic extract from aerial parts of O. sericeum exhibited significant cytotoxicity activity against the breast cancer cells (MCF-7) with significantly decreased cell viability (28.76 ± 11.31%) . The petroleum ether and aqueous extracts of O. hispidum roots have shown significant anticancer actions against HepG2 liver cancer cell lines . 9.5. Enzyme Inhibitory Activity 9.5.1. Antidiabetic Activity A literature search revealed multiple research works that confirmed the anti-diabetics properties of Onosma species as the in vitro antidiabetic activity of Onosma species was reported based on its inhibitory potentials on α-amylase and glucosidase enzymes. The ethyl acetate extraction of aerial parts of O. gigantea showed higher α-amylase and glucosidase inhibitory activity (15.98 and 1.07 μmol/g) than that (410.50 and 6.75 μmol/g) and (1320.53 and 5.16μmol/g) of methanol and water extracts, respectively . The α-amylase inhibitory activity from MeOH extracts of O. aucheriana and O. sericea were reported higher (2.50 and 2.51 mg/mL, respectively) than that (3.15 mg/mL) of O. frutescens . The ethyl acetate extraction of O. ambigens aerial parts showed stronger α-amylase inhibitory activity (IC 50 : 2.64 mg/mL) than that (2.98 and 16.34 mg/mL) for methanol and water extracts, respectively . The methanol extracts of O. lycaonica and O. papillosa aerial parts exhibited significant α-amylase inhibitory concentration (IC 50 : 2.57 and 2.40 mg/mL) and glucosidase inhibition (IC 50 : 2.60 and 2.61 mg/mL), respectively . The ethyl acetate extract of O. pulchra aerial parts showed higher α-amylase inhibitory activity (2.40 mg/mL) than that (5.47 and 19.23 mg/mL) of methanol and water extracts, respectively . The aerial part extraction of O. rigidum showed higher glucosidase and lower α-amylase enzyme inhibitory activity than that of O. trapezuntea extracts . The MeOH aerial extracts of O. stenoloba exhibited higher α-amylase and lower glucosidase inhibitory activity (0.89 and 43.47 mmol/g) than that (1.26 and 33.38 mmol/g) of O. sericea , respectively . The hydroalcoholic extract of the aerial part of O. Dichroanthum was reported to have anti-diabetic and anti-neuropathy properties based on its ability to down regulation of the MDA and Glutathione levels in homogenized tissues of brain and liver in a rat experiment . The petroleum ether, chloroform, and methanol extracts of O. hispidum wall roots have shown significant anticancer actions with inhibitory percentages reported as 70, 58, and 50%, respectively. Meanwhile, the superiority of petroleum ether extracts has been linked with its higher polyphenolic contents . 9.5.2. Alzheimer’s Disease The protective effect of Onosma species against Alzheimer’s disease was reported depending on its inhibitory activity on acetylcholinesterase (AChE) and butyrylcholinesterase (BChE) enzymes. The ethyl acetate extraction of aerial parts of O. gigantea showed higher AChE and BChE inhibitory activity (2.76 and 6.87 μmol/g, respectively) than that (31.57 and 1.82 μmol/g, respectively) of methanol extracts . The isolated hispidone and (2S)-5,2-dihydroxy-7,5-dimethoxyflavanone from methanol extractions of whole plant parts of O. hispida showed significant inhibitory activity against AChE (11.6 and 15.7 mg/mL, respectively) and BChE (28.0 and 7.9 mg/mL, respectively) enzymes . The aerial part extracts of O. lycaonica and O. papillosa exhibited significant AChE inhibition activity (IC 50 :1.32 and increased BChE inhibitory activity (2.31 ± 0.04 and 2.07 ± 0.1 (2.31 ± 0.04 and 2.07 ± 0.08 mg GALAEs/g extracts), respectively . The MeOH extraction of O. rigidum aerial parts showed higher AChE and lower BChE inhibitory activity than that of O. trapezuntea extracts . O. sericea aerial part extracts showed higher inhibitory activity on AChE (3.74 mg/g) and BChE (0.51 mg/g) than that (4.34 and 3.44 mg/g) for O. stenoloba , respectively . 9.5.3. Anti-Tyrosinase Activity Tyrosinase enzymes are well-known for their participation in melanin biosynthesis, and hypersecretion accompanied by accumulation of melanin pigments may lead to hyperpigmentation disorders and photo carcinogenesis . The ethyl acetate partition of aerial parts of O. gigantea showed higher tyrosinase inhibitory activity (0.15 μmol/g) than that (0.49 and 10.48μmol/g) of methanol and water extracts, respectively . The tyrosinase inhibitory activity of methanol extracts of O. aucheriana aerial parts was higher (2.19 mg/mL) than that (2.23 and 2.40 mg/mL) of O. sericea and O. frutescence , respectively . The methanol partition of aerial parts of O. ambigens showed higher tyrosinase inhibitory activity (2.81 mg/mL) than that (3.79 and 4.45 mg/mL) of water and ethyl acetate extracts, respectively . Onosma lycaonica and O. papillosa aerial extracts have been reported as tyrosinase inhibitors with IC 50 values 2.20 and 2.05 mg/mL, respectively . The methanol extracts of O. pulchra aerial parts showed higher tyrosinase inhibitory activity (2.47 mg/mL) than that (3.77 and 4.35 mg/mL) of ethyl acetate and water extracts, respectively . The aerial part extracts of O. rigidum and O. trapezuntea showed comparable tyrosinase inhibitory potentials activity . A previous study also reported modest tyrosinase inhibitory activity (136.35 and 135.68 mg/g) for methanol extracts of aerial parts of O. sericea and O. stenoloba , respectively . The ethyl acetate extracts of O. isauricum showed higher tyrosinase inhibitory activity (19.96 mg/g kojic acid equivalents) than that (15.33 and 14.83 mg/g) of methanol and water extracts, respectively . 9.5.4. Anti-Lipoxygenases Activity Lipoxygenases enzymes are known to catalyze oxidation of polyunsaturated fatty acids (linoleic, linolenic, and arachidonic acid) yielding hydroperoxides. Such reactions may be favorable, but also lipoxygenases may interact undesirably. Aromatic compounds are major yields of lipoxygenase reactions that can interfere with food properties, mainly during long-term storage. Lipoxygenase’s impact on unsaturated fatty acids may lead to off-flavor/off-odor formation, leading to food spoilage. Furthermore, lipoxygenase is considered as an important enzyme in stimulation of inflammatory reactions in the human body by playing as a key factor in the biosynthesis of many bio-regulatory compounds such as hydroxyeicosatetraenoic acids (HETEs), leukotrienes, lipoxins, and hepoxylines that were linked to major diseases such as cancer, stroke, and heart and brain diseases . Therefore, searching for natural products that could target this enzyme has become a continuous scientific mission to prevent such diseases. The onosmins A (2-[(4-methylbenzyl)amino]benzoic acid and B (methyl 2-[(4-methylbenzyl)amino]benzoate) compounds isolated from the n-hexane-soluble fraction of ethanol extracts of O. hispida whole plant showed significant lipoxygenase inhibitory activity (IC 50 : 24.0 and 36.2 μM) . The essential oils isolated from roots of O. sieheana showed appreciable antibacterial activity against gram negative bacteria ( Escherichia coli (MIC: 125 μg/mL) and Pseudomonas aeruginosa (MIC: 125 μg/mL) and gram positive bacteria ( Staphylococcus aureus (MIC: 125 μg/mL) and Bacillus subtilis (MIC: 250 μg/mL)) . The n-hexane–dichloromethane mixture extracts of O. argentatum roots showed antibacterial activity against Bacillus subtilis , Escherichia coli , and Staphylococcus aureus with MIC values 28, 13, and 32 μg/mL, respectively . The chloroform fraction of O. khyberianum whole plant parts showed significant antibacterial activity against Salmonella typhi , Shigella dysenteriae , and Vibrio cholera inhibition zone 28, 26, 26 mm, respectively. Ethanol fraction of O. khyberianum demonstrated significant antiradical activity against Shigella dysenteriae (21 mm) and Vibrio cholera (20 mm), while the least active fraction of O. khyberianum n-hexane showed activity against Vibrio cholera , S. aureus , and Shigella dysenteriae (inhibition zone: 12, 9, 8 mm, respectively) but completely inactive against Salmonella and E. coli . The crude ethanolic extracts of O. hispidum roots showed significant antibacterial activity against several gram positive and gram negative bacteria ( Corynebacterium diphtheria , C. diphtheriticum , Micrococcus lysodiecticus , S. aureus , S. epidermidis , S. saprophyticus , Enterococcus faecalis , E. faecalis 2400 , E. faecium , Streptococcus pneumonia , and S. pyogenes ) with inhibition zone range between 18–20 mm . The isolated naphtshoquinones (deoxyshikonin, isobutyrylshikonin, α- methylbutyrylshikonin, acetylshikonin, β-hydroxyisovalerylshikonin, 5,8- O -dimethyl isobutyrylshikonin, and 5,8- O -dimethyl deoxyshikonin) from O. visanii roots showed significant antibacterial activity against gram negative bacteria ( Citrobacter koseri , Hafnia alvei , maltophilia , Yersinia intermedia , Ps. proteolytica , and Stenotrophomonas ) and gram positive bacteria ( Bacillus megaterium , Enterococcus faecalis , S. epidermidis , Microbacterium arborescens , and Micrococcus luteus ) with MIC 50 and MIC 90 values between range 4.27–68.27 μg/mL and 4.77–76.20 μg/mL, respectively . The antibacterial activity (MIC values) from methanol extract of aerial parts of O. sericea and O. stenoloba were between 2.5−10 mg/mL. Both Onosma extracts had moderate antibacterial activity only on a few strains, namely A. chroococcum and E. coli with MIC values 2.5 and 5 mg/L, respectively. O. sericea extract exhibited low activity on gram positive strain M. lysodeikticus with MIC 10 mg/mL, while O. stenoloba extract showed notable antibacterial action on E. faecalis and A. tumefaciens with MIC values 5 and 10 mg/mL, respectively . Antifungal activity of methanolic extracts of O. sericea and O. stenoloba aerial parts against fungal strains Phialophare fastigiata and Fusarium oxysporum has been reported as 2.5 and 5 μg/mLof MIC, respectively. Furthermore, the methanol extracts of O. sericea exhibited moderate activity (MIC range of 2.5−5 μg/mL) on Penicillium canescens FSB 24 and P. cyclopium FSB 23, while O. stenoloba had antifungal activity only against P. cyclopium (MIC 10 μg/mL). Moreover, the same study showed antifungal potentials (MIC 10 μg/mL) of O. sericea against Trichoderma longibrachiatum FSB 13 and Trichoderma harzianum FSB 12. Meanwhile, increased concentration (10 μg/mL) of Onosma extracts showed inactivity against Aspergillus niger FSB 31, Aspergillus glaucus FSB 32, Doratomyces stemonitis FSB 41, Phialophora fastigiata FSB 81, Alternaria alternata FSB 51, and Fusarium oxysporum FSB 91 . The methanol extracts from aerial parts of O. griffithii exhibit antifungal activity against Aspergillus flavus (55%) and Fusarium solani (40%). Meanwhile, the chloroformic extracts showed better antifungal activity against A. flavus (59%) and Fusarium solani (60%) . The antifungal activity of O. kheberianum against three fungal strains, Fusarium oxysporum , Alternaria alternate , and A. flavus were reported as 18, 13, and 7 mm, respectively, for ethanol fractions and 17, 11, and 9 mm, respectively, for chloroform fractions . A previous study also showed a lack of antifungal activity of n-hexane–dichloromethane extracts of O. argentatum roots against Trichophyton tonsurans , Trichophyton interdigitale , Microphyton gypseum , and Candida albicans . The essential oils from O. sieheana Hayek roots showed significant antifungal activity against yeast strains Candida glabrata and C. albicans , and the authors linked this activity with their phytoconstituents, namely Monoterpenes, such as cymene and thymol . The essential oils from O. chlorotricum roots exhibit higher antifungal activity (21 and 19.3 mean of inhibition zones (mm) against C. albicans and C. glaberata , respectively) than that of essential oils from O. microcarpum roots . The O. paniculatum cells showed strong response to fungal elicitors from Aspergillus sp., in an attempt to accelerate shikonin derivative formation and inversely arrest plant cell growth, which resulted in a slight change in shikonin contents . Onosma species have been comprehensively studied and researchers have revealed that they are a promising resources of antioxidants using various types of extraction and solvent methods . The O. ambigens aerial part extracts exhibited notable antioxidant action in the phosphomolybdenum, CUPRAC, FRAP, DPPH, and ABTS assays with values of 1.65, 0.95, 0.52, 1.86, and 1.45 mg/mL, respectively . The antioxidant activity of O. gigantea were significant in phosphomolybdenum (134.31 μmol trolox (TEs)/g air dry matter (adm)), chelating effect (32.97 μmol (EDTAEs)/g adm), on DPPH (32.14 μmol TEs/g adm) and ABTS (58.68 μmol TEs/g adm)), and reducing power (CUPRAC (50.23 μmol TEs/g adm) and FRAP (40.96 μmol TEs/g adm)) assays . The water extract of the aerial part of O. pulchra showed significant antioxidant actions in DPPH, ABTS, CUPRAC, and ferrous ion chelating tests (3.90, 2.55, 2.20, and 1.23 mg/mL, respectively). Meanwhile, the Phosphomolybdenum and FRAP assays showed superiority of MeOH extract (1.98 and 1.02 mg/mL, respectively) . The ethanol extract of aerial parts of O. bracteatum showed significant radical quenching activity in superoxide radical scavenging (EC 50 : 115.14 μg/mL) and lipid peroxidation (EC 50 : 199.33 μg/mL) assays . The methanol extracts of O. mutabilis showed higher antioxidant activity than that of water and ethyl acetate fractions, respectively, in which the antioxidant values for methanol extracts were 1.45 ± 0.05, 3.54 ± 0.064, 2.33 ± 0.045, 1.12 ± 0.023, and 1.62 ± 0.079 mg/mL in phosphomolybdenum, DPPH scavenging, ABTS, FRAP, and CUPRAC reducing, respectively . The methanol extract of aerial parts of O. frutescens showed significantly higher antioxidant activity in DPPH (1.14 mg/mL), ABTS (1.04 mg/mL), CUPRAC (0.53 mg/mL), FRAP (0.35 mg/mL), and phosphomolybdenum (1.18 mg/mL) tests than that (1.75,1.50, 0.87, 0.55, 1.97 mg/mL) and (2.18, 1.87, 0.99, 0.63, 1.92 mg/mL) for O. sericea and O. aucheriana , respectively. The ferrous ion chelating assays showed superiority of O. aucheriana (IC 50 : 2.57 mg/mL) over O. frutescens (4.68 mg/mL) and O. sericea (6.18 mg/mL) . The aqueous extract of O. aucheriana roots showed significant antioxidant activity in radical quenching activity (ABTS, DPPH) with IC50 values as 9.89 and 17.73 μg/mL. Additionally, the same species showed notable lipid peroxidation inhibition, and hydroxyl radical scavenging actions with IC50 values, 23.41 and 31.09 μg/mL, respectively . The methanolic extracts of O. trapezuntea aerial parts showed stronger antioxidant activity (IC 50 : 3.05 mg/mL in DPPH and 7.19 mg/mL in ABTS) than that (IC 50 : 2.63 mg/mL in DPPH and 5.23 mg/mL in ABTS) of O. rigidum . The O. argentatum root extracts (0.1% concentration) by n-hexane–dichloromethane mixture (1:1) showed significant 98% antioxidant activity (IC 50 : 0.0076% w / v ) by thiobarbituric acid (TBA) . The methanol extract of aerial parts of O. lycaonica Hub. -Mor. exhibited stronger antioxidant activity in 1,1-diphenyl-2-picrylhydrazyl scavenging activity (2.69 ± 0.10 mg/mL), cupric reducing antioxidant power (1.10 ± 0.01), ferric reducing antioxidant power (0.69 ± 0.01 mg/mL), and ferrous ion chelating activity (2.32 ± 0.16 mg/mL) than that of O. papillosa . However, the O. papillosa showed lower IC 50 or EC 50 values for phosphomolybdenum (1.90 ± 0.07 mg/mL) when compared to O. lycaonica (2.05 ± 0.07 mg/mL), which could be related to their phytochemical contents as O. lycaonica had higher phenolic contents, with (43.5 ± 1.5 mg (gallic acid equivalent)/g extracts), whereas O. papillosa was higher in flavonoids (32.9 ± 0.3 mg (quercetin equivalent)/g extracts) . The aerial part ethanol extracts of O. hookeri showed the same 2,2-diphenyl-1-picrylhydrazyl (77.77 ± 1.44 μg/mL) scavenging activity as butylated hydroxy toluene (72.70 ± 1.04 μg/mL), but slightly weaker 2,2′-azino-bis-3-ethylbenzthiazoline-6-sulphonic acid (553.56 ± 2.78 μg/mL) scavenging activity and total antioxidant capacity than that of BHT (51.44 ± 1.37 μg/mL), while the ethyl acetate fraction of O. hookeri showed better ABTS scavenger, with IC 50 value of 84.83 ± 1.37 μg/mL . The aerial part MeOH extracts of O. sericea significant antioxidant activity in DPPH scavenging (130.23 ± 5.31 mg TE/g extract), ABTS scavenging (235.53 ± 4.62 mg TE/g extract), FRAP (215.65 ± 2.51 mg TE/g extract), CUPRAC (359.63 ± 14.83 mg TE/g extract), total antioxidant capacity (2.46 ± 0.35 mmol TE/g extract), metal chelating activity (24.65 ± 2.21 mgEDTAE/g extract), while O. stenoloba stronger activity with values 53.96 ± 0.78, 95.60 ± 2.30, 76.48 ± 3.26, 142.88 ± 1.49 mg TE/g, 1.16 ± 0.05 mmol TE/g, and 5.51 ± 0.81 mg EDTAE/g in the same essays, respectively . The aerial part extract of O. isauricum exhibited significant antioxidant actions with superiority of its methanol extracts in DPPH (34.75 mg/mL) and CUPRAC (0.643 mg/mL), ferric reducing powers (0.211 mg/mL), ABTS (188.68 mgTE/g extract), superoxide radical scavenging ability (97.50 mgTE/g extract), and total antioxidant ability (86.02 mgAAE/g extract) than that (31.44 mg/mL, 0.471 mg/mL, 0.237 mg/mL, 130.91 mgTE/g, 159.92 mgTE/g, 55.36 mgAAE/g) and (4.69 mg/mL, 0.078 mg/mL, 0.021 mg/mL, 131.94 mgTE/g, 103.23 mgTE/g, 31.17 mgAAE/g extract) for water and ethyl extracts, respectively . The results of antioxidant investigations of O. mollis showed significant radical scavenging actions phosphomolybdenum, DPPH, and ABTS, (2.01, 3.33, 2.30 mg/mL, respectively) while reducing power activity, CUPRAC and FRAP, were found as 1.48 and 0.79 mg/mL, respectively . For the past decades, several studies have confirmed the traditional usage of the Onosma species as cytotoxic agents, and mammalian cancer cell division was inhibited by its extracts and isolated compounds . The methanol extract of O. mutabilis aerial parts indicated significant anticancer activity against prostate (DU-145), mammary (MCF-7), and cervical cancer (Hep2c) cells with IC 50 values as 35.67 ± 0.15, 28.79 ± 0.23, and 41.83 ± 0.21 μg/mL, respectively . The crude extracts of O. aucheriana showed significant cytotoxicity activity against human rhabdomyosarcoma, human cervix carcinoma Hep2c, and from murine fibroblast (L2OB) cell lines with IC 50 values range between 25.54 to 50.57 μg/mL . The isolated compounds acetylshikonin, dimethylacrylshikonin, α-methylbutyrylshikonin, and isovalerylshikonin from the roots of O. paniculata showed appreciable anticancer activity against human CCRF-CEM leukemia, MDA-MB-231 breast cancer, human U251 glioblastoma, HCT 116 colon cancer, and human melanoma (SBcl2, WM35, WM9, WM164) cell lines with IC 50 values ranging between 600 nM to 70 μM . The isolated naphtshoquinones α-methylbutyrylshikonin and acetylshikonin compounds from O. visanii roots demonstrated stronger cytotoxic activity against MDAMB-231 cells (IC 50 : 86.0 μg/mL and 80.2 μg/mL, respectively) than that of 118.9, 204.6, 424.7, 391.6, and 411.5 μg/mL of Deoxyshikonin, β-Hydroxyisovalerylshikonin, Isobutyrylshikonin, 5,8- O -Dimethyl deoxyshikonin, and 5,8- O -Dimethyl isobutyrylshikonin, respectively. Additionally, all compounds except 5,8- O -Dimethyl deoxyshikonin, and 5,8- O -Dimethyl isobutyrylshikonin reduced viability of MDA-MB-231 cells after 48 h of incubation. Furthermore, α-methylbutyrylshikonin demonstrated the higher anticancer activity against HCT116 cells (IC 50 : 15.2 μg/mL) than that 97.8 μg/mL, 24.6 μg/mL and 30.9 μg/mL of Deoxyshikonin, Acetylshikonin, and β-Hydroxyisovalerylshikonin, respectively . The effect of Onosma bracteatum has been studied against different cancer cell lines and the results showed that various concentrations (0.055, 0.11, 0.22, 0.44, 0.88, 1.7, and 3.52 µg/mL) of O. bracteatum decreased viability of cells in a time- and dose-dependent protocol . Furthermore, the hydrochloric root extracts of O. dichroanthum Boiss. roots have shown significant anticancer actions against gastric cancer cells . Moreover, O. paniculata has shown notable cytotoxicity activity against a number of cancer lines and linked their action with its ability to accelerate apoptosis . The 50 µg/mL ethanolic extract from aerial parts of O. sericeum exhibited significant cytotoxicity activity against the breast cancer cells (MCF-7) with significantly decreased cell viability (28.76 ± 11.31%) . The petroleum ether and aqueous extracts of O. hispidum roots have shown significant anticancer actions against HepG2 liver cancer cell lines . 9.5.1. Antidiabetic Activity A literature search revealed multiple research works that confirmed the anti-diabetics properties of Onosma species as the in vitro antidiabetic activity of Onosma species was reported based on its inhibitory potentials on α-amylase and glucosidase enzymes. The ethyl acetate extraction of aerial parts of O. gigantea showed higher α-amylase and glucosidase inhibitory activity (15.98 and 1.07 μmol/g) than that (410.50 and 6.75 μmol/g) and (1320.53 and 5.16μmol/g) of methanol and water extracts, respectively . The α-amylase inhibitory activity from MeOH extracts of O. aucheriana and O. sericea were reported higher (2.50 and 2.51 mg/mL, respectively) than that (3.15 mg/mL) of O. frutescens . The ethyl acetate extraction of O. ambigens aerial parts showed stronger α-amylase inhibitory activity (IC 50 : 2.64 mg/mL) than that (2.98 and 16.34 mg/mL) for methanol and water extracts, respectively . The methanol extracts of O. lycaonica and O. papillosa aerial parts exhibited significant α-amylase inhibitory concentration (IC 50 : 2.57 and 2.40 mg/mL) and glucosidase inhibition (IC 50 : 2.60 and 2.61 mg/mL), respectively . The ethyl acetate extract of O. pulchra aerial parts showed higher α-amylase inhibitory activity (2.40 mg/mL) than that (5.47 and 19.23 mg/mL) of methanol and water extracts, respectively . The aerial part extraction of O. rigidum showed higher glucosidase and lower α-amylase enzyme inhibitory activity than that of O. trapezuntea extracts . The MeOH aerial extracts of O. stenoloba exhibited higher α-amylase and lower glucosidase inhibitory activity (0.89 and 43.47 mmol/g) than that (1.26 and 33.38 mmol/g) of O. sericea , respectively . The hydroalcoholic extract of the aerial part of O. Dichroanthum was reported to have anti-diabetic and anti-neuropathy properties based on its ability to down regulation of the MDA and Glutathione levels in homogenized tissues of brain and liver in a rat experiment . The petroleum ether, chloroform, and methanol extracts of O. hispidum wall roots have shown significant anticancer actions with inhibitory percentages reported as 70, 58, and 50%, respectively. Meanwhile, the superiority of petroleum ether extracts has been linked with its higher polyphenolic contents . 9.5.2. Alzheimer’s Disease The protective effect of Onosma species against Alzheimer’s disease was reported depending on its inhibitory activity on acetylcholinesterase (AChE) and butyrylcholinesterase (BChE) enzymes. The ethyl acetate extraction of aerial parts of O. gigantea showed higher AChE and BChE inhibitory activity (2.76 and 6.87 μmol/g, respectively) than that (31.57 and 1.82 μmol/g, respectively) of methanol extracts . The isolated hispidone and (2S)-5,2-dihydroxy-7,5-dimethoxyflavanone from methanol extractions of whole plant parts of O. hispida showed significant inhibitory activity against AChE (11.6 and 15.7 mg/mL, respectively) and BChE (28.0 and 7.9 mg/mL, respectively) enzymes . The aerial part extracts of O. lycaonica and O. papillosa exhibited significant AChE inhibition activity (IC 50 :1.32 and increased BChE inhibitory activity (2.31 ± 0.04 and 2.07 ± 0.1 (2.31 ± 0.04 and 2.07 ± 0.08 mg GALAEs/g extracts), respectively . The MeOH extraction of O. rigidum aerial parts showed higher AChE and lower BChE inhibitory activity than that of O. trapezuntea extracts . O. sericea aerial part extracts showed higher inhibitory activity on AChE (3.74 mg/g) and BChE (0.51 mg/g) than that (4.34 and 3.44 mg/g) for O. stenoloba , respectively . 9.5.3. Anti-Tyrosinase Activity Tyrosinase enzymes are well-known for their participation in melanin biosynthesis, and hypersecretion accompanied by accumulation of melanin pigments may lead to hyperpigmentation disorders and photo carcinogenesis . The ethyl acetate partition of aerial parts of O. gigantea showed higher tyrosinase inhibitory activity (0.15 μmol/g) than that (0.49 and 10.48μmol/g) of methanol and water extracts, respectively . The tyrosinase inhibitory activity of methanol extracts of O. aucheriana aerial parts was higher (2.19 mg/mL) than that (2.23 and 2.40 mg/mL) of O. sericea and O. frutescence , respectively . The methanol partition of aerial parts of O. ambigens showed higher tyrosinase inhibitory activity (2.81 mg/mL) than that (3.79 and 4.45 mg/mL) of water and ethyl acetate extracts, respectively . Onosma lycaonica and O. papillosa aerial extracts have been reported as tyrosinase inhibitors with IC 50 values 2.20 and 2.05 mg/mL, respectively . The methanol extracts of O. pulchra aerial parts showed higher tyrosinase inhibitory activity (2.47 mg/mL) than that (3.77 and 4.35 mg/mL) of ethyl acetate and water extracts, respectively . The aerial part extracts of O. rigidum and O. trapezuntea showed comparable tyrosinase inhibitory potentials activity . A previous study also reported modest tyrosinase inhibitory activity (136.35 and 135.68 mg/g) for methanol extracts of aerial parts of O. sericea and O. stenoloba , respectively . The ethyl acetate extracts of O. isauricum showed higher tyrosinase inhibitory activity (19.96 mg/g kojic acid equivalents) than that (15.33 and 14.83 mg/g) of methanol and water extracts, respectively . 9.5.4. Anti-Lipoxygenases Activity Lipoxygenases enzymes are known to catalyze oxidation of polyunsaturated fatty acids (linoleic, linolenic, and arachidonic acid) yielding hydroperoxides. Such reactions may be favorable, but also lipoxygenases may interact undesirably. Aromatic compounds are major yields of lipoxygenase reactions that can interfere with food properties, mainly during long-term storage. Lipoxygenase’s impact on unsaturated fatty acids may lead to off-flavor/off-odor formation, leading to food spoilage. Furthermore, lipoxygenase is considered as an important enzyme in stimulation of inflammatory reactions in the human body by playing as a key factor in the biosynthesis of many bio-regulatory compounds such as hydroxyeicosatetraenoic acids (HETEs), leukotrienes, lipoxins, and hepoxylines that were linked to major diseases such as cancer, stroke, and heart and brain diseases . Therefore, searching for natural products that could target this enzyme has become a continuous scientific mission to prevent such diseases. The onosmins A (2-[(4-methylbenzyl)amino]benzoic acid and B (methyl 2-[(4-methylbenzyl)amino]benzoate) compounds isolated from the n-hexane-soluble fraction of ethanol extracts of O. hispida whole plant showed significant lipoxygenase inhibitory activity (IC 50 : 24.0 and 36.2 μM) . A literature search revealed multiple research works that confirmed the anti-diabetics properties of Onosma species as the in vitro antidiabetic activity of Onosma species was reported based on its inhibitory potentials on α-amylase and glucosidase enzymes. The ethyl acetate extraction of aerial parts of O. gigantea showed higher α-amylase and glucosidase inhibitory activity (15.98 and 1.07 μmol/g) than that (410.50 and 6.75 μmol/g) and (1320.53 and 5.16μmol/g) of methanol and water extracts, respectively . The α-amylase inhibitory activity from MeOH extracts of O. aucheriana and O. sericea were reported higher (2.50 and 2.51 mg/mL, respectively) than that (3.15 mg/mL) of O. frutescens . The ethyl acetate extraction of O. ambigens aerial parts showed stronger α-amylase inhibitory activity (IC 50 : 2.64 mg/mL) than that (2.98 and 16.34 mg/mL) for methanol and water extracts, respectively . The methanol extracts of O. lycaonica and O. papillosa aerial parts exhibited significant α-amylase inhibitory concentration (IC 50 : 2.57 and 2.40 mg/mL) and glucosidase inhibition (IC 50 : 2.60 and 2.61 mg/mL), respectively . The ethyl acetate extract of O. pulchra aerial parts showed higher α-amylase inhibitory activity (2.40 mg/mL) than that (5.47 and 19.23 mg/mL) of methanol and water extracts, respectively . The aerial part extraction of O. rigidum showed higher glucosidase and lower α-amylase enzyme inhibitory activity than that of O. trapezuntea extracts . The MeOH aerial extracts of O. stenoloba exhibited higher α-amylase and lower glucosidase inhibitory activity (0.89 and 43.47 mmol/g) than that (1.26 and 33.38 mmol/g) of O. sericea , respectively . The hydroalcoholic extract of the aerial part of O. Dichroanthum was reported to have anti-diabetic and anti-neuropathy properties based on its ability to down regulation of the MDA and Glutathione levels in homogenized tissues of brain and liver in a rat experiment . The petroleum ether, chloroform, and methanol extracts of O. hispidum wall roots have shown significant anticancer actions with inhibitory percentages reported as 70, 58, and 50%, respectively. Meanwhile, the superiority of petroleum ether extracts has been linked with its higher polyphenolic contents . The protective effect of Onosma species against Alzheimer’s disease was reported depending on its inhibitory activity on acetylcholinesterase (AChE) and butyrylcholinesterase (BChE) enzymes. The ethyl acetate extraction of aerial parts of O. gigantea showed higher AChE and BChE inhibitory activity (2.76 and 6.87 μmol/g, respectively) than that (31.57 and 1.82 μmol/g, respectively) of methanol extracts . The isolated hispidone and (2S)-5,2-dihydroxy-7,5-dimethoxyflavanone from methanol extractions of whole plant parts of O. hispida showed significant inhibitory activity against AChE (11.6 and 15.7 mg/mL, respectively) and BChE (28.0 and 7.9 mg/mL, respectively) enzymes . The aerial part extracts of O. lycaonica and O. papillosa exhibited significant AChE inhibition activity (IC 50 :1.32 and increased BChE inhibitory activity (2.31 ± 0.04 and 2.07 ± 0.1 (2.31 ± 0.04 and 2.07 ± 0.08 mg GALAEs/g extracts), respectively . The MeOH extraction of O. rigidum aerial parts showed higher AChE and lower BChE inhibitory activity than that of O. trapezuntea extracts . O. sericea aerial part extracts showed higher inhibitory activity on AChE (3.74 mg/g) and BChE (0.51 mg/g) than that (4.34 and 3.44 mg/g) for O. stenoloba , respectively . Tyrosinase enzymes are well-known for their participation in melanin biosynthesis, and hypersecretion accompanied by accumulation of melanin pigments may lead to hyperpigmentation disorders and photo carcinogenesis . The ethyl acetate partition of aerial parts of O. gigantea showed higher tyrosinase inhibitory activity (0.15 μmol/g) than that (0.49 and 10.48μmol/g) of methanol and water extracts, respectively . The tyrosinase inhibitory activity of methanol extracts of O. aucheriana aerial parts was higher (2.19 mg/mL) than that (2.23 and 2.40 mg/mL) of O. sericea and O. frutescence , respectively . The methanol partition of aerial parts of O. ambigens showed higher tyrosinase inhibitory activity (2.81 mg/mL) than that (3.79 and 4.45 mg/mL) of water and ethyl acetate extracts, respectively . Onosma lycaonica and O. papillosa aerial extracts have been reported as tyrosinase inhibitors with IC 50 values 2.20 and 2.05 mg/mL, respectively . The methanol extracts of O. pulchra aerial parts showed higher tyrosinase inhibitory activity (2.47 mg/mL) than that (3.77 and 4.35 mg/mL) of ethyl acetate and water extracts, respectively . The aerial part extracts of O. rigidum and O. trapezuntea showed comparable tyrosinase inhibitory potentials activity . A previous study also reported modest tyrosinase inhibitory activity (136.35 and 135.68 mg/g) for methanol extracts of aerial parts of O. sericea and O. stenoloba , respectively . The ethyl acetate extracts of O. isauricum showed higher tyrosinase inhibitory activity (19.96 mg/g kojic acid equivalents) than that (15.33 and 14.83 mg/g) of methanol and water extracts, respectively . Lipoxygenases enzymes are known to catalyze oxidation of polyunsaturated fatty acids (linoleic, linolenic, and arachidonic acid) yielding hydroperoxides. Such reactions may be favorable, but also lipoxygenases may interact undesirably. Aromatic compounds are major yields of lipoxygenase reactions that can interfere with food properties, mainly during long-term storage. Lipoxygenase’s impact on unsaturated fatty acids may lead to off-flavor/off-odor formation, leading to food spoilage. Furthermore, lipoxygenase is considered as an important enzyme in stimulation of inflammatory reactions in the human body by playing as a key factor in the biosynthesis of many bio-regulatory compounds such as hydroxyeicosatetraenoic acids (HETEs), leukotrienes, lipoxins, and hepoxylines that were linked to major diseases such as cancer, stroke, and heart and brain diseases . Therefore, searching for natural products that could target this enzyme has become a continuous scientific mission to prevent such diseases. The onosmins A (2-[(4-methylbenzyl)amino]benzoic acid and B (methyl 2-[(4-methylbenzyl)amino]benzoate) compounds isolated from the n-hexane-soluble fraction of ethanol extracts of O. hispida whole plant showed significant lipoxygenase inhibitory activity (IC 50 : 24.0 and 36.2 μM) . 10.1. Parasiticidal Activity The antileishmanial activities of the crude methanol extract of O. griffithii and its fractions were statistically significant ( p < 0.05) against the Leishmania promastigotes , Pakistani isolates in comparison with the standard drug called Pentamidine . 10.2. Anti-Inflammatory and Analgesic Activity The chloroform extracts from roots of O. aucheranum , O. isauricum , and O. tauricum showed 28.0%, 34.3%, and 15.6% inhibitory action in p-benzoquinone-induced abdominal constriction experiment, while the ethanol extracts of O. isauricum and O. sericeum demonstrated inhibition action of 24.6% and 27.5%, respectively, in the same test. The chloroform and ethanol extracts of O. isauricum and ethanol extract of O. sericeum also showed significant inhibitory activity, ranging between 12.3–27.3%, 10.5–25.3%, 8.2–22.6%, respectively, in a carrageenan-induced hind paw edema model at 100 mg/kg dose without gastric damage, and the activity was very comparable to indomethacin (32.0–38.4% inhibition) as a standard sample . The chloroform extracts of O. aucheranum and O. isauricum and ethanolic extracts of O. isauricum and O. sericeum exhibited notable antinociceptive activity; 28.0%, 34.3%, 24.6%, and 27.5% inhibition, respectively, against p-benzoquinone-induced abdominal contractions, without induction of any sign of gastric lesion . The methanol extraction of aerial parts of O. bracteatum showed potent analgesic activity by inducing significant increase in the latency period in a dose-dependent manner at different doses at 1, 2, and 3 h (with superiority of 500 mg/kg i.e., 258.9% ( p < 0.05) at 3 h) post feeding, respectively, in a tail flick test. Furthermore, the methanol extract of O. bracteatum showed significant analgesic effect at 500 mg/kg body weight dose by inducing 54% inhibition ( p < 0.05) in comparison to 45.9% inhibition activity for standard Diclofenac sodium (5 mg/kg body weight) . 10.3. Gastric-Ulcerogenic Activity The chloroform and ethanol extracts from O. aucheranum , O. isauricum , O. sericeum , and O. tauricum roots did not cause any gastric lesions or bleeding in the stomach of mice in a 48-h experiment . 10.4. Treatment and Prevention of COVID-19 The Onosma phytochemicals, deoxyshikonin, 3-hydroxy-isovaleryl shikonin, propionyl shikonin, and acetyl shikonin showed significant binding affinities for the Mpro enzyme based on the molecular docking studies using two distinct approaches, in which a SiteMap module of Maestro was used to detect the possible ligand binding sites for the Mpro enzyme. Docking simulations and molecular mechanics suggest that shikonin derivatives might be effective anti-SARS-CoV-2 compounds . The antileishmanial activities of the crude methanol extract of O. griffithii and its fractions were statistically significant ( p < 0.05) against the Leishmania promastigotes , Pakistani isolates in comparison with the standard drug called Pentamidine . The chloroform extracts from roots of O. aucheranum , O. isauricum , and O. tauricum showed 28.0%, 34.3%, and 15.6% inhibitory action in p-benzoquinone-induced abdominal constriction experiment, while the ethanol extracts of O. isauricum and O. sericeum demonstrated inhibition action of 24.6% and 27.5%, respectively, in the same test. The chloroform and ethanol extracts of O. isauricum and ethanol extract of O. sericeum also showed significant inhibitory activity, ranging between 12.3–27.3%, 10.5–25.3%, 8.2–22.6%, respectively, in a carrageenan-induced hind paw edema model at 100 mg/kg dose without gastric damage, and the activity was very comparable to indomethacin (32.0–38.4% inhibition) as a standard sample . The chloroform extracts of O. aucheranum and O. isauricum and ethanolic extracts of O. isauricum and O. sericeum exhibited notable antinociceptive activity; 28.0%, 34.3%, 24.6%, and 27.5% inhibition, respectively, against p-benzoquinone-induced abdominal contractions, without induction of any sign of gastric lesion . The methanol extraction of aerial parts of O. bracteatum showed potent analgesic activity by inducing significant increase in the latency period in a dose-dependent manner at different doses at 1, 2, and 3 h (with superiority of 500 mg/kg i.e., 258.9% ( p < 0.05) at 3 h) post feeding, respectively, in a tail flick test. Furthermore, the methanol extract of O. bracteatum showed significant analgesic effect at 500 mg/kg body weight dose by inducing 54% inhibition ( p < 0.05) in comparison to 45.9% inhibition activity for standard Diclofenac sodium (5 mg/kg body weight) . The chloroform and ethanol extracts from O. aucheranum , O. isauricum , O. sericeum , and O. tauricum roots did not cause any gastric lesions or bleeding in the stomach of mice in a 48-h experiment . The Onosma phytochemicals, deoxyshikonin, 3-hydroxy-isovaleryl shikonin, propionyl shikonin, and acetyl shikonin showed significant binding affinities for the Mpro enzyme based on the molecular docking studies using two distinct approaches, in which a SiteMap module of Maestro was used to detect the possible ligand binding sites for the Mpro enzyme. Docking simulations and molecular mechanics suggest that shikonin derivatives might be effective anti-SARS-CoV-2 compounds . Application of natural products and their metabolites as chemically diverse starting building blocks has been a major driving force in drug discovery over the last century. However, the use of natural products is not linked only to the modern era, as most folkloric medicines have plant-derived extracts. Moreover, the technological advancement and new technical development for isolation and identification of the natural bioactive compounds in herbs have motivated scientists to investigate and use them as nutrients and nutraceuticals, as well as curatives. The genus Onosma , known to be widespread worldwide, has a history of medicinal uses against different diseases in the folk medicine system of several civilizations. In this review, the authors rediscover the genus Onosma by detailing the important isolated and identified chemical compounds and extracts, including naphthoquinone (33), flavonoids (30), hydrocarbon (23), phenolic (22), ester (17), alkaloids (20), terpenoids (10), carboxylic acid (11), fatty acids (9), aromatics (12), and liganin (5). The Onosma phytoconstituents that are considered as potential leads amenable for drug development were reported as rosmarinic acid, apigenin, ferulic acid, protocatechuic acid, chlorogenic acid, caffeic acid, p-coumaric acid, vanillic acid. Several biological activities were reported from Onosma compounds and extracts, including, Genotoxicity and Mutagenicity , antifungal, antibacterial, antioxidant, anticancer, antidiabetic, anti-Alzheimer, anti-tyrosinase, anti-lipoxygenases, parasiticidal, anti-inflammatory, and gastric-ulcerogenic activities. Finally, despite the fact that rosmarinic acid is reported as the most detectable compound in the Onosma species, it was not found in other species such as O. echioides , O. hookeri , O. heterophylla , and O. erecta , requiring further investigation for more confirmation by profiling many other species for comparison. |
null | 08dccc9c-e2bc-48b8-8a7a-07ce7fa5f9ac | 8457475 | Pharmacology[mh] | Asthma is a common chronic inflammation of the airways manifesting as shortness of breath, wheeze and chest tightness that can lead to death, especially in older adults. With around 339 million affected people worldwide and an increase in incidence over recent decades, asthma is still under-diagnosed and under treated. The wide variation observed within asthma-related traits, including age of onset, respiratory symptoms, risk factors, frequency of exacerbations, lung function, comorbidities, and underlying inflammatory patterns, provide difficulty for diagnosis and are believed to contribute to heterogeneous treatment response. Although asthma is recognised as a heterogenous disease, conventional stratification of relevant asthma subgroups is mainly limited to Th2-high (eosinophilic) or Th2-low (non-eosinophilic) according to the level of type 2 helper T cell (Th2) cytokine mediated eosinophilic airway inflammation. Other current biomarkers of airway inflammation, such as immunoglobulin E (IgE), or fraction of exhaled nitric oxide (FeNO), are also used clinically to target treatment and predict future risk . In the recent years, new clinical biomarkers of asthma phenotype such as disease severity, exacerbation triggers and comorbidities have been widely used, with the age of onset of asthma and genetic risk factors recently emerging . Early-onset asthma is typically Th2-high atopic or allergic and responsive to steroids. Late-onset asthma is more diverse than early-onset asthma and presents with both Th2-high and Th2-low types with the latter being associated to non-atopy or lifestyle (smoking, obesity) and being characterised by a non-response to steroid therapy . Although asthma shares common genetics with allergic diseases , none all individuals with allergic asthma respond to steroids . This add to the emerging evidence of asthma genetic heterogeneity defining different underlying biological mechanisms, which should help predict the patients’ likely responses to treatment . Leukotriene modifiers are pharmacological therapies for the treatment of asthma and allergic rhinitis. As established through numerous clinical trials, drugs such as montelukast have been shown to be effective in improving lung function, asthma control and quality of life, by reducing airway hyperresponsiveness and eosinophilia . Recent guidelines for the management of asthma in children and adults from the National Institute for Health and Clinical Excellence (NICE), highlighted leukotriene receptor antagonists (LTRA) as a key part of asthma management and recommended their use as the first line add-on therapy for persistent asthmatic patients who still show symptoms despite low dose of inhaled corticosteroids (ICS) . However, a range of 35 to 78 percent of the patients treated with leukotriene inhibitors such as montelukast were found unresponsive across several independent studies . Asthma is a very heterogenous disease resulting from a complex interplay between genetic susceptibility and environment. Despite the ongoing efforts there are many patients with poorly controlled asthma and with treatment based on symptomatology rather than underlying disease pathobiology. A precision medicine approach, based on genetic profiles, biomarkers measures and clinical assessments, is vital to target treatments to the patient population most likely to benefit. Pharmacogenetic studies are therefore essential to the understanding of the genetic interplay in drug response and identification of individuals best targeted treatments. Pharmacogenetic studies have investigated candidate genes in the leukotriene pathway, showing that interpatient variability might be influenced by genetic variation. Variants in ALOX5 , LTA4H , LTC4S , ABCC1 , CYSLTR2 , and SLCO2B1 genes may contribute to the heterogeneity of response to leukotriene modifiers, including montelukast . To date, only the effect of a small number of genetic variants has been validated in independent studies . We focused our analysis on a particular variant in the regulatory region of the leukotriene-A 4 hydrolase (EC3.3.2.6) gene ( LTA4H ), the single nucleotide polymorphism (SNP) rs2660845. This variant has been shown to influence montelukast response through the prevention of exacerbations in a cohort of young adults with asthma from the United States as well as improvement of pulmonary parameters such as peak expiratory flow (PEF) and forced expiratory volume in one second (FEV1) in a Japanese cohort . However, the sample size of those studies was small, and the association observed between the rs2660845 SNP and responsiveness to montelukast has not been validated. In this study, we utilised a very large dataset, consisting of previously described, ethnically diverse case-control studies (GALA II and SAGE), asthma cohorts (BREATHE and PAGES), a randomised clinical trial (RCT) (Tayside RCT) and two large real-life cohorts (the UKBiobank and GoSHARE), to explore the association of rs2660845 with lack of montelukast treatment response. We also wished to determine if age of asthma onset status further refined this putative pharmacogenomic association to provide the beginnings of a precision medicine strategy for the treatment of asthma with montelukast.
Study design Asthmatic patients treated with montelukast for at least 6 months were included in this study. We defined early-onset as individuals ≤18 years old at asthma diagnosis and late-onset for individuals over 18 years old at asthma diagnosis. Early-onset were recruited from five cross-sectional studies (PAGES, BREATHE, Tayside RCT, GALA II and SAGE) as well as from one longitudinal study (GoSHARE, the Genetics of Scottish Health Registry ). BREATHE, Tayside RCT and PAGES’ patients were recruited from primary and secondary care in Scotland. BREATHE and Tayside RCT details of enrolment have been presented in detail previously . PAGES, the Pediatric Asthma Gene Environment Study, was an exploratory study of the genetic variation and exposure in asthmatic Scottish children from various ethnicities . The Genes-environments and Admixture in Hispanics/Latinos (GALA II) study and the Study of African-Americans, Asthma, Genes, and Environments (SAGE) recruited children from community- and clinic-based centers, with GALA II focused on Hispanics/Latinos and SAGE focused on African Americans . BREATHE, PAGES, GALA II and SAGE studies were all part of the Pharmacogenetics in Childhood Asthma (PiCA) consortium . For all studies from the PiCA consortium and Tayside RCT, asthmatic individuals under montelukast treatment were selected by a physician and information such as age, sex, BMI and OCS use, hospitalization and exacerbation were available from a questionnaire filled at patient enrolment. In GoSHARE, asthma patients were selected from a dataset containing complete electronic medical records (EMR), prescription information, hospital, and emergency room records (International Classification of Diseases 10th revision [ICD-10] codes J45) from Tayside, Scotland . Late-onset individuals were recruited from the UKBiobank and GoSHARE. In the UKBiobank, selected asthma patients were British white individuals from self-reported questionnaires and hospital records (ICD10: J45) released in October 2019 by querying an AstraZeneca in-house protected UKBiobank database server. Other information on UKBiobank individuals such as age of onset of asthma and sex were extracted from self-reported questionnaires whereas montelukast usage was extracted from the recent release of drug prescriptions (October 2019). https://www.sciencedirect.com/science/article/pii/S2213260019300554-bib9 The studies included in this article comply with the Declaration of Helsinki, and locally appointed ethics committee have approved respective research protocols and consent procedures. Study participants for GoSHARE provided written informed consent to their data as well as spare blood, left after routine venepuncture, to be held in NHS databases and used for suitable research projects. For the UKBiobank, written consent form was filled by participants on their enrolment in the different selected centre and witnessed and signed by an NHS staff member. Written informed consent was obtained from the participants and/or parent/guardian as relevant for PAGES, BREATHE, Tayside RCT, SAGE and GALAII studies. All data were fully anonymized before granted access. PAGES was approved by the Cornwall and Plymouth Research Ethics Committee (Plymouth, United Kingdom). GoSHARE, BREATHE and Tayside RCT were approved by the Tayside Committee on Medical Research Ethics (Dundee, United Kingdom). The UKBiobank was approved by the National Research Ethics Committee. The Human Research Protection Program Institutional Review Board of the University of California, San Francisco (San Francisco, United States) approved GALA II and SAGE (ethics approval numbers: 217802 and 210362, respectively). Asthma treatment was categorized following British Thoracic Society/Scottish Intercollegiate Guideline Network (BTS/SIGN) guidelines . Only individuals on montelukast therapy were selected for further analysis . Both late-onset and early-onset patients were selected from the GoSHARE study. As age of asthma onset was not reported in the EMR for GoSHARE, we use age at first salbutamol, age at first Inhaled corticosteroid and age at first montelukast prescription, all three over 18 years old to characterize the late-onset group GoSHARE(a) and all three under 18 years old to characterize the early-onset group GoSHARE(b). DNA collection, extraction and analysis For BREATHE and Tayside RCT, saliva was collected in commercially available kits (Oragene, DNA Genotech Ontario, Canada). DNA was prepared using the Qiagen DNeasy 96 kit (Qiagen, Manchester, UK) and genotypes were determined in the University of Dundee laboratory using TaqMan™ based allelic discrimination arrays on an ABI 7700 sequence detection system (ThermoFisher, Waltham, USA). For GoSHARE (a and b) respectively, DNA was extracted from blood samples taken at recruitment or from a prescribed blood test and genotyped using the Infinium Global Screening Array v2 (Illumina, San Diego, USA). Under the UKBiobank project, all recruited individuals were genotyped using a purpose-designed genotypic array: the UKBiobank Axiom Array. As part of the PiCA consortium, PAGES DNA was selected and genotyped in the Spanish National Genotyping Center (CEGEN-PRB3-ISCIII), using the Axiom Precision Medicine Research Array (PMRA) (Affymetrix, Thermo Fisher Scientific Inc, Waltham, MA). GALA II and SAGE were genotyped using the Axiom LAT1 array (World Array 4, Affymetrix, Santa Clara, CA, United States), as described elsewhere . Studies with genome-wide genotyping data were subjected to standard quality control procedures, ensuring that SNP genotyping call rate was above 95% and the absence of deviations from Hardy-Weinberg equilibrium (p>10–6) within control subjects. Samples with discrepancy between genetic sex and reported sex and with family relatedness were removed from the analyses. Binary risk of having an asthma exacerbation The association between the variant of interest and the risk of having an asthma exacerbation was tested. A binary variable related to the absence or presence of at least one exacerbation event in a time frame of 6 to 12 months (0 = no exacerbation, 1 = at least one exacerbation event) was used . Asthma exacerbations were defined based on the American Thoracic Society (ATS)/European Respiratory Society (ERS) guidelines as episodes of worsening of asthma symptoms which require a short course (3–5 days) of oral systemic corticosteroids (OCS) use, hospitalizations or emergency department (ED) visits . The definition was then adapted to each study design and information available. Cross sectional studies collected data on asthma treatment either from pharmacy records, parent/patient-reported medication use, or completed study questionnaires, 6 months for BREATHE, Tayside RCT and PAGES or 12 months before enrolment for GALAII and SAGE. For PAGES, BREATHE and Tayside RCT, the definition of exacerbation was at least one of the following in the previous six months of recruitment: hospital admission, short course of oral corticosteroids (OCS) or absence from school, all due to asthma symptoms and verified by a general practitioner or a nurse. For GALA II and SAGE, asthma exacerbations were defined by the presence of at least one of the following events in the 12 months preceding the study inclusion as available: need to seek emergency asthma care, hospitalization, or the administration of OCS due to asthma symptoms. Response designed on prescription-based information in GoSHARE and UKBiobank was gathered in a time frame of 12 months after date of first montelukast prescription. For the UKBiobank and GoSHARE (a and b), the definition of exacerbation was at least one of the following within 12 months after first montelukast prescription date: hospital admission for asthma symptoms (ICD10: J45), emergency room visit (ER) for asthma symptoms (ICD10: J45), and two or more prescriptions of OCS. Non responders or cases, under montelukast treatment, have at least one course of oral corticosteroids (OCS), hospitalisation, emergency room visit and/or school absence within 6 months (T = time) before date of enrolment for BREATHE, PAGES and Tayside RCT. For GALA II and SAGE, non-responders or cases have at least one course of OCS, hospitalisation and/or emergency room visit within 12 months before date of enrolment. For UKB and GoSHARE, non-responders or cases have at least 2 course of OCS, one hospitalisation and/or one emergency room visit within 12 months after date of first montelukast prescription. To verify the effect of rs2660845 on montelukast response and not just on exacerbation, we tested the association between rs2660845 and exacerbation in non-montelukast users. In late-onset, as we have longitudinal data for GoSHARE(a) cohort, individuals previously selected as being under montelukast treatment where used. The binary effect of experiencing an exacerbation was defined 12 months before start of therapy. In early-onset asthma, genetic data from non-montelukast users were only available from PAGES, BREATHE GALA and SAGE cohorts. Individuals were selected as not being under montelukast treatment at enrolment. The binary risk of having an asthma exacerbation was defined as described before for PAGES, BREATHE and GoSHARE. An asthma exacerbation was defined described before as well for PAGES and BREATHE. For GoSHARE (a) and asthma exacerbation was described as at least one of the following within 12 months before first montelukast prescription date: hospital admission, emergency room visit, course of OCS and discontinuation of the drug. Study details are presented in . Statistical analysis Statistical analysis of PAGES, BREATHE, Tayside RCT, the UKBiobank and GoSHARE (a and b) data were performed in SAS 9.3 (SAS Institute, Cary, NC, USA ) and PLINK 1.9 for GALA II and SAGE. Binary logistic regression models were used to test the association between rs2660845 and asthma exacerbation status. Information related to age, gender, body mass index (BMI), asthma, age onset and exacerbation status 12 months before the start of therapy as well as the first two principal components of the genotype matrix, to control for population stratification, were included as covariates where appropriate and available for each study . The effect of the LTA4H variant was considered additive for this analysis. Statistically significant associations were considered for p-values (P) lower than 0.05. Regarding the number of individuals selected we were sufficiently powered (>80%) to detect an OR of 1.5 or above based on our power calculation . Meta-analysis Association results obtained from late-onset and early-onset studies were meta-analyzed separately to investigate the age of onset effect. For early-onset, populations of European origin were meta-analyzed together as well as with Latin/Hispanic and African American populations. Overall pooled odds ratios from all early-onset patients from GoSHARE(b), BREATHE, Tayside RCT, PAGES, SAGE, and GALA II studies, together with 95% confidence interval of the association between rs2660845 genotype and asthma exacerbation status, were obtained using a random effects model, because the phenotype definition as well as study characteristics varied among studies . Whereas for late-onset Europeans studies (GoSHARE(a) and the UKBiobank), as the phenotype definition and baseline characteristics (age, gender, exacerbation percentage) were similar, a fixed effects model was used. The analysis was performed using the metafor package in R . We tested the heterogeneity among studies by means of the measure of inconstancy (I 2 ) values as low (0–25%), moderate (25–50%) and high (50–75%). A threshold of P<0.05 was used to assess the statistical significance of the main effect association. All association tests were performed using an additive model.
Asthmatic patients treated with montelukast for at least 6 months were included in this study. We defined early-onset as individuals ≤18 years old at asthma diagnosis and late-onset for individuals over 18 years old at asthma diagnosis. Early-onset were recruited from five cross-sectional studies (PAGES, BREATHE, Tayside RCT, GALA II and SAGE) as well as from one longitudinal study (GoSHARE, the Genetics of Scottish Health Registry ). BREATHE, Tayside RCT and PAGES’ patients were recruited from primary and secondary care in Scotland. BREATHE and Tayside RCT details of enrolment have been presented in detail previously . PAGES, the Pediatric Asthma Gene Environment Study, was an exploratory study of the genetic variation and exposure in asthmatic Scottish children from various ethnicities . The Genes-environments and Admixture in Hispanics/Latinos (GALA II) study and the Study of African-Americans, Asthma, Genes, and Environments (SAGE) recruited children from community- and clinic-based centers, with GALA II focused on Hispanics/Latinos and SAGE focused on African Americans . BREATHE, PAGES, GALA II and SAGE studies were all part of the Pharmacogenetics in Childhood Asthma (PiCA) consortium . For all studies from the PiCA consortium and Tayside RCT, asthmatic individuals under montelukast treatment were selected by a physician and information such as age, sex, BMI and OCS use, hospitalization and exacerbation were available from a questionnaire filled at patient enrolment. In GoSHARE, asthma patients were selected from a dataset containing complete electronic medical records (EMR), prescription information, hospital, and emergency room records (International Classification of Diseases 10th revision [ICD-10] codes J45) from Tayside, Scotland . Late-onset individuals were recruited from the UKBiobank and GoSHARE. In the UKBiobank, selected asthma patients were British white individuals from self-reported questionnaires and hospital records (ICD10: J45) released in October 2019 by querying an AstraZeneca in-house protected UKBiobank database server. Other information on UKBiobank individuals such as age of onset of asthma and sex were extracted from self-reported questionnaires whereas montelukast usage was extracted from the recent release of drug prescriptions (October 2019). https://www.sciencedirect.com/science/article/pii/S2213260019300554-bib9 The studies included in this article comply with the Declaration of Helsinki, and locally appointed ethics committee have approved respective research protocols and consent procedures. Study participants for GoSHARE provided written informed consent to their data as well as spare blood, left after routine venepuncture, to be held in NHS databases and used for suitable research projects. For the UKBiobank, written consent form was filled by participants on their enrolment in the different selected centre and witnessed and signed by an NHS staff member. Written informed consent was obtained from the participants and/or parent/guardian as relevant for PAGES, BREATHE, Tayside RCT, SAGE and GALAII studies. All data were fully anonymized before granted access. PAGES was approved by the Cornwall and Plymouth Research Ethics Committee (Plymouth, United Kingdom). GoSHARE, BREATHE and Tayside RCT were approved by the Tayside Committee on Medical Research Ethics (Dundee, United Kingdom). The UKBiobank was approved by the National Research Ethics Committee. The Human Research Protection Program Institutional Review Board of the University of California, San Francisco (San Francisco, United States) approved GALA II and SAGE (ethics approval numbers: 217802 and 210362, respectively). Asthma treatment was categorized following British Thoracic Society/Scottish Intercollegiate Guideline Network (BTS/SIGN) guidelines . Only individuals on montelukast therapy were selected for further analysis . Both late-onset and early-onset patients were selected from the GoSHARE study. As age of asthma onset was not reported in the EMR for GoSHARE, we use age at first salbutamol, age at first Inhaled corticosteroid and age at first montelukast prescription, all three over 18 years old to characterize the late-onset group GoSHARE(a) and all three under 18 years old to characterize the early-onset group GoSHARE(b).
For BREATHE and Tayside RCT, saliva was collected in commercially available kits (Oragene, DNA Genotech Ontario, Canada). DNA was prepared using the Qiagen DNeasy 96 kit (Qiagen, Manchester, UK) and genotypes were determined in the University of Dundee laboratory using TaqMan™ based allelic discrimination arrays on an ABI 7700 sequence detection system (ThermoFisher, Waltham, USA). For GoSHARE (a and b) respectively, DNA was extracted from blood samples taken at recruitment or from a prescribed blood test and genotyped using the Infinium Global Screening Array v2 (Illumina, San Diego, USA). Under the UKBiobank project, all recruited individuals were genotyped using a purpose-designed genotypic array: the UKBiobank Axiom Array. As part of the PiCA consortium, PAGES DNA was selected and genotyped in the Spanish National Genotyping Center (CEGEN-PRB3-ISCIII), using the Axiom Precision Medicine Research Array (PMRA) (Affymetrix, Thermo Fisher Scientific Inc, Waltham, MA). GALA II and SAGE were genotyped using the Axiom LAT1 array (World Array 4, Affymetrix, Santa Clara, CA, United States), as described elsewhere . Studies with genome-wide genotyping data were subjected to standard quality control procedures, ensuring that SNP genotyping call rate was above 95% and the absence of deviations from Hardy-Weinberg equilibrium (p>10–6) within control subjects. Samples with discrepancy between genetic sex and reported sex and with family relatedness were removed from the analyses.
The association between the variant of interest and the risk of having an asthma exacerbation was tested. A binary variable related to the absence or presence of at least one exacerbation event in a time frame of 6 to 12 months (0 = no exacerbation, 1 = at least one exacerbation event) was used . Asthma exacerbations were defined based on the American Thoracic Society (ATS)/European Respiratory Society (ERS) guidelines as episodes of worsening of asthma symptoms which require a short course (3–5 days) of oral systemic corticosteroids (OCS) use, hospitalizations or emergency department (ED) visits . The definition was then adapted to each study design and information available. Cross sectional studies collected data on asthma treatment either from pharmacy records, parent/patient-reported medication use, or completed study questionnaires, 6 months for BREATHE, Tayside RCT and PAGES or 12 months before enrolment for GALAII and SAGE. For PAGES, BREATHE and Tayside RCT, the definition of exacerbation was at least one of the following in the previous six months of recruitment: hospital admission, short course of oral corticosteroids (OCS) or absence from school, all due to asthma symptoms and verified by a general practitioner or a nurse. For GALA II and SAGE, asthma exacerbations were defined by the presence of at least one of the following events in the 12 months preceding the study inclusion as available: need to seek emergency asthma care, hospitalization, or the administration of OCS due to asthma symptoms. Response designed on prescription-based information in GoSHARE and UKBiobank was gathered in a time frame of 12 months after date of first montelukast prescription. For the UKBiobank and GoSHARE (a and b), the definition of exacerbation was at least one of the following within 12 months after first montelukast prescription date: hospital admission for asthma symptoms (ICD10: J45), emergency room visit (ER) for asthma symptoms (ICD10: J45), and two or more prescriptions of OCS. Non responders or cases, under montelukast treatment, have at least one course of oral corticosteroids (OCS), hospitalisation, emergency room visit and/or school absence within 6 months (T = time) before date of enrolment for BREATHE, PAGES and Tayside RCT. For GALA II and SAGE, non-responders or cases have at least one course of OCS, hospitalisation and/or emergency room visit within 12 months before date of enrolment. For UKB and GoSHARE, non-responders or cases have at least 2 course of OCS, one hospitalisation and/or one emergency room visit within 12 months after date of first montelukast prescription. To verify the effect of rs2660845 on montelukast response and not just on exacerbation, we tested the association between rs2660845 and exacerbation in non-montelukast users. In late-onset, as we have longitudinal data for GoSHARE(a) cohort, individuals previously selected as being under montelukast treatment where used. The binary effect of experiencing an exacerbation was defined 12 months before start of therapy. In early-onset asthma, genetic data from non-montelukast users were only available from PAGES, BREATHE GALA and SAGE cohorts. Individuals were selected as not being under montelukast treatment at enrolment. The binary risk of having an asthma exacerbation was defined as described before for PAGES, BREATHE and GoSHARE. An asthma exacerbation was defined described before as well for PAGES and BREATHE. For GoSHARE (a) and asthma exacerbation was described as at least one of the following within 12 months before first montelukast prescription date: hospital admission, emergency room visit, course of OCS and discontinuation of the drug. Study details are presented in .
Statistical analysis of PAGES, BREATHE, Tayside RCT, the UKBiobank and GoSHARE (a and b) data were performed in SAS 9.3 (SAS Institute, Cary, NC, USA ) and PLINK 1.9 for GALA II and SAGE. Binary logistic regression models were used to test the association between rs2660845 and asthma exacerbation status. Information related to age, gender, body mass index (BMI), asthma, age onset and exacerbation status 12 months before the start of therapy as well as the first two principal components of the genotype matrix, to control for population stratification, were included as covariates where appropriate and available for each study . The effect of the LTA4H variant was considered additive for this analysis. Statistically significant associations were considered for p-values (P) lower than 0.05. Regarding the number of individuals selected we were sufficiently powered (>80%) to detect an OR of 1.5 or above based on our power calculation .
Association results obtained from late-onset and early-onset studies were meta-analyzed separately to investigate the age of onset effect. For early-onset, populations of European origin were meta-analyzed together as well as with Latin/Hispanic and African American populations. Overall pooled odds ratios from all early-onset patients from GoSHARE(b), BREATHE, Tayside RCT, PAGES, SAGE, and GALA II studies, together with 95% confidence interval of the association between rs2660845 genotype and asthma exacerbation status, were obtained using a random effects model, because the phenotype definition as well as study characteristics varied among studies . Whereas for late-onset Europeans studies (GoSHARE(a) and the UKBiobank), as the phenotype definition and baseline characteristics (age, gender, exacerbation percentage) were similar, a fixed effects model was used. The analysis was performed using the metafor package in R . We tested the heterogeneity among studies by means of the measure of inconstancy (I 2 ) values as low (0–25%), moderate (25–50%) and high (50–75%). A threshold of P<0.05 was used to assess the statistical significance of the main effect association. All association tests were performed using an additive model.
Study characteristics Study type and baseline characteristics of the participants for each study are presented in . A total of 3,594 individuals (2,514 late-onset and 1,080 early-onset) with asthma from multiple independent studies were used. More than 50% of the participants were female in SAGE, the UKBiobank and GoSHARE (a and b) (55%, 66.5%, 56% and 79% female respectively) and male in BREATHE, Tayside RCT, PAGES and GALA II (61%, 63%, 60%, 57% male respectively). The proportion of early-onset and late-onset individuals experiencing an exacerbation before being on montelukast was between 12% and 22% for BREATHE, Tayside RCT, GoSHARE and UKBiobank but was 67%, 72% and 73% respectively for PAGES, GALA II and SAGE. The minor allele frequency of rs2660845 in the UKB, GoSHARE, BREATHE, Tayside RCT, PAGES, GALA II and SAGE, study populations was 0.27, 0.27, 0.27, 0.28, 0.26, 0.46 and 0.32 respectively. As a control analysis, we tested the association between rs2660845 and non-montelukast users in both late-onset and early-onset . No significant association was observed between rs2660845 and exacerbation rate in individuals with early or late onset not under montelukast treatment (Tables and ). Association between rs2660845 and exacerbation status under montelukast treatment in Europeans with early-onset asthma We assessed the risk of having an exacerbation whilst on montelukast treatment using an additive model. shows the results of logistic regression models assessing the effect of rs2660845 on exacerbation status. The genotypic effect was noted in three out of four European early-onset asthma studies. In BREATHE and Tayside RCT, individuals carrying at least a copy of rs2660845 G allele had respectively 4.4 (95%CI: 1.77–10.96, P = 0.05) and 9.6 (95%CI:1.00–92.19, P = 0.001) times the odds of having an exacerbation, in an additive model adjusted for age at recruitment and BMI. In GoSHARE (b), and PAGES the associations were not significant in a model adjusted for age at first LTRA prescribed and exacerbation 12 months before date of first montelukast prescription (respectively Odds-ratio (OR) 4.5, 95% confidence interval (CI): 0.77–26.24; P = 0.0913 and OR = 0.959; 95%CI: 0.427–2.15; P = 0.668). Association of rs2660845 with exacerbation status under montelukast treatment in Latino/Hispanic and African American early-onset asthma ethnic groups In the Latino/Hispanic population, GALA II, the association was not significant (OR = 1.04; 95%CI: 0.78–1.39; P = 0.788). The African American population (SAGE) presented an opposite nominal association with an OR of 0.27 (95%CI: 0.09–0.80; P = 0.019) in a model adjusted by age of asthma onset, gender and the first two principal components . Meta-analyses of early-onset asthma cohorts A meta-analysis of European early-onset cohorts PAGES, GoSHARE (b), BREATHE, Tayside RCT revealed that carriers of at least one G allele of rs2660845 SNP have 2.92 (95%CI: 1.04–8.18, P = 0.0412) times the odds of experiencing an exacerbation . When adding GALA II (Latino/Hispanic) and SAGE (African American), where rs2660845 MAF is higher than in European cohorts, no significant association between rs2660845 SNP and exacerbation status was found (OR = 1.60, 95%CI: 0.61–4.19; P = 0.342) . Lack of association between rs2660845 genotype and exacerbation status under montelukast treatment in Europeans late-onset asthma and meta-analysis No association between exacerbation risk and LTA4H rs2660845 was seen in any of the late-onset studies . A meta-analysis of the observed effects across late-onset populations of the UKBiobank and GoSHARE(a) showed no significant association between rs2660845 genotype and exacerbation status under montelukast treatment (OR = 1.02, 95%CI: 0.87–1.19 P = 0.833) .
Study type and baseline characteristics of the participants for each study are presented in . A total of 3,594 individuals (2,514 late-onset and 1,080 early-onset) with asthma from multiple independent studies were used. More than 50% of the participants were female in SAGE, the UKBiobank and GoSHARE (a and b) (55%, 66.5%, 56% and 79% female respectively) and male in BREATHE, Tayside RCT, PAGES and GALA II (61%, 63%, 60%, 57% male respectively). The proportion of early-onset and late-onset individuals experiencing an exacerbation before being on montelukast was between 12% and 22% for BREATHE, Tayside RCT, GoSHARE and UKBiobank but was 67%, 72% and 73% respectively for PAGES, GALA II and SAGE. The minor allele frequency of rs2660845 in the UKB, GoSHARE, BREATHE, Tayside RCT, PAGES, GALA II and SAGE, study populations was 0.27, 0.27, 0.27, 0.28, 0.26, 0.46 and 0.32 respectively. As a control analysis, we tested the association between rs2660845 and non-montelukast users in both late-onset and early-onset . No significant association was observed between rs2660845 and exacerbation rate in individuals with early or late onset not under montelukast treatment (Tables and ).
We assessed the risk of having an exacerbation whilst on montelukast treatment using an additive model. shows the results of logistic regression models assessing the effect of rs2660845 on exacerbation status. The genotypic effect was noted in three out of four European early-onset asthma studies. In BREATHE and Tayside RCT, individuals carrying at least a copy of rs2660845 G allele had respectively 4.4 (95%CI: 1.77–10.96, P = 0.05) and 9.6 (95%CI:1.00–92.19, P = 0.001) times the odds of having an exacerbation, in an additive model adjusted for age at recruitment and BMI. In GoSHARE (b), and PAGES the associations were not significant in a model adjusted for age at first LTRA prescribed and exacerbation 12 months before date of first montelukast prescription (respectively Odds-ratio (OR) 4.5, 95% confidence interval (CI): 0.77–26.24; P = 0.0913 and OR = 0.959; 95%CI: 0.427–2.15; P = 0.668).
In the Latino/Hispanic population, GALA II, the association was not significant (OR = 1.04; 95%CI: 0.78–1.39; P = 0.788). The African American population (SAGE) presented an opposite nominal association with an OR of 0.27 (95%CI: 0.09–0.80; P = 0.019) in a model adjusted by age of asthma onset, gender and the first two principal components .
A meta-analysis of European early-onset cohorts PAGES, GoSHARE (b), BREATHE, Tayside RCT revealed that carriers of at least one G allele of rs2660845 SNP have 2.92 (95%CI: 1.04–8.18, P = 0.0412) times the odds of experiencing an exacerbation . When adding GALA II (Latino/Hispanic) and SAGE (African American), where rs2660845 MAF is higher than in European cohorts, no significant association between rs2660845 SNP and exacerbation status was found (OR = 1.60, 95%CI: 0.61–4.19; P = 0.342) .
No association between exacerbation risk and LTA4H rs2660845 was seen in any of the late-onset studies . A meta-analysis of the observed effects across late-onset populations of the UKBiobank and GoSHARE(a) showed no significant association between rs2660845 genotype and exacerbation status under montelukast treatment (OR = 1.02, 95%CI: 0.87–1.19 P = 0.833) .
We report the results of the first study evaluating the association of the SNP rs2660845 with asthma exacerbations whilst on montelukast treatment, in late-onset and early-onset asthma patients, from different populations and ethnic backgrounds. We analyzed data from 3,594 individuals with asthma, 2,514 of which were late-onset and 1,080 early-onset. Although no evidence of rs2660845 association with asthma was found in late-onset cohorts of European ancestry, we observed a significant association (p = 0.0412) and heterogenous response in early-onset asthma carriers of rs2660845 G allele with the European cohorts PAGES, GoSHARE(b), BREATHE and Tayside RCT (heterogeneity P = 0.0364). Higher heterogeneity was found with ethnically different early-onset populations for association between rs2660845 and exacerbation whilst on montelukast treatment (heterogeneity P = 0.0007), with the odds ratio being lower in the African American cohort SAGE and not significant in the Latino/Hispanic cohort GALA II. As our study identifies the rs2660845 genotype as associated with positive, negative and neutral effect on outcome in different populations, mechanistic studies would be required to confirm its relevance to LTRA response in different ethnic groups. However, rs2660845 represents a putative biomarker for prediction of montelukast response in Europeans with early-onset asthma. Since early-onset and late-onset asthma present fundamental differences both in the establishment of disease, as well as genetic variation and expression , we chose to analyse the effect of LTA4H rs2660845 SNP in each onset group. We found that the age of onset plays a role in the response to montelukast in European early-onset asthma patients, where the SNP rs2660845 was significantly associated with heterogenous montelukast response as measured by exacerbation rate. These results were consistent with previous studies where early-onset asthma is presented as more genetic and allergic driven than late-onset which tends to be more environmental and lung centred . There are significant disparities in asthma prevalence, mortality and drug response between ethnic groups . Here we reported different results from four European early-onset cohorts, one Latino/Hispanic and one African American, with rs2660845 MAF higher in GALA II and SAGE populations ( consistent with gnomAD ) compared to the European studies. Previous sequencing of LTA4H in GALA II revealed the prevalence of five polymorphisms differing from those found in a European cohort by Holloway et al . It is possible that ancestry-specific differences in genetic architecture (linkage disequilibrium patterns) may explain why rs2660845 SNP shows no or significant opposite effects in different populations. Furthermore, heterogeneous effects may also be combined with differences in the environmental background between African American, Latino and European populations as well as insufficient sample size . Also as PAGES, GALAII and SAGE, presented a higher exacerbation percentage before montelukast treatment, it is possible that those higher exacerbation percentage explain the high heterogeneity found between early-onset asthma cohorts and so the difference in the outcome. As environmental factors can modulate the effect of genetic variants upon asthma patients or drug response, and we were not able to investigate these cohorts more in depth to determine what environmental factor was driving this difference, it is important to note that we might underestimate the effect of rs2660845 SNP on montelukast response in the European early-onset meta-analysis. Another important driver of the heterogeneity found here is the asthma exacerbation definition. An indirect role of rs2660845 on asthma exacerbation has been suggested by Rao et al ., with the inactivation of LTA4H resulting in less exacerbations . Leukotriene-B 4 (LTB 4 ), the product of the degradation of leukotriene-A 4 by LTA4H, plays a role in recruiting CD4 + , CD8 + and T-cells and thus triggering lung inflammation and airway responsiveness . Therefore, an inactivation of LTA4H would affect those inflammatory cells recruitment by lowering LTB 4 concentration in the blood. As LTB 4 levels in the airways have been shown to increase in blood and bronchoalveolar lavage of asthma patients and to correlate with asthma severity , we examined mRNA expression data from whole blood in the GTEx , eQTLGen and BIOSQTL databases. We found that rs2660845 SNP is associated with LTA4H expression, with AA genotype lowering the expression ( and ). rs2660845 is present in the 5’UTR regulatory region of LTA4H , indicative of a putative cis-eQTL effect. However, the gene expression data analyzed was collected from adults of European ancestry ( for GTEx) and no significant association was found between rs2660845 and montelukast response in an adult cohort of early-onset asthma from the UKBiobank ( and Tables). As LTA4H expression and effects have been shown to be different in late-onset and early-onset asthma , it is possible that the results would be significant in a pediatric study. A possible mechanism in early-onset asthma is presented in . By down regulating LTA4H activity, individuals with a rs2660845 AA genotype have a reduced LTB 4 blood concentration resulting in less LTB 4 driven exacerbations, simultaneously montelukast, acting as a cysteinyl receptor antagonist, will prevent other products of LTA 4 degradation by leukotriene C4 synthase from binding to the receptor, thus reducing exacerbations through this arm of the LTA 4 pathway. Therefore, LTA4H inhibitors, 5-Lipoxygenase Activating Protein inhibitors and Leukotriene-B 4 receptor antagonists might represent a potential therapeutic strategy that could modulate key aspects of early-onset asthma . Our study has limitations that should also be considered. A fine mapping of the LTA4H locus or haplotype analysis as reported by Holloway and colleagues may help our understand the relationship between genetic variants of LTA4H , but was prevented by either GWAS or haplotype data being unavailable for all cohorts. We decided to focus on rs2660845 as literature indicated a putative association with outcomes for asthma treatment and replicated this result in our European early-onset cohorts despite a high heterogeneity between studies. Differences in patient recruitment (GP selection vs prescriptions-base selection, GP selection between asthma studies) and definition of asthma exacerbation despite LTRA treatment such as time frame variation of 6 to 12 months between longitudinal cohorts and asthma study as well as between asthma studies, exacerbation markers used in only one or two cohorts (school absence for BREATHE, PAGES and Tayside RCT, at least two OCS prescriptions for UKB and GoSHARE) explained a large part of this heterogeneity. Within ethnic group differences could also be a driver of the high heterogeneity found here as it can be more significant than heterogeneity across different ethnic groups. Moreover, another potential source of variation came from the use of retrospective information about the occurrence or absence of asthma exacerbations, partly based on self-reporting within asthma studies. Another important limitation is that we were not able to account for any aetiology of asthma exacerbations as well as exposure to potential environmental triggers such as smoking, one of the most important environmental factors knowing its impact on lungs. Unfortunately, the information was not collected in a similar way for all studies, with only second-hand smoking information available for the pediatric cohorts (PAGES, BREATHE, Tayside RCT, GALAII and SAGE) and primary smoking status only available for a subset of GoSHARE and the UKBiobank cohorts. Finally, we have also assumed that treatment has been assigned based on similar prescribing criteria and did not take into consideration specific LTRA dose and type or any index of treatment adherence (BTS/SIGN guidelines) since information related to these variables was not available for most of the studies included. We also did not investigate other treatments and it is possible that individuals with increased risk of exacerbation might not have received LTRA treatment which would tend to underestimate the effect of the interaction between montelukast and rs2660845 for exacerbations. As other treatments were not taken into account, the effect found in the early-onset asthma group could also be overestimated and mostly linked to a more serious disease. Altogether, this heterogeneity in data availability could represent a potential interpretation bias in terms of response to asthma treatment. To our knowledge, this is the only study of leukotriene response to utilize populations of different age of onset and different ethnic backgrounds. These results, together with published studies in other European and Latino/Hispanic cohorts , support a strong involvement of age of onset, LTA4H and LTB4 production in the pathophysiology of asthma. Our results add to the growing evidence of age of asthma onset as a robust clinical biomarker and moreover suggest stratifying treatment by age of asthma onset and rs2660845 status can form the basis of a precision medicine strategy to select those patients most likely to benefit from montelukast treatment.
S1 Table Power of the sample size for combined cohorts to detect increases in OR for asthma exacerbation. GoSHARE (a) is the late-onset population (>18 years-old); GoSHARE (b) is the early-onset population (< = 18 years-old). (DOCX) Click here for additional data file. S2 Table Selection of asthmatic patients in GoSHARE by treatment steps. Step 1: inhaled short-acting β2-agonists (SABA) on demand; Step 2: regular inhaled steroids (ICS) plus SABA on demand; Step 3: regular inhaled long-acting β2-agonists (LABA) (salmeterol or formoterol) plus ICS with SABA on demand; Step 4: oral montelukast with SABA on demand (plus ICS plus/or regular LABA). (DOCX) Click here for additional data file. S3 Table Covariates association with asthma exacerbation binary trait. For GALA II and SAGE, betas 95%(CI) are reported for quantitative variables; * Traits with a P <0.05 were used as covariates in the logistic regression. (DOCX) Click here for additional data file. S4 Table Details of late-onset and early-onset in three non-montelukast user populations. 1 Exacerbation within 6 months; 2 Exacerbation within 12 months; OCS: Oral Corticosteroids; ER: Emergency Room visit; GoSHARE (a) is the late-onset population (>18 years-old); GoSHARE (b) is the early-onset population (< = 18 years-old). (DOCX) Click here for additional data file. S5 Table Details of early-onset asthma adult montelukast user from the UKBiobank. 1 Exacerbation within 6 months; 2 Exacerbation within 12 months; OCS: Oral Corticosteroids; ER: Emergency Room visit; Patients were diagnosed as having early-onset asthma; Montelukast prescription records were only available as adults. (DOCX) Click here for additional data file. S6 Table Association between rs2660845 and asthma exacerbation in early-onset UKBiobank individuals 12 months after taking montelukast prescription as adults. Patients were diagnosed as having early-onset asthma; Montelukast prescription records were only available as adults. (DOCX) Click here for additional data file. S7 Table Cis-eQTL effect of rs2660845 on LTA4H expression in whole blood from adult cohorts. BIOSQTL: The Biobank-Based Integrative Omics Study Quantitative Trait Locus; eQTLGen: expression Quantitative Trait Loci Genetic; FDR: False Discovery Rate; ID: Gencode Identifier. (DOCX) Click here for additional data file. S8 Table Genotype counts for rs2660845 SNP by exacerbation status. 1 Patients were diagnosed as having early-onset asthma but with montelukast prescription records only available as adults. 2 GoSHARE individuals with exacerbation events in a year before first prescription of montelukast. 3 Individuals from BREATHE and PAGES cohorts not under montelukast treatment. GoSHARE(a): individuals with age at first salbutamol, age at first inhaled corticosteroid and age at first montelukast prescription, all three over 18 years old. GoSHARE(b): individuals with age at first salbutamol, age at first inhaled corticosteroid and age at first montelukast prescription, all three under or at 18 years old. (DOCX) Click here for additional data file. S1 Fig Age (a) and ethnicity (b) distributions in the GTEx portal (V8). (TIF) Click here for additional data file. S2 Fig Box plot showing the cis -eQTL effect of rs2660845 on LTA 4 H expression in whole blood (beta = -0.020, P-value = 0.46). (GTEx v8). (TIF) Click here for additional data file. S1 Data (DOCX) Click here for additional data file.
|
Dispersal shapes compositional and functional diversity in aquatic microbial communities | a7987aa7-ae6d-41f7-a905-f6009eb5180e | 11651098 | Microbiology[mh] | Microbial community assembly involves stochastic processes like passive dispersal, drift, and diversification, leading to random fluctuations in species abundances , and deterministic, niche-related phenomena that affect species through abiotic and biotic selection . The relative importance of different assembly processes may shift during succession, from predominantly stochastic during initial colonization to increasingly deterministic in mature communities . This transition is disrupted if extremely low and high dispersal rates override deterministic abiotic selection . For instance, dispersal limitation promoted the emergence of distinct microbial assemblages in identical bioreactor habitats and lake water microcosms . Conversely, homogenizing dispersal mitigated the effects of abiotic selection and increased community similarity . Moreover, dispersal need not be a continuous process but also encompasses extremes like transient isolation or total coalescence . While the interplay of dispersal and biotic selection is not fully understood, evidence suggests that antagonistic and facilitative interspecific interactions, modified by historical contingency , shape community dynamics upon habitat colonization . Moreover, similar climax communities may emerge despite moderate dispersal , indicating that biotic interactions may act as stabilizing filters for community structure. Metacommunities are networks of local communities interconnected by source-sink dynamics . The spatial insurance hypothesis predicts that dispersal mitigates the negative effect of suboptimal local (abiotic or biotic) conditions . Neighboring communities may rescue depleted ones via dispersal to circumvent local extinction and preserve metacommunity diversity. However, connectivity is a double-edged sword, as the isolation of local communities may protect species from superior competitors . Such competitively weak rare species may promote functional diversity , and thus possibly affect community performance. We experimentally investigated the interplay between dispersal-related processes and biotic interactions in shaping bacterial community structure and functioning. Dispersal limitation served to generate parallel bacterial assemblages with contrasting composition and functioning at identical environmental conditions . We hypothesized (i) that functional differences and beta diversity among stochastically assembled communities would increase during semi-continuous cultivation in the absence of dispersal, but (ii) that subsequent homogenizing dispersal would cause a decrease in beta and gamma diversity, as well as a reduction in functional variability due to biotic selection of disproportionally competitive populations from the metacommunity. Finally, we assessed the implications of species interactions on the observed decrease in beta and gamma diversity, and functional variability, following homogenizing dispersal (i.e., mixing of all communities). Initial colonization and compositional changes during semi-continuous growth cycles We conducted a 34-day culture experiment in artificial lake water using 20 parallel freshwater bacterial assemblages over six semi-continuous growth cycles . The putative 16S reads represented 0.04% ± 0.04% (mean ± standard deviation) of total metagenomic reads per sample . Altogether, 118 bacterial genera were identified, with Pseudomonas (10.9%), Flavobacterium (9.2%), Aeromonas (6.7%), Acidovorax (5.9%), and Limnohabitans (5.9%) being the most abundant ones . Experimentally induced dispersal limitation (growth cycle 0, C0) resulted in communities with distinct structures . Acidovorax , Aeromonas, and Pseudomonas were both, abundant (>30% of reads) and prevalent (present in 20, 15, and 15 microcosms, respectively). C0 was also characterized by genera that were only abundant in single microcosms, such as Rhodoferax , Caulobacter , Rheinheimera, and Deefgea . Initially prevalent taxa like Acidovorax and Pseudomonas persisted from the growth cycle 1 to 6 (C1–C6). Some rarer taxa, such as Pelomonas , Cellvibrio , Paucibacter , and Duganella , also increased in abundance over the growth cycles but remained restricted to few microcosms . In C6, half of the communities were dominated by Acidovorax , Pseudomonas, or Aeromonas (i.e., >40% of total read numbers), while the other half were unique with respect to the most abundant taxon. Local operational taxonomic unit (OTU) richness per microcosm significantly decreased throughout the cultivation cycles, from 111 ± 33 OTUs at C0 to 78 ± 38 OTUs at C6 (linear mixed model, P < 0.001). By contrast, the total number of OTUs in all microcosms tended to remain stable from C0 to C6 (CV = 5.1%, ). Most C0 communities were more dissimilar than expected by chance (β RC > 0.95). The proportion of community pairs more dissimilar than expected by chance gradually decreased over the cycles (from 73% to 46% of the pairwise comparisons). By contrast, pairs more similar than by chance (β RC < −0.95) increased, from 11% at C0 to 19% at C6 . Community type affects changes in composition and bulk parameters We categorized communities according to dominant taxon. Acidovorax -type communities transitioned from predominantly dissimilar pairs at C0 (β RC > 0.95) to increasingly more similar in subsequent cycles . An even steeper decrease of β RC was observed for Pseudomonas -type communities, with all pairwise comparisons being <−0.95 by growth cycle 4 (C4). Aeromonas -type communities were already highly similar (β RC < −0.95) at C0 and remained stable over the growth cycles . The three community types together harbored 82% of all genera detected at C6 . The three dominant genera only co-occurred during the early growth cycles . The community types exhibited similar OTU richness (repeated measurement analysis of variance [ANOVA], P > 0.05), but Aeromonas -type communities were less even (Pielou’s evenness) than the other two types and the unique communities (repeated measurement ANOVA, P < 0.001) . Twice as many genera were exclusive to the Acidovorax -type communities than to those of the other two types . Bacterial abundance per microcosm did not correlate with biomass (Fig. S2): bacterial abundance decreased over the growth cycles , whereas communities produced increasingly more biomass, i.e., larger cells . Cellobiose consumption in the microcosms did not significantly change over cycles (cycle, repeated measurement ANOVA, P = 0.126, ; ), but carbon use efficiency (CUE) increased 1.5-fold. . Community type affected both, magnitude and temporal changes of some community bulk parameters . Acidovorax -type communities had higher bacterial abundances than Pseudomonas and Aeromonas types . Bacterial abundances decreased over the growth cycles in Acidovorax - and Pseudomonas -type communities but increased in the Aeromonas -type ones . The overall increase in biomass was mainly driven by the steep rise in the Acidovorax -type communities . By contrast, the rate of CUE increases only slightly varied between community types . Communities of unique composition showed high variability in bulk parameters amongst each other: some unique communities had more than three times higher biomasses and cellobiose consumption rates than the average of the common community types . Changes in genus-level competitiveness over time (Elo-rating) We used Elo-rating (originally developed to rank chess players across multiple tournaments ) as an index to assess the overall performance of individual taxa within the metacommunity over time, i.e., how often they occurred in the microcosms, and their relative abundances in these communities. The Elo-rating was used to generate rank distributions of genera with >0.1% relative abundances in at least one microcosm between C0 and C6 . Several successful primary colonizers (rating above 75th percentile at C0), such as Flavobacterium , Pseudarcicella , and Aquabacterium, significantly declined in their Elo-rating, whereas Rhodoferax became more competitive . While most of the initially less competitive genera did not significantly change or even decreased in Elo-rating (e.g., Polaromonas , Rheinheimera, and Methylibium ), several others, such as Paucibacter , Duganella , Variovorax , Bosea, and Pelomonas significantly improved in competitive performance over the cycles (Spearman rank correlation, P < 0.05; ). The Elo-rating of Acidovorax, Aeromonas, and Pseudomonas were all above the median. While the ratings of the former two did not change over the cycles, Pseudomonas slightly but significantly decreased (from 1,039 to 996, Spearman rank correlation, P < 0.05; ), reflecting its increasingly restricted occurrence across microcosms. Acidovorax had the third-highest Elo-rating in C0, and the highest one by a large margin in C6. Homogenizing dispersal event We experimentally induced a homogenizing dispersal event (growth cycle 7, C7) using the 20 parallel microbial communities from C6 . Homogenizing dispersal produced highly similar microbial assemblages (β RC < −0.95; ) dominated by the genera that already formed the highest abundances in 50% of the C6 microcosms ( Pseudomonas , Acidovorax , and Aeromonas, ). Moreover, 20% of genera from C6 were not detected in C7, including some that represented a sizable fraction of the C7 inoculum (e.g., Caulobacter , ). A neutral and a competitive model were compared in their power to predict the community composition after the homogenizing dispersal event (C7, ). The neutral model was based on the principle of mass effects , i.e., the respective abundances of each genus in C6 microcosms explained their abundances in C7. It successfully predicted 20% of the genera, representing ~44.8% ± 9.9% of abundances . The competitiveness model, based on genus-specific Elo-ratings from C6, most accurately predicted the C7 abundances of three of the top 10 most abundant genera ( Pelomonas , Bosea, and Cellvibrio ), accounting for 11.2% ± 2.3% of abundances. Only Pseudomonas outperformed both predictors, whereas many rare genera performed worse than predicted by either model . Bulk community parameters of the C7 communities were highly similar . Biomass and CUE in C7 matched the average values of the C6 communities , while total cell abundances and cellobiose consumption were ~13% and ~10% lower than before homogenizing dispersal . We processed the metagenomes from C6 and C7 to retrieve metagenome-assembled genomes (MAGs) that defined the C6 “community types,” which were affiliated with the most abundant taxa across the microcosms. From the dominant genera Pseudomonas , Acidovorax, and Aeromonas , we obtained 4, 15, and 4 MAGs from C6 and 4, 19, and 21 from C7, respectively . Additionally, we retrieved nearly single MAGs from the most abundant taxa in microcosms of unique composition at C6, including the genera Caulobacter, Cellvibrio, Dugannella, and Pararheinheimera . Based on the ANI distance of the MAGs, a single Pseudomonas genotype, P. azotoformans, dominated in C7 communities . Acidovorax and Aeromonas were each represented by two mutually exclusive genotypes, A. temperans or A. soli , and A. hydrophila or A. bestiarum , respectively . Genomic traits of dominant community members We assessed genes for secretion systems, amino acid biosynthesis, and cellobiose degradation in the dominant genera at C7 and genera dominating at least one C6 microcosm . Type 2 secretion systems (T2SS) were found in all MAGs. Only MAGs affiliated with Pseudomonas , Aeromonas, and Acidovorax featured T3SS, T5SS, and T6SS, indicating strong competitive traits . Of the studied taxa, only A. hydrophila was prototrophic for amino acids. Pseudomonas and Acidovorax MAGs each lacked synthesis pathways for specific amino acids . Aeromonas MAGs and P. azotoformans MAGs contained genes involved in cellobiose consumption (bgl-B and bgl-X, respectively), whereas Acidovorax MAGs did not. All MAGs dominating single microcosms at C6 were prototrophic for amino acids and featured genes coding for β-glucosidases . We conducted a 34-day culture experiment in artificial lake water using 20 parallel freshwater bacterial assemblages over six semi-continuous growth cycles . The putative 16S reads represented 0.04% ± 0.04% (mean ± standard deviation) of total metagenomic reads per sample . Altogether, 118 bacterial genera were identified, with Pseudomonas (10.9%), Flavobacterium (9.2%), Aeromonas (6.7%), Acidovorax (5.9%), and Limnohabitans (5.9%) being the most abundant ones . Experimentally induced dispersal limitation (growth cycle 0, C0) resulted in communities with distinct structures . Acidovorax , Aeromonas, and Pseudomonas were both, abundant (>30% of reads) and prevalent (present in 20, 15, and 15 microcosms, respectively). C0 was also characterized by genera that were only abundant in single microcosms, such as Rhodoferax , Caulobacter , Rheinheimera, and Deefgea . Initially prevalent taxa like Acidovorax and Pseudomonas persisted from the growth cycle 1 to 6 (C1–C6). Some rarer taxa, such as Pelomonas , Cellvibrio , Paucibacter , and Duganella , also increased in abundance over the growth cycles but remained restricted to few microcosms . In C6, half of the communities were dominated by Acidovorax , Pseudomonas, or Aeromonas (i.e., >40% of total read numbers), while the other half were unique with respect to the most abundant taxon. Local operational taxonomic unit (OTU) richness per microcosm significantly decreased throughout the cultivation cycles, from 111 ± 33 OTUs at C0 to 78 ± 38 OTUs at C6 (linear mixed model, P < 0.001). By contrast, the total number of OTUs in all microcosms tended to remain stable from C0 to C6 (CV = 5.1%, ). Most C0 communities were more dissimilar than expected by chance (β RC > 0.95). The proportion of community pairs more dissimilar than expected by chance gradually decreased over the cycles (from 73% to 46% of the pairwise comparisons). By contrast, pairs more similar than by chance (β RC < −0.95) increased, from 11% at C0 to 19% at C6 . We categorized communities according to dominant taxon. Acidovorax -type communities transitioned from predominantly dissimilar pairs at C0 (β RC > 0.95) to increasingly more similar in subsequent cycles . An even steeper decrease of β RC was observed for Pseudomonas -type communities, with all pairwise comparisons being <−0.95 by growth cycle 4 (C4). Aeromonas -type communities were already highly similar (β RC < −0.95) at C0 and remained stable over the growth cycles . The three community types together harbored 82% of all genera detected at C6 . The three dominant genera only co-occurred during the early growth cycles . The community types exhibited similar OTU richness (repeated measurement analysis of variance [ANOVA], P > 0.05), but Aeromonas -type communities were less even (Pielou’s evenness) than the other two types and the unique communities (repeated measurement ANOVA, P < 0.001) . Twice as many genera were exclusive to the Acidovorax -type communities than to those of the other two types . Bacterial abundance per microcosm did not correlate with biomass (Fig. S2): bacterial abundance decreased over the growth cycles , whereas communities produced increasingly more biomass, i.e., larger cells . Cellobiose consumption in the microcosms did not significantly change over cycles (cycle, repeated measurement ANOVA, P = 0.126, ; ), but carbon use efficiency (CUE) increased 1.5-fold. . Community type affected both, magnitude and temporal changes of some community bulk parameters . Acidovorax -type communities had higher bacterial abundances than Pseudomonas and Aeromonas types . Bacterial abundances decreased over the growth cycles in Acidovorax - and Pseudomonas -type communities but increased in the Aeromonas -type ones . The overall increase in biomass was mainly driven by the steep rise in the Acidovorax -type communities . By contrast, the rate of CUE increases only slightly varied between community types . Communities of unique composition showed high variability in bulk parameters amongst each other: some unique communities had more than three times higher biomasses and cellobiose consumption rates than the average of the common community types . We used Elo-rating (originally developed to rank chess players across multiple tournaments ) as an index to assess the overall performance of individual taxa within the metacommunity over time, i.e., how often they occurred in the microcosms, and their relative abundances in these communities. The Elo-rating was used to generate rank distributions of genera with >0.1% relative abundances in at least one microcosm between C0 and C6 . Several successful primary colonizers (rating above 75th percentile at C0), such as Flavobacterium , Pseudarcicella , and Aquabacterium, significantly declined in their Elo-rating, whereas Rhodoferax became more competitive . While most of the initially less competitive genera did not significantly change or even decreased in Elo-rating (e.g., Polaromonas , Rheinheimera, and Methylibium ), several others, such as Paucibacter , Duganella , Variovorax , Bosea, and Pelomonas significantly improved in competitive performance over the cycles (Spearman rank correlation, P < 0.05; ). The Elo-rating of Acidovorax, Aeromonas, and Pseudomonas were all above the median. While the ratings of the former two did not change over the cycles, Pseudomonas slightly but significantly decreased (from 1,039 to 996, Spearman rank correlation, P < 0.05; ), reflecting its increasingly restricted occurrence across microcosms. Acidovorax had the third-highest Elo-rating in C0, and the highest one by a large margin in C6. We experimentally induced a homogenizing dispersal event (growth cycle 7, C7) using the 20 parallel microbial communities from C6 . Homogenizing dispersal produced highly similar microbial assemblages (β RC < −0.95; ) dominated by the genera that already formed the highest abundances in 50% of the C6 microcosms ( Pseudomonas , Acidovorax , and Aeromonas, ). Moreover, 20% of genera from C6 were not detected in C7, including some that represented a sizable fraction of the C7 inoculum (e.g., Caulobacter , ). A neutral and a competitive model were compared in their power to predict the community composition after the homogenizing dispersal event (C7, ). The neutral model was based on the principle of mass effects , i.e., the respective abundances of each genus in C6 microcosms explained their abundances in C7. It successfully predicted 20% of the genera, representing ~44.8% ± 9.9% of abundances . The competitiveness model, based on genus-specific Elo-ratings from C6, most accurately predicted the C7 abundances of three of the top 10 most abundant genera ( Pelomonas , Bosea, and Cellvibrio ), accounting for 11.2% ± 2.3% of abundances. Only Pseudomonas outperformed both predictors, whereas many rare genera performed worse than predicted by either model . Bulk community parameters of the C7 communities were highly similar . Biomass and CUE in C7 matched the average values of the C6 communities , while total cell abundances and cellobiose consumption were ~13% and ~10% lower than before homogenizing dispersal . We processed the metagenomes from C6 and C7 to retrieve metagenome-assembled genomes (MAGs) that defined the C6 “community types,” which were affiliated with the most abundant taxa across the microcosms. From the dominant genera Pseudomonas , Acidovorax, and Aeromonas , we obtained 4, 15, and 4 MAGs from C6 and 4, 19, and 21 from C7, respectively . Additionally, we retrieved nearly single MAGs from the most abundant taxa in microcosms of unique composition at C6, including the genera Caulobacter, Cellvibrio, Dugannella, and Pararheinheimera . Based on the ANI distance of the MAGs, a single Pseudomonas genotype, P. azotoformans, dominated in C7 communities . Acidovorax and Aeromonas were each represented by two mutually exclusive genotypes, A. temperans or A. soli , and A. hydrophila or A. bestiarum , respectively . We assessed genes for secretion systems, amino acid biosynthesis, and cellobiose degradation in the dominant genera at C7 and genera dominating at least one C6 microcosm . Type 2 secretion systems (T2SS) were found in all MAGs. Only MAGs affiliated with Pseudomonas , Aeromonas, and Acidovorax featured T3SS, T5SS, and T6SS, indicating strong competitive traits . Of the studied taxa, only A. hydrophila was prototrophic for amino acids. Pseudomonas and Acidovorax MAGs each lacked synthesis pathways for specific amino acids . Aeromonas MAGs and P. azotoformans MAGs contained genes involved in cellobiose consumption (bgl-B and bgl-X, respectively), whereas Acidovorax MAGs did not. All MAGs dominating single microcosms at C6 were prototrophic for amino acids and featured genes coding for β-glucosidases . Compositional and functional variability in the absence of dispersal Dispersal limitation during initial colonization of the microcosms (Cycle 0) led to a set of stochastically assembled communities with high β-diversity . Subsequent semi-continuous cultivation (i.e., zero dispersal rates) revealed contrasting effects of local isolation on different levels of diversity . Our findings align with observations in anaerobic bioreactors, where stochastically assembled communities decreased in richness during the transition to a deterministic regime . They also experimentally support microbial metacommunity models predicting that the absence of dispersal will strengthen local biotic selection . Other experimental systems, such as freshwater nematode metacommunities, maintained stable diversity levels despite prolonged local isolation . This difference in our findings is probably due to resource availability: our experimental system relied on a limited number of resources (cellobiose and glucose), thereby promoting biotic selection and diversity loss. By contrast, nematodes could exploit a wide range of resources, including bacteria, microphytobenthos, protists, meiofauna, or organic debris . This likely led to reduced substrate competition and niche separation, which in turn stabilized diversity during segregation. Interestingly, total compartmentalization rarely resulted in the extinction of OTUs (or their decrease below the detection level of our method) at the metacommunity level, as illustrated by stable γ-diversity over the growth cycles . Instead, it led to a decline in OTU occurrence or even to “endemism” within single microcosms, illustrating both their redundant roles in most local assemblages and their likely dependence on positive biotic interactions to circumvent elimination . This gradual “purging” of OTUs from microcosm communities during their transition to more deterministic assembly processes was also the main driver of increasing β-diversity . The simultaneous increase in biomass and CUE of microcosm communities during growth cycles contrasted with stable cellobiose consumption rates . This suggests that their improved performance was not directly driven by specialized taxa that could degrade the primary resource, but is best attributed to increasing efficiency of using the “common good,” cellobiose-derived glucose. Other biotic interactions might have also contributed to reduced energy waste, e.g., enhanced cross-feeding on secreted metabolites , or energy reallocation from metabolically costly competitive traits to growth yield due to reduced interspecific competition (decrease in α-diversity; ) . Our findings also speak for MacArthur’s minimization principle in communities developing under competition at stable conditions: unutilized resources decreased with community maturity due to niche complementarity among species . While this concept has received little attention , an experimental study using synthetic phytoplankton communities confirmed its predictions regarding biomass production . Our results extend these findings by showing that cellobiose-derived carbon was increasingly fixed into microbial biomass across growth cycles, irrespective of community structure . Finally, microevolutionary adaptation toward more efficient glucose consumption could also explain the increasing efficiency. Originally designed for assessing dyadic interactions within game tournaments , Elo-rating has been used in biology to assess the social structure in primates . Our implementation demonstrated suitability for metacommunity analysis by clearly highlighting Acidovorax as the overall “winner” across multiple communities . More importantly, it gave insight into subtle community re-arrangements during growth cycles that would have been challenging to detect without context-dependent measure, e.g., the increasing importance of Paucibacter , Bosea , or Pelomonas , and the concomitant decline of Flavobacterium , Pseudodarcicella , or Aquabacterium . Additionally, it was the best predictor for the performance of three of the top ten most abundant genera after metacommunity mixing . Thus, Elo-rating could be an additional tool to assess the overall success of taxa in metacommunities based on their competitive performance within and among local assemblages. Community types Stochastic assembly processes can generate compositional and functionally distinct communities . We show that dispersal limitation within metacommunities may produce recurrent community types with different carrying capacities , evenness , and subsets of exclusively associated taxa . We defined types from genera dominating three or more microcosms ( Acidovorax , Pseudomonas, and Aeromonas ) . These genera are typically members of the rare aquatic biosphere that proliferate upon input of organic matter or in substrate-rich microniches . Acidovorax was initially seeded into all microcosms, and all local populations survived over the growth cycles. Since these bacteria lack a known cellobiose degradation mechanism , cellobiose-derived glucose must have been available to them as a “common good.” By contrast, both Pseudomonas and Aeromonas were dispersal-limited and more vulnerable to biotic selection. In general, the community types self-stabilized: while initial stochastic dispersal established the state for subsequent development ( , C0), the biological interaction cycles resulted in their deterministic “purification.” This led to stable or increasing within-type similarity against a background of increasing metacommunity-level β-diversity . Effects of homogenizing dispersal on the metacommunity Experimental homogenizing dispersal increased α-diversity and similarity (lower β-diversity) of local microcosm communities but led to a reduction of total metacommunity (γ) diversity . These observations do not align with the theoretical predictions for a fully connected metacommunity subjected to high dispersal rates . Thus, the effects of a singular coalescence event differ from the source-sink dynamics resulting from a continuous process of connectivity. Moderate local species sorting in post-coalescence microcosms was suggested by the large proportion of abundances of individual genera explained by the parent communities (i.e., by the neutral model, ). The appearance of novel positive interactions among previously allopatric populations may also have contributed to the increased local diversity . Since our experiment was limited to a single growth cycle after coalescence, we cannot assess if this high initial diversity was only temporary. A gradual loss of diversity over a 6-week period was demonstrated in an experimental study of mixed soil and carcass communities . Upon coalescence, the heterogeneous assemblages transitioned to novel, more uniform communities that differed from all source communities . Homogenization of synthetic bacterial communities has been observed already at low dispersal rates . Comparable findings have been reported from long-term field observations at the landscape scale: the anthropogenic connection of freshwater bodies (related to the construction of a reservoir) led to the homogenization of the zooplankton metacommunity . Coalescence also led to functional uniformity : community performance didn’t improve after mixing but instead stabilized around the median value of the parent communities . This contrasts with previous observations where the best-performing parent community dictated both, the structure and function of post-coalescence methanogenic assemblages . The loss of functional variability most likely resulted from the disproportional decline or extinction of functionally distinct taxa that dominated in single C6 communities and significantly contributed to this variability . These “endemic” populations, Rheinheimera , Duganella, and Caulobacter , proved to be extremely vulnerable to competitive exclusion . Thus, our experimental observations shed light on how homogenizing dispersal can affect species trait distributions and lead to a loss of functional variability at the metacommunity level , thereby potentially altering ecosystem functioning through the replacement of specialists at the expense of generalists and functionally inefficient species . The post-coalescence dominance of one genotype of Pseudomonas from a single isolated microcosm, P. azotoformans, conspicuously exceeded our predictions . The analysis of the corresponding MAGs revealed that P. azotoformans was the only abundant community member that featured T5SS and T6SS. These secretion systems confer competitive advantages to pseudomonads by delivering effectors such as nucleases, amidases, hydrolases, or phospholipases to neighboring bacterial cells and the external milieu . Taken together, our findings suggest that dispersal limitation may play a key role in defining community performance, by stochastically segregating highly efficient “bottom-up” specialists from taxa that outcompete them via negative biotic interactions . This holds relevance for a rational selection of stable microbial assemblages for both industrial and ecosystem restoration purposes . Specifically, we demonstrate the feasibility of a “top-down” design approach to optimize degradation efficiency in synthetic communities by producing rare variants that outperform the more common types: the highest levels of cellobiose degradation occurred in a unique dominated community stable over the six growth cycles (i.e., Caulobacter ; ) but did not survive community coalescence . Our findings thus provide a potential alternative to classical bottom-up approaches , by allowing for intrinsic biotic relationships from initial stochastic assembly to serve as stabilizing force during deterministic selection . Dispersal limitation during initial colonization of the microcosms (Cycle 0) led to a set of stochastically assembled communities with high β-diversity . Subsequent semi-continuous cultivation (i.e., zero dispersal rates) revealed contrasting effects of local isolation on different levels of diversity . Our findings align with observations in anaerobic bioreactors, where stochastically assembled communities decreased in richness during the transition to a deterministic regime . They also experimentally support microbial metacommunity models predicting that the absence of dispersal will strengthen local biotic selection . Other experimental systems, such as freshwater nematode metacommunities, maintained stable diversity levels despite prolonged local isolation . This difference in our findings is probably due to resource availability: our experimental system relied on a limited number of resources (cellobiose and glucose), thereby promoting biotic selection and diversity loss. By contrast, nematodes could exploit a wide range of resources, including bacteria, microphytobenthos, protists, meiofauna, or organic debris . This likely led to reduced substrate competition and niche separation, which in turn stabilized diversity during segregation. Interestingly, total compartmentalization rarely resulted in the extinction of OTUs (or their decrease below the detection level of our method) at the metacommunity level, as illustrated by stable γ-diversity over the growth cycles . Instead, it led to a decline in OTU occurrence or even to “endemism” within single microcosms, illustrating both their redundant roles in most local assemblages and their likely dependence on positive biotic interactions to circumvent elimination . This gradual “purging” of OTUs from microcosm communities during their transition to more deterministic assembly processes was also the main driver of increasing β-diversity . The simultaneous increase in biomass and CUE of microcosm communities during growth cycles contrasted with stable cellobiose consumption rates . This suggests that their improved performance was not directly driven by specialized taxa that could degrade the primary resource, but is best attributed to increasing efficiency of using the “common good,” cellobiose-derived glucose. Other biotic interactions might have also contributed to reduced energy waste, e.g., enhanced cross-feeding on secreted metabolites , or energy reallocation from metabolically costly competitive traits to growth yield due to reduced interspecific competition (decrease in α-diversity; ) . Our findings also speak for MacArthur’s minimization principle in communities developing under competition at stable conditions: unutilized resources decreased with community maturity due to niche complementarity among species . While this concept has received little attention , an experimental study using synthetic phytoplankton communities confirmed its predictions regarding biomass production . Our results extend these findings by showing that cellobiose-derived carbon was increasingly fixed into microbial biomass across growth cycles, irrespective of community structure . Finally, microevolutionary adaptation toward more efficient glucose consumption could also explain the increasing efficiency. Originally designed for assessing dyadic interactions within game tournaments , Elo-rating has been used in biology to assess the social structure in primates . Our implementation demonstrated suitability for metacommunity analysis by clearly highlighting Acidovorax as the overall “winner” across multiple communities . More importantly, it gave insight into subtle community re-arrangements during growth cycles that would have been challenging to detect without context-dependent measure, e.g., the increasing importance of Paucibacter , Bosea , or Pelomonas , and the concomitant decline of Flavobacterium , Pseudodarcicella , or Aquabacterium . Additionally, it was the best predictor for the performance of three of the top ten most abundant genera after metacommunity mixing . Thus, Elo-rating could be an additional tool to assess the overall success of taxa in metacommunities based on their competitive performance within and among local assemblages. Stochastic assembly processes can generate compositional and functionally distinct communities . We show that dispersal limitation within metacommunities may produce recurrent community types with different carrying capacities , evenness , and subsets of exclusively associated taxa . We defined types from genera dominating three or more microcosms ( Acidovorax , Pseudomonas, and Aeromonas ) . These genera are typically members of the rare aquatic biosphere that proliferate upon input of organic matter or in substrate-rich microniches . Acidovorax was initially seeded into all microcosms, and all local populations survived over the growth cycles. Since these bacteria lack a known cellobiose degradation mechanism , cellobiose-derived glucose must have been available to them as a “common good.” By contrast, both Pseudomonas and Aeromonas were dispersal-limited and more vulnerable to biotic selection. In general, the community types self-stabilized: while initial stochastic dispersal established the state for subsequent development ( , C0), the biological interaction cycles resulted in their deterministic “purification.” This led to stable or increasing within-type similarity against a background of increasing metacommunity-level β-diversity . Experimental homogenizing dispersal increased α-diversity and similarity (lower β-diversity) of local microcosm communities but led to a reduction of total metacommunity (γ) diversity . These observations do not align with the theoretical predictions for a fully connected metacommunity subjected to high dispersal rates . Thus, the effects of a singular coalescence event differ from the source-sink dynamics resulting from a continuous process of connectivity. Moderate local species sorting in post-coalescence microcosms was suggested by the large proportion of abundances of individual genera explained by the parent communities (i.e., by the neutral model, ). The appearance of novel positive interactions among previously allopatric populations may also have contributed to the increased local diversity . Since our experiment was limited to a single growth cycle after coalescence, we cannot assess if this high initial diversity was only temporary. A gradual loss of diversity over a 6-week period was demonstrated in an experimental study of mixed soil and carcass communities . Upon coalescence, the heterogeneous assemblages transitioned to novel, more uniform communities that differed from all source communities . Homogenization of synthetic bacterial communities has been observed already at low dispersal rates . Comparable findings have been reported from long-term field observations at the landscape scale: the anthropogenic connection of freshwater bodies (related to the construction of a reservoir) led to the homogenization of the zooplankton metacommunity . Coalescence also led to functional uniformity : community performance didn’t improve after mixing but instead stabilized around the median value of the parent communities . This contrasts with previous observations where the best-performing parent community dictated both, the structure and function of post-coalescence methanogenic assemblages . The loss of functional variability most likely resulted from the disproportional decline or extinction of functionally distinct taxa that dominated in single C6 communities and significantly contributed to this variability . These “endemic” populations, Rheinheimera , Duganella, and Caulobacter , proved to be extremely vulnerable to competitive exclusion . Thus, our experimental observations shed light on how homogenizing dispersal can affect species trait distributions and lead to a loss of functional variability at the metacommunity level , thereby potentially altering ecosystem functioning through the replacement of specialists at the expense of generalists and functionally inefficient species . The post-coalescence dominance of one genotype of Pseudomonas from a single isolated microcosm, P. azotoformans, conspicuously exceeded our predictions . The analysis of the corresponding MAGs revealed that P. azotoformans was the only abundant community member that featured T5SS and T6SS. These secretion systems confer competitive advantages to pseudomonads by delivering effectors such as nucleases, amidases, hydrolases, or phospholipases to neighboring bacterial cells and the external milieu . Taken together, our findings suggest that dispersal limitation may play a key role in defining community performance, by stochastically segregating highly efficient “bottom-up” specialists from taxa that outcompete them via negative biotic interactions . This holds relevance for a rational selection of stable microbial assemblages for both industrial and ecosystem restoration purposes . Specifically, we demonstrate the feasibility of a “top-down” design approach to optimize degradation efficiency in synthetic communities by producing rare variants that outperform the more common types: the highest levels of cellobiose degradation occurred in a unique dominated community stable over the six growth cycles (i.e., Caulobacter ; ) but did not survive community coalescence . Our findings thus provide a potential alternative to classical bottom-up approaches , by allowing for intrinsic biotic relationships from initial stochastic assembly to serve as stabilizing force during deterministic selection . Sampling site and experimental design Water for the inoculum of the experiment was collected at 5 m depth from the prealpine oligo-mesotrophic Lake Zurich (Switzerland) on 4 October 2019. It was prefiltered using 0.8 µm pore size filters (polycarbonate membrane, Whatman, Maidstone, UK) via a peristaltic pump (Ismatec, Wertheim, Germany) to exclude potential grazers and other eukaryotes from the microcosms. The experiment comprised three phases: initial colonization of sterile microcosm environments, six growth cycles in semi-continuous culture, and a final “coalescence event” . Bacterial communities were grown in artificial lake water (ALW) supplemented with glucose (10 µmol L −1 ) and its dimer, cellobiose (100 µmol L −1 ). This setup aimed to mimic the natural pool of dissolved organic carbon in aquatic systems, i.e., low concentrations of labile and higher concentrations of recalcitrant compounds . Microcosms consisted of 200 mL Erlenmeyer flasks incubated at 20°C in dark conditions. At the end of each growth cycle, samples for substrate utilization, bacterial growth, and biomass were collected. For the initial colonization phase (growth cycle 0, hereafter C0), the filtered lake water was inoculated into ALW (1:100), homogenized, and distributed over 20 microcosms. This procedure promotes the dispersal limitation of rare lake bacteria that thrive in the provided environmental conditions . Microcosms were incubated for 6 days until bacteria reached the stationary phase. For the semi-continuous cultivation phase, 20 mL from each microcosm was transferred into 180 mL of substrate-supplemented ALW in a new microcosm. These cultures were incubated for 4 days between subsequent transfers, for altogether six growth cycles (C1 to C6; ). In the final phase, 20 mL from each of the 20 communities were mixed to simulate homogenizing dispersal, diluted with ALW (1:10), homogenized, and distributed across 20 microcosms. These microcosms were incubated for 4 days (C7; ). Bacterial abundances and biomass For bacterial enumeration, 1 mL portions were fixed with formaldehyde (2% final concentration), stored at 4°C, and measured within 24 h. Fixed samples were stained with SYBR Green and analyzed on a CytoFLEX flow cytometer (Beckman Coulter, Indianapolis, IN, USA). For biomass determination, 50 mL aliquots were filtered onto precombusted 0.22 µm pore size GF/F filters (Tisch Scientific, 450°C for 6 h) and stored in small aluminum containers at −20°C until analysis. The total organic carbon was quantified on a dry combustion module cavity ring-down spectrometer (Picarro Inc, Santa Clara, CA, USA). Filters were combusted at 950°C, and the resulting CO 2 was quantified. Standards with a known C-content ( Miscanthus ) served as the reference for calibration. Substrate quantification Glucose and cellobiose concentrations were determined by high-performance liquid chromatography (1260 Infinity series, Agilent Technologies, Santa Clara, CA, USA) coupled with mass spectrometry (API 5000 triple quadrupole, AB Sciex, Baden, Switzerland; HPLC‒MS). Aliquots (1.5 mL) were filtered through 0.1 µm membrane filters (Polyethersulfone, Infochroma AG, Goldau, Switzerland) and stored at −20°C until analysis. Measurements were conducted as described , using sucralose (2 µmol L −1 ) as the internal standard. Data were acquired using Analyst v1.6.1 software (AB Sciex), and chromatograms were analyzed via MultiQuant v2.1 (AB Sciex). Carbon use efficiency CUE was calculated as the ratio between the biomass produced and the corresponding amount of carbon (combined concentrations of glucose and cellobiose) consumed during each cycle. DNA extraction At the end of cycles C0, C1, C4, C6, and C7, 100 mL from each microcosm were filtered onto a 0.22 µm pore size filter (GPWP, Millipore, Darmstadt, Germany), stored at −20°C until DNA extraction with the DNeasy PowerBiofilm Kit (Qiagen, Germany). Metagenomic DNA was sequenced using the Illumina shotgun NovaSeq 6000 platform (2 × 150 pb, NOVOGENE, Cambridge, United Kingdom). 16S rRNA genotyping For bacterial community structure analysis, we retrieved reads from the C0, C1, C4, C6, and C7 metagenome sequences mapped to the 16S rRNA gene using published pipelines . Forward and reverse reads were merged using BBmerge v38.86 at default settings and filtered by length (>200 bp) using BBduck v38.86 . These pre-processed reads were queried against the SILVA SSU database using MMseqs2 (e-value 1e −3 ) to identify RNA-like sequences. Bona fide 16S rRNA sequences were further compared by blastn (e-value 1e −5 ) with SSU-ALIGN v0.1 ( http://eddylab.org/software/ssu-align/ ) against the SILVA 99NR database v138.1 . OTUs were constructed by BLAST (v2.9.0) analysis of the identified 16S rRNA sequences against SILVA that simultaneously had identity values >97% and alignment lengths ≥80% . Reads were rarefied to the read count of the lowest sample (2,911; ). Genome assembly and functional annotation Raw Illumina reads were quality and adapter trimmed using BBduck v38.86 (qtrim=rl trimq=30). Reads were assembled per sample using MEGAHIT v1.2.9 (defaults settings, k-mer 29, 39, 49, 59, 69, 79, 89, 99). The metagenomic reads were mapped using BBmap v38.86 against the assembled contigs. The abundance profile of assembled contigs was used for binning with MetaBAT2 . Completeness and contamination were assessed by CheckM v1.2.2 . Bins with contamination <5% were considered MAGs for further analysis . MAGs were taxonomically classified with GTDB-tk v1.4.0 software against the GTBD database release R07-RS207. Coding sequences were predicted via Prokka v1.12 and annotated using the Kyoto Encyclopedia of Genes and Genomes (KEGG) . Metabolic reconstruction and genomic traits analysis were conducted with the KEGG mapping tools ( https://www.genome.jp/kegg/mapper/reconstruct.html ) using the previously annotated KO numbers. Proteins involved in cellobiose degradation (beta-glucosidases and cellobiose phosphorylases) we identified using the UniProtKB/Swiss-Prot protein database (release 2024_02 v1). Genomic traits associated with biological interactions We searched for genomic traits associated with positive and negative biological interactions. Amino acid biosynthesis pathways were analyzed to detect auxotrophic taxa. Auxotrophies for essential metabolites increase metabolic interdependencies within microbial communities, thereby promoting positive interactions . Genes associated with bacterial secretion systems were assessed as proxies for competitive advantages . Since some secretion systems can be associated with positive interaction (i.e., cell-cell communication), we operationally classified them into weak or strong competitive traits. Weak competitive traits comprised type I, II, and IV secretion systems (T1SS, T2SS, T4SS) that relate to host-pathogen interactions, and participate in bacterial genetic exchange . In contrast, strong competitive traits encompassed type III, V, and VI secretion systems (T3SS, T5SS, T6SS), which can provide a direct competitive advantage by enhancing survival and invasion capacity through the release of toxins and effectors into the environment or neighboring cells . Assessment of competitiveness and predictions for the homogenizing dispersal event Competition among genera in our low-complexity communities at C0, C1, C4, and C6 was assessed by a multiplayer version of the Elo-rating index used to compare the performance of players across multiple matches in gaming . For calculations, we used the multielo-package v0.4.0 implemented in Python ( https://github.com/djcunningham0/multielo ). Elo-ratings rely on the accuracy of a scoring function, for which we fitted an exponential decay function to the distribution of bacterial read numbers from cycle C0 to C6 ( R 2 = 0.90, P < 0.001; ). Each genus obtained a score (So G ) based on its ranking in the community. Subsequently, the Elo rating per genus (Elo G ) in each microcosm was calculated as (1) E l o G , n = E l o G , n − 1 + K ( N − 1 ) ( S o G − S e G ) , where K (default = 32) corresponds to the sum of points per microcosm after all pairwise “matches,” N is the number of bacterial genera, and Se G is the per genus expected score if all community members have the same winning probability. Elo-rating was calculated per cycle and sequentially updated through the 20 microcosms. Because the order of the microcosms can influence Elo-rating results, we randomized the order of microcosms ( n = 1,000), and the average Elo-rating was reported. The effects of the homogenizing dispersal event at the genus level were assessed by comparing whether the final proportions of genera in C7 were better predicted by competitiveness (their Elo-ratings in C6) or neutral processes (their respective abundances in the inoculum for C7). To calculate the competitive scenario, we first scaled the Elo-ratings in C6 by subtracting the minimum rating: (2) E l o G ′ = E l o G − min ( E l o G ) The scaled Elo-rating was normalized from 0 to 1 for further comparison with the observed abundances: (3) E l o . n o r m G ′ = E l o G ′ ∑ ( E l o G ′ ) × m e d i a n ( C e l l c o u n t s C 6 ) , where 𝐸𝑙𝑜.𝑛𝑜𝑟𝑚 G ′ represents the expected abundance of the individual genus after the homogenizing dispersal event according to their competitiveness. The expected abundance of the individual genus according to the neutral scenario was estimated by multiplying the relative read number per genus ( G ) with cell abundances per microcosm ( i ) from the C6. These results were summed up across microcosms ( n ), as follows: (4) A b u n d a n c e . n e u t r a l G = ∑ i n ( r e l a t i v e r e a d n u m b e r s G , i × c e l l c o u n t G , i ) Abundance.neutral G was normalized by the sum of cell counts across microcosms ( n = 20). Statistical analysis Statistical analyses were conducted using R . The modified Raup–Crick index (β RC ) was used to assess the importance of community assembly processes using the Bray‒Curtis distance at the OTU level with the R package NST . The β RC performs a pairwise evaluation of community turnover based on a null model in which taxa are randomly shuffled among all communities. It indicates whether community pairs are more (β RC < −0.95) or less (β RC > 0.95) similar than expected by chance, or if turnover does not differ from the stochastic assembly (|β RC | < 0.95). Community pairs more similar than by chance are expected to be influenced by either homogenizing dispersal or homogeneous selection, while community pairs less similar than by chance are influenced by dispersal limitation. Distinct community types were defined by genera that were both, abundant and prevalent , i.e., that had the highest read proportions of all genera at C6 in at least three microcosms. The average community composition of these types was derived from C6 microcosms. One-way repeated-measurement ANOVAs were performed to evaluate if community type affected alpha diversity (Richness, Pielou’s evenness) and bulk properties (bacterial abundance, biomass, and CUE). Normality and homoscedasticity were tested by Kolmogorov‒Smirnov and Levene tests, respectively. Linear mixed models were fitted to the bulk property values of community types from C0 to C6 to assess potential in or decrease during the semi-continuous cultivation phase (“lme” function, R package nlm). Spearman rank correlations of Elo-rating scores versus cycle number were used to test if the competitive performance of individual genera significantly changed between C0 and C6. One-sample Wilcoxon or one-sample t-tests (depending on data distribution) were performed to assess if the abundances of individual genera after the homogenizing dispersal event in the C7 microcosms ( n = 20) were more accurately predicted by the neutral or the competitive model (or by neither). The abundances predicted by either model were considered null hypotheses ( h 0 ). If neither predictor deviated from h 0 , the model yielding the higher P -value was selected. Genera with significantly higher or lower abundances than predicted by both models were classified as over- or underperforming, respectively. Multiple testing was adjusted for by the Benjamini-Hochberg method. Water for the inoculum of the experiment was collected at 5 m depth from the prealpine oligo-mesotrophic Lake Zurich (Switzerland) on 4 October 2019. It was prefiltered using 0.8 µm pore size filters (polycarbonate membrane, Whatman, Maidstone, UK) via a peristaltic pump (Ismatec, Wertheim, Germany) to exclude potential grazers and other eukaryotes from the microcosms. The experiment comprised three phases: initial colonization of sterile microcosm environments, six growth cycles in semi-continuous culture, and a final “coalescence event” . Bacterial communities were grown in artificial lake water (ALW) supplemented with glucose (10 µmol L −1 ) and its dimer, cellobiose (100 µmol L −1 ). This setup aimed to mimic the natural pool of dissolved organic carbon in aquatic systems, i.e., low concentrations of labile and higher concentrations of recalcitrant compounds . Microcosms consisted of 200 mL Erlenmeyer flasks incubated at 20°C in dark conditions. At the end of each growth cycle, samples for substrate utilization, bacterial growth, and biomass were collected. For the initial colonization phase (growth cycle 0, hereafter C0), the filtered lake water was inoculated into ALW (1:100), homogenized, and distributed over 20 microcosms. This procedure promotes the dispersal limitation of rare lake bacteria that thrive in the provided environmental conditions . Microcosms were incubated for 6 days until bacteria reached the stationary phase. For the semi-continuous cultivation phase, 20 mL from each microcosm was transferred into 180 mL of substrate-supplemented ALW in a new microcosm. These cultures were incubated for 4 days between subsequent transfers, for altogether six growth cycles (C1 to C6; ). In the final phase, 20 mL from each of the 20 communities were mixed to simulate homogenizing dispersal, diluted with ALW (1:10), homogenized, and distributed across 20 microcosms. These microcosms were incubated for 4 days (C7; ). For bacterial enumeration, 1 mL portions were fixed with formaldehyde (2% final concentration), stored at 4°C, and measured within 24 h. Fixed samples were stained with SYBR Green and analyzed on a CytoFLEX flow cytometer (Beckman Coulter, Indianapolis, IN, USA). For biomass determination, 50 mL aliquots were filtered onto precombusted 0.22 µm pore size GF/F filters (Tisch Scientific, 450°C for 6 h) and stored in small aluminum containers at −20°C until analysis. The total organic carbon was quantified on a dry combustion module cavity ring-down spectrometer (Picarro Inc, Santa Clara, CA, USA). Filters were combusted at 950°C, and the resulting CO 2 was quantified. Standards with a known C-content ( Miscanthus ) served as the reference for calibration. Glucose and cellobiose concentrations were determined by high-performance liquid chromatography (1260 Infinity series, Agilent Technologies, Santa Clara, CA, USA) coupled with mass spectrometry (API 5000 triple quadrupole, AB Sciex, Baden, Switzerland; HPLC‒MS). Aliquots (1.5 mL) were filtered through 0.1 µm membrane filters (Polyethersulfone, Infochroma AG, Goldau, Switzerland) and stored at −20°C until analysis. Measurements were conducted as described , using sucralose (2 µmol L −1 ) as the internal standard. Data were acquired using Analyst v1.6.1 software (AB Sciex), and chromatograms were analyzed via MultiQuant v2.1 (AB Sciex). CUE was calculated as the ratio between the biomass produced and the corresponding amount of carbon (combined concentrations of glucose and cellobiose) consumed during each cycle. At the end of cycles C0, C1, C4, C6, and C7, 100 mL from each microcosm were filtered onto a 0.22 µm pore size filter (GPWP, Millipore, Darmstadt, Germany), stored at −20°C until DNA extraction with the DNeasy PowerBiofilm Kit (Qiagen, Germany). Metagenomic DNA was sequenced using the Illumina shotgun NovaSeq 6000 platform (2 × 150 pb, NOVOGENE, Cambridge, United Kingdom). For bacterial community structure analysis, we retrieved reads from the C0, C1, C4, C6, and C7 metagenome sequences mapped to the 16S rRNA gene using published pipelines . Forward and reverse reads were merged using BBmerge v38.86 at default settings and filtered by length (>200 bp) using BBduck v38.86 . These pre-processed reads were queried against the SILVA SSU database using MMseqs2 (e-value 1e −3 ) to identify RNA-like sequences. Bona fide 16S rRNA sequences were further compared by blastn (e-value 1e −5 ) with SSU-ALIGN v0.1 ( http://eddylab.org/software/ssu-align/ ) against the SILVA 99NR database v138.1 . OTUs were constructed by BLAST (v2.9.0) analysis of the identified 16S rRNA sequences against SILVA that simultaneously had identity values >97% and alignment lengths ≥80% . Reads were rarefied to the read count of the lowest sample (2,911; ). Raw Illumina reads were quality and adapter trimmed using BBduck v38.86 (qtrim=rl trimq=30). Reads were assembled per sample using MEGAHIT v1.2.9 (defaults settings, k-mer 29, 39, 49, 59, 69, 79, 89, 99). The metagenomic reads were mapped using BBmap v38.86 against the assembled contigs. The abundance profile of assembled contigs was used for binning with MetaBAT2 . Completeness and contamination were assessed by CheckM v1.2.2 . Bins with contamination <5% were considered MAGs for further analysis . MAGs were taxonomically classified with GTDB-tk v1.4.0 software against the GTBD database release R07-RS207. Coding sequences were predicted via Prokka v1.12 and annotated using the Kyoto Encyclopedia of Genes and Genomes (KEGG) . Metabolic reconstruction and genomic traits analysis were conducted with the KEGG mapping tools ( https://www.genome.jp/kegg/mapper/reconstruct.html ) using the previously annotated KO numbers. Proteins involved in cellobiose degradation (beta-glucosidases and cellobiose phosphorylases) we identified using the UniProtKB/Swiss-Prot protein database (release 2024_02 v1). We searched for genomic traits associated with positive and negative biological interactions. Amino acid biosynthesis pathways were analyzed to detect auxotrophic taxa. Auxotrophies for essential metabolites increase metabolic interdependencies within microbial communities, thereby promoting positive interactions . Genes associated with bacterial secretion systems were assessed as proxies for competitive advantages . Since some secretion systems can be associated with positive interaction (i.e., cell-cell communication), we operationally classified them into weak or strong competitive traits. Weak competitive traits comprised type I, II, and IV secretion systems (T1SS, T2SS, T4SS) that relate to host-pathogen interactions, and participate in bacterial genetic exchange . In contrast, strong competitive traits encompassed type III, V, and VI secretion systems (T3SS, T5SS, T6SS), which can provide a direct competitive advantage by enhancing survival and invasion capacity through the release of toxins and effectors into the environment or neighboring cells . Competition among genera in our low-complexity communities at C0, C1, C4, and C6 was assessed by a multiplayer version of the Elo-rating index used to compare the performance of players across multiple matches in gaming . For calculations, we used the multielo-package v0.4.0 implemented in Python ( https://github.com/djcunningham0/multielo ). Elo-ratings rely on the accuracy of a scoring function, for which we fitted an exponential decay function to the distribution of bacterial read numbers from cycle C0 to C6 ( R 2 = 0.90, P < 0.001; ). Each genus obtained a score (So G ) based on its ranking in the community. Subsequently, the Elo rating per genus (Elo G ) in each microcosm was calculated as (1) E l o G , n = E l o G , n − 1 + K ( N − 1 ) ( S o G − S e G ) , where K (default = 32) corresponds to the sum of points per microcosm after all pairwise “matches,” N is the number of bacterial genera, and Se G is the per genus expected score if all community members have the same winning probability. Elo-rating was calculated per cycle and sequentially updated through the 20 microcosms. Because the order of the microcosms can influence Elo-rating results, we randomized the order of microcosms ( n = 1,000), and the average Elo-rating was reported. The effects of the homogenizing dispersal event at the genus level were assessed by comparing whether the final proportions of genera in C7 were better predicted by competitiveness (their Elo-ratings in C6) or neutral processes (their respective abundances in the inoculum for C7). To calculate the competitive scenario, we first scaled the Elo-ratings in C6 by subtracting the minimum rating: (2) E l o G ′ = E l o G − min ( E l o G ) The scaled Elo-rating was normalized from 0 to 1 for further comparison with the observed abundances: (3) E l o . n o r m G ′ = E l o G ′ ∑ ( E l o G ′ ) × m e d i a n ( C e l l c o u n t s C 6 ) , where 𝐸𝑙𝑜.𝑛𝑜𝑟𝑚 G ′ represents the expected abundance of the individual genus after the homogenizing dispersal event according to their competitiveness. The expected abundance of the individual genus according to the neutral scenario was estimated by multiplying the relative read number per genus ( G ) with cell abundances per microcosm ( i ) from the C6. These results were summed up across microcosms ( n ), as follows: (4) A b u n d a n c e . n e u t r a l G = ∑ i n ( r e l a t i v e r e a d n u m b e r s G , i × c e l l c o u n t G , i ) Abundance.neutral G was normalized by the sum of cell counts across microcosms ( n = 20). Statistical analyses were conducted using R . The modified Raup–Crick index (β RC ) was used to assess the importance of community assembly processes using the Bray‒Curtis distance at the OTU level with the R package NST . The β RC performs a pairwise evaluation of community turnover based on a null model in which taxa are randomly shuffled among all communities. It indicates whether community pairs are more (β RC < −0.95) or less (β RC > 0.95) similar than expected by chance, or if turnover does not differ from the stochastic assembly (|β RC | < 0.95). Community pairs more similar than by chance are expected to be influenced by either homogenizing dispersal or homogeneous selection, while community pairs less similar than by chance are influenced by dispersal limitation. Distinct community types were defined by genera that were both, abundant and prevalent , i.e., that had the highest read proportions of all genera at C6 in at least three microcosms. The average community composition of these types was derived from C6 microcosms. One-way repeated-measurement ANOVAs were performed to evaluate if community type affected alpha diversity (Richness, Pielou’s evenness) and bulk properties (bacterial abundance, biomass, and CUE). Normality and homoscedasticity were tested by Kolmogorov‒Smirnov and Levene tests, respectively. Linear mixed models were fitted to the bulk property values of community types from C0 to C6 to assess potential in or decrease during the semi-continuous cultivation phase (“lme” function, R package nlm). Spearman rank correlations of Elo-rating scores versus cycle number were used to test if the competitive performance of individual genera significantly changed between C0 and C6. One-sample Wilcoxon or one-sample t-tests (depending on data distribution) were performed to assess if the abundances of individual genera after the homogenizing dispersal event in the C7 microcosms ( n = 20) were more accurately predicted by the neutral or the competitive model (or by neither). The abundances predicted by either model were considered null hypotheses ( h 0 ). If neither predictor deviated from h 0 , the model yielding the higher P -value was selected. Genera with significantly higher or lower abundances than predicted by both models were classified as over- or underperforming, respectively. Multiple testing was adjusted for by the Benjamini-Hochberg method. |
Automatic Segmentation of the Jaws Used in Guided Insertion of Orthodontic Mini Implants to Improve Their Stability and Precision | a1f28f25-66aa-4acc-b926-f32e90cc88f7 | 11509293 | Dentistry[mh] | The area of medicine is undergoing a revolution due to the rapid developments in medical imaging. Cone-beam computed tomography (CBCT), intraoral and face scanners, dental 3D printing, and artificial intelligence are all examples of how quickly digital dentistry is developing. In medical image investigations, it is frequently necessary to differentiate or segment objects, organs, or structures from their surrounding background . The analysis of organs and structures in medical images is becoming increasingly significant in the field of diagnosis and in providing guidance for minimally invasive surgical and therapeutic interventions. For orthodontic diagnosis and treatment planning, it is essential to obtain precise segmentation of the jaw and teeth in CBCT scans . Traditional manual or automated thresholding-based methods, incorporated into commercial or open-source 3D software (Blue Sky Plan software version 4.12.13 (64-bit) (Blue Sky Bio LLC, Libertyville, IL, USA). applications, primarily drive the segmentation process that models these 3D surfaces . In order to facilitate its practical application, the segmentation process should be automated by software. One advantage of segmentation is its ability to generate 3D visible structures, facilitating the verification of the implant simulation’s site through the identification of its spatial surroundings. Micro-implants (OMIs), which are also called temporary anchorage devices (TADs), mini implants, or mini screws in the field of orthodontics, have been used to make complicated orthodontic movements possible . Numerous orthodontic operations that were previously used to manage anchoring have been made simpler by the orthodontic mini screws that give a skeletal anchorage, and the side effects of many orthodontic appliances have been thus minimized . Although these mini screws were originally inserted manually, thanks to the benefits of digitization, especially in particularly difficult cases, guided insertion is now used . Previous studies have assessed the precision of computer-guided mini implant insertion . Bae et al., in a study using cadaver jaws, discovered that 20% of the direct implant insertions resulted in contact with the roots; in contrast, there were no incidences of this in the guided implant placement group, as shown in . The accuracy results show that, in comparison to implants inserted manually, implants inserted with guidance systems offer significantly better precision, as reported in computer-guided implantology studies . Even in the hands of less skilled medical professionals, CAD/CAM templates enable more accuracy and consistency in the insertion of mini screws, as well as improved orientation and depth of the cortical insertion. This can lead to a shorter recovery period following surgery and a lower chance of damage to adjacent anatomical tissues, all of which can improve patient comfort . While the use of surgical guides contributes to the accuracy of mini implant placement, the stability of these implants also depends on a qualitative examination of the bone structures present at the insertion site, including the identification of bicortical and tricortical structures situated where the vestibular wall of the jaw and the maxillary sinus floor meet or where the nasal floor meets the maxillary sinus floor . Additionally, for maxillary distalization, the infrazygomatic ridge area is the place where the mini implants are placed, particularly in cases of class II malocclusion . The spaces in the jaw between the two premolars and the space between the second premolar and the first molar are the most often used locations for mini implant insertion . When considering the utilization of mini implants in the mandible for the purpose of distalizing the molars, it has been determined that the most advantageous location for implantation is situated medially and distally to the second molar . This study aims to examine three cases, two in the maxilla and one in the mandible, with the help of automated segmentation. A study model is utilized to simulate the mini implant with the help of in space visibility using 3D rendering, offering an exhaustive examination of the interactions between the mini implant and the surrounding structures. In the first two cases, we will examine the maxilla in the mesial and distal regions of the second premolar. The mesial position relative to the second premolar anatomically corresponds to the location of the intersection between the nasal cortical and the palatal cortical at the mini implant’s tip. In this particular case, our objective is to determine the region of intersection between the two cortical structures in order to properly situate the implant in that location, which improves its stability. The position distal to the second premolar corresponds to the intersection between the maxillary sinus cortical and the maxillary vestibular cortical at the mini implant’s tip. Additionally, in this case, we will look for the point where the two corticals connect, to provide stability. Additionally, we will look for a safety zone where the mini implant passes between the two roots of the corresponding neighboring teeth, at least one millimeter from each one . In the mesial zone of the second molar, the third case will be located on the mandibular buccal shelf , as this insertion place was determined to be the most suitable for the mandible . A variety of software applications to automate segmentation have been developed . Specifically for the dental field, we identified two automatic segmentation software applications: BlueSkyPlan version 4.12.13 (64-bit) (Blue Sky Bio LLC, Libertyville, IL, USA) and Diagnocat (Diagnocat Inc., San Francisco, CA, USA) . In contrast to the second, which requires payment for segmentation, we have utilized the first one, since it provides automatic segmentation for free and is open source . An inherent constraint of the existing technique employed for simulating the placement of orthodontic mini implants is its capability to conduct this simulation exclusively within the sections of the 3D image in Multi-Planar Reconstruction (MPR) visualization, instead of in the spatial domain within the 3D rendering interface. The number of sections in the imaging software is limited. Typically, the mentioned orientations include the sagittal, coronal, axial, cross-sectional, and panoramic orientations. If the implant is simulated in certain previous sections, its orientation will be limited due to the absence of oblique planes, where the mini implant may be rotated in these sections. Only the possibility of its spatial orientation in all three planes gives it the security of insertion where it is desired and where it meets the conditions of stability and precision. The practical method proposed in this study solves this issue by providing the potential for unrestricted spatial orientation of the mini implant.
We conducted the present study using the Blue Sky Plan software version 4.12.13 (64-bit) (Blue Sky Bio LLC, Libertyville, IL, USA). The workflow diagram is shown in . We first selected the Model Master module. Then, we selected the “Import Patient CT Scan” option and imported the patient’s DICOM image into the application. The DICOM 3D files utilized in this work have a field of view of 5 × 8 cm, a slice thickness of 0.2 mm, and voxel dimensions of 0.2 × 0.2 × 0.2 mm. The acquisition was performed using an X-ray tube current of 8 mA and a KVP of 89, utilizing the PaX-i3D model from Vatech Company Ltd. (Hwaseong, Republic of Korea). The software manufacturers do not supply the grayscale threshold values utilized for segmentation. These are not adjustable to different values during the start of the automatic segmentation command. If we believed that aligning the datasets was essential to obtain an occlusal plane closer to the horizontal for the reslice function, we did so. Then, we selected the “Automatic Jaw Segmentation” feature from the Tools menu after closing the Wizard that opens in the subsequent window; next, selected the mandible or maxilla to be segmented; and then clicked the button “Start Automatic Segmentation”. The operating system setting determines how long segmentation takes, but it usually takes two to three minutes. We clicked the “Create surface” button once the automatic segmentation was complete. The quality of segmentation in the MPR sections can be checked by modifying the segmented image using the inflate and deflate functions provided by the software by adding or removing a voxel on the entire segmented contour. We navigated to the Advanced tab from the Model Master tab to simulate mini implants, and selected the type of mini implant from the Implant Library or used a custom generic implant that has the same dimensions as the original model in case the corresponding model was not available. The dimensions of the customized implant can be as close to those of the real implant as possible, with a cylindrical or conical forms. Generic implants, created with software slightly larger than original implants, offer increased safety at the insertion site. One limitation is that the lateral spaces between the roots and the mini implant are on the order of tenths of millimeters, and when the mini implants are simulated, it is frequently required to know the real dimensions of these spaces. Thus, compared to generic customized implants, the original implants from the Implant Library are more appropriate for simulating implants. Once the implant has been positioned at the desired place in 3D rendering, the jaw can be hidden, and only the teeth and the implant become visible in 3D renderings, thanks to the separate teeth viewing feature. The software’s placement capability allows the implant to move freely in three dimensions in any direction, allowing for the selection of the most advantageous position in relation to the available spaces and distances from nearby roots. The implant simulation can be corrected from the 3D rendering window in cases where there is an excessively tight closeness between the implant and the tooth roots, thanks to the 3D imaging of the segmented structures. It is more difficult to visualize the spaces between the roots and the mini implant in the MPR view because of the varied angles of the root axes. These spaces should not be, ideally, smaller than 0.6 mm, to avoid contact with neighboring roots .
From the CBCT DICOM file, a, we can obtain unique STL files for each tooth, as well as for the mandible and maxilla, or for the mandibular canal, by using automated segmentation in BlueSkyPlan software (Blue Sky Bio LLC., Libertyville, IL, USA). They can be studied either together or separately ( a–d). In the first case of this study, we could locate the ideal position for the simulated mini implant between the two premolars, which was equally spaced from the surrounding roots, and acquire the bicortical structure made up of the sinus maxillary floor and the nasal floor. The buccal side, the palatal side of the mini implant, and the two premolars are shown a,b, respectively. In c, the STL file of the maxilla is sectioned with a plane that contains the mini implant’s axis to verify the image of the biortical position. d provides evidence supporting the positioning of the mini implant tip within the bicortical area, in the cross-sectional section. The mini implant’s angle of 57 degrees with the horizontal is within the range studied in the article by Wang et al. . In the second investigated case, which refers to the placement of a mini implant between the first molar and the second premolar, the circumstances correspond with those of the initial case in relation to the situations observed during the implant simulation. a displays a section that crosses the axis of the implant, revealing the distance from the neighboring roots exceeding the threshold of 0.6 mm. b provides illustrations of the implant’s movement in three-dimensional space during his movement to determine the ideal simulation position. In c, the maxilla STL file is sectioned with a plane containing the mini implant’s axis to confirm the rendering of the bicortical position. d demonstrates that the mini implant tip is positioned within the bicortical area in the cross-sectional section. The mini implant’s angle of 45 degrees with the horizontal is within the range analyzed in the publication by Wang et al. . In the third case, with the insertion of the mini implant between the first and second mandibular molars, the aim was to avoid the mandibular canal and simulate the mini implant, keeping away from the path of movement of the molars towards distalization . It is important to show caution when using automatic segmentation by verifying the boundaries of the jaw bone and the outside boundaries of the teeth in the MPR software sections in order to ensure that the segmented contours closely correspond to the real dimensions observed solely within the sections. illustrates the segmentation of the same mandible using two distinct software applications. The segmentation in a was performed using BlueSkyPlan software (Blue Sky Bio LLC, Libertyville, IL, USA), while the segmentation in b was performed using Diagnocat software (Diagnocat Inc., San Francisco, CA, USA). We imported the STL file of the segmented mandible from Diagnocat (Diagnocat Inc., San Francisco, CA, USA) into the Blue Sky Plan software (Blue Sky Bio LLC, Libertyville, IL, USA) to achieve overlap of the two segmented mandibles. We successfully completed the desired match by utilizing the automated overlay function of the Blue Sky Plan software (Blue Sky Bio LLC, Libertyville, IL, USA), as shown in c. The cross-sectional section presented in d reveals that there are no major differences seen in the cortical area. The distinctions between them are evident only at the alveolar process, where the vestibular bone shows a notably thin structure. The boundaries of the first segmentation made in the BlueSkyPlan (Blue Sky Bio LLC, Libertyville, IL, USA) are highlighted in orange in the cross-sectional section, while the second segmentation from the Diagnocat (Diagnocat Inc., San Francisco, CA, USA) is highlighted in green.
Many techniques to semi-automatically segment different anatomic structures in CBCT scans have been proposed in the past few decades . The constantly progressing software providing automatic segmentation addresses the necessity for mini implant insertion precision in very small locations. Among these automated techniques are statistical shape models, morphologic snakes, random forests, region seeding, edge detection, and watershed segmentation. In the research paper by Wang et al., they stated that there is still a lack of fully automated segmentation methods that can simultaneously segment both anatomic structures in CBCT scans (i.e., multiclass segmentation) . A lot of research has been conducted using convolutional neural network segmentation , manual segmentation , or semi-automatic segmentation from CBCT images. The process of segmentation, other than automatic methods, is time-consuming, necessitates the use of complex applications and a high level of knowledge for its operation, and is also expensive . These images can be simply examined in terms of their proximity to neighboring structures or their connections with other anatomical processes. It is much harder to see these spaces between the roots and the mini implant in the MPR view because of the varied angles of the root axes. The necessity to improve the precision of simulating mini implant insertion, particularly in small spaces, has led to research efforts focused on their insertion utilizing virtual navigation systems. The utilization of augmented reality technology in navigation procedures has been found to have a significant impact on the precision of orthodontic self-drilling mini implant placement, leading to a reduction in intraoperative problems when compared to the traditional free-hand technique . Contact between the mini screws and the tooth root next to the insertion site was one of the most frequently reported problems . The quantity and quality of bone, as well as the location of the mini implants in the bone, all affect the manner in which the primary stability of the implants occurs. This stability can be maximized by considering the thickness and properties of the bone when choosing the implantation site . In studies about inserting mini implants in the palate, artificial intelligence and automatic segmentation have been used . The differences among various techniques employed by the operators for the insertion of orthodontic mini implants relates to the utilization of the conventional freehand approach versus the guided procedure that incorporates surgical guidance. Three-dimensional images produced by automatic segmentation can be rapidly and simply used as study models. This article names only two software programs that perform automatic segmentation due to a lack of information on others, particularly those that offer free and user-friendly automatic segmentation for doctors. The BlueskyPlan software (Blue Sky Bio LLC, Libertyville, IL, USA) is an open-source application combining multiple modules for the manufacturing of orthodontic aligners, surgical guides, cephalometric analyses, and dental crowns and bridges. It remains free until the user chooses to export a set of aligners or surgical guides. The Diagnocat software (Diagnocat Inc., San Francisco, CA, USA) operates on a platform that allows users to upload a 3D DICOM dataset, which then performs automatic segmentation based on the selected criteria: maxilla, mandible, teeth, etc. The Diagnocat software is not free; a fee is charged for each segmentation. Most software options require advanced computer skills for effective automatic segmentation. The segmentation method used for this study is reproducible, as it utilizes an algorithm from automated segmentation software that necessitates no supplementary configurations for running. The automatic segmentation operates using default commands and non-configurable commands, resulting in identical segmentation outcomes. The method is applicable in everyday situations; however it necessitates a medium-term learning curve concerning the movement of implants, bones, or teeth as objects in augmented reality. The method is accessible to anyone at no cost. This article only allowed for a presentation of the concept. Among the shortcomings of this method are the following: a medium-term learning curve for operators unfamiliar with handling 3D rendering volumes, the necessity of equipping computers with processors capable of performing segmentation in a relatively brief period, and the requirement for full-capacity operation of resource-intensive software dependent on computer specifications.
1. Utilizing open-source software that offers free automatic segmentation is quite helpful for studying 3D CBCT DICOM images. 2. Placing an implant in a 3D rendering view significantly decreases the limitations associated with simulating its insertion within the software’s available section plans. 3. Using this method, the implant and the tooth may be studied as two three-dimensional objects whose positions can be changed in accordance with different needs. 4. When comparing the evaluation of the simulation of implants in 3D rendering images and their visualization in MPR to the three examples provided in this research, the volumetric visualization method provided here is superior. The automatic segmentation presented in this study topic is useful for those who use software for creating surgical guides for the insertion of orthodontic mini implants, as well as for medical professionals who wish to use the 3D segmentation images as a study model to create an appropriate treatment plan. The automatic segmentation performed by the computer using the software is substantially preferable than semi-automatic or manual segmentations regarding the time resources available to the physician during medical procedures.
|
Editorial: TCM approaches in cellular endocrinology | 51b498d4-2499-4e99-95a2-7621d6db5c16 | 10292213 | Physiology[mh] | The author confirms being the sole contributor of this work and has approved it for publication. |
How could metabolomics change pediatric health? | 0b4654cb-062f-4897-8878-f36016876ed1 | 7099833 | Pediatrics[mh] | In the last years, “Omics” technologies (including genomics, transcriptomics, proteomics and metabolomics) gained popularity and completely modified the scientific approach, due to their role in the description of complexes biological systems, through the simultaneous and usually noninvasive analysis of a large amount of data. Among such tools, metabolomics represents a growing and expanding scientific discipline and a promising area of research. Metabolomics (also known as Metabonomics), and the word metabolite, share as their root the ancient Greek word, metabolì , meaning ‘ change ’. Metabolomics, an innovative analytical profiling technique, investigates and detects the whole and comprehensive set of molecules of low molecular weight (including sugars, lipids, small peptides, vitamins and amino acids) present in cells, tissues, organs, and biological fluids, portraying their phenotypes. Thus, the global qualitative and quantitative analysis of all metabolites in biological fluids or tissues is highly useful in human health, allowing the evaluation of specific and individual metabolic responses to many pathophysiological stimuli including drugs, environmental changes, lifestyle, diseases and other epigenetics factors . Metabolomics, detecting broad classes of metabolites, provides a comprehensive functional phenotype integrating clinical features and genetic and non-genetic factors, such as environment, metabolites from symbiotic organisms including gut microflora and xenobiotics . Currently, more than 28.600 articles dealing with metabolomics can be found on PubMed. It is considered a technology that will contribute the world change and, in 2018, 1 Euro out of 18 spent in medicine research were intended for metabolomics. The human being should be evaluated in a holistic way, and this could be only performed through the integration of genome, transcriptome, proteome, metabolome, microbiome, fungome, virome, bacteriofagome, epigenome, exposome, phenome, diseasome and more. These topics represents the evolution of scientific technologies applied to medicine and clinical research. In the last decades, intuition-based medicine was replaced by the evidence based- and precision-based medicine. The holistic view of humans as a system biology, the integrate approach and “omics” technologies played a great role in such transition . The main advantages of such technique are a high diagnostic power (detecting changes in metabolites’ concentration and dynamic changes during time), a high rapidity (detecting changes occurring in seconds instead of minutes or hours), and easiness. Moreover, it is a “holistic” (the same metabolite in different samples or tissues at the same time from the same platform) and less expensive science (excluding the initial cost of the instrumentation). On the contrary, the traditional laboratory methodologies offer markers of diseases often showing a low sensitivity or a late appearance. Metabolic fingerprints can be generated by several analytical techniques, including nuclear magnetic resonance spectrometry ( 1 H-NMR), gas chromatography mass spectrometry (GC-MS) and liquid chromatography mass spectrometry (LC-MS), providing metabolic information potentially related to physiological states or pathological conditions of an organism. In example, variations of these fingerprints determined by changes in concentration of specific metabolites can allow the early detection of the onset of diseases. A strength point of metabolomics is the detection of rapid daily variability in the metabolic fingerprint, representing a metabolic “snapshot.” The final phenotype is determined by environmental effects, like diet, age, lifestyle, drugs or diseases, on the individual genome, giving origin to a unique transcriptome, proteome and a highly sensible metabolome . Metabolomics is what really happened, since metabolites are the final products of the interactions of genes, RNAs and proteins, while genomics represents what could potentially happen and proteomics is the picture of what is happening. It could be affirmed that metabolomics approach is the equivalent of investigating in a personal dumpster. Innovative, interesting and promising fields of metabolomics applications are the investigation of physiological status and the diagnosis of a disease, the identification of perturbed pathways due to disease or treatment, the evaluation of the response to drugs (Pharma-metabolomics) and the monitoring of the effects of nutrition (Nutri-metabolomics), the discovery of new and specific biomarkers, the classification of different phenotypes and functional genomics, and, finally, the characterization of natural or artificial products’ composition. Metabolomics also allows the study of disease phenotypes, molecular pathophysiology and cellular metabolism through metabolic profiling. Big data, provided by ‘omics’ tools, analyzed through the technology of machine intelligence, have also been named the black box of medicine, providing the so called “artificial intuition”. Machine techniques can decode the high volume of data generated from a limited number of subjects, comparing them to traditional analytic approaches . From a drop into the ocean (old biochemistry) to an ocean into a drop (metabolomics). In fact, the modern analytical technologies allow the identification of patterns that confer significantly more information than the measurement of a single parameter, as a bar code contains more information than a single number. Many authors recently evidenced the strong correlation between metabolomics and clinical research . In metabolomics experiment, the setting, the patients and the types of samples should be carefully chosen. Samples require a correct storage and, after processing, undergo a multivariate statistical analysis whose results are interpreted (Scale free networks) to individuate significant associations or formulate new hypothesis. Samples that can be analyzed in prenatal and perinatal metabolomics analysis are collected from mothers (amniotic fluid, placenta, blood, urine, breast milk, erythrocytes, hair, and vaginal secretions) or from the neonate (urine, blood, saliva, broncho-alveolar fluid, exhaled air condensate, stools, and umbilical cord) . In metabolomics, the complex systems are analyzed using a scale-free topology. Technically, networks expand through the addition of recent vertices, and vertices attach preferentially to well-connected sites. Scale-free networks can be exactly determined using vital features such as a disease or a patient. This allows the comprehension of complex metabolomics systems . Metabolic networks can correlate interlinking metabolites, revealing novel key pathways and the scale free network is useful to identify unexpected pathophysiological mechanisms. A metabolomics fingerprint (a reflection of the whole metabolome in a biofluid) characterizing a single sample constitutes such a strong characteristic of each subject as to allow its identification with 100% probability. It can be obtained through statistical analysis performed on NMR or GS spectra of multiple samples and points out an invariant part characteristic of each subject. Metabolomics fingerprints can be obtained in a fast, untargeted, and highly reproducible manner, detecting with high sensitivity information regarding the smallest concentration changes of several metabolites at the same time. Thus, an individual metabolic phenotype exists and is the key point of metabolomics clinical studies. However, changes related to pathological stimuli may be difficult to distinguish from physiological variations . In conclusion, metabolomics could help in the optimization of individualized therapy and nutrition; assessing drug-related efficacy or toxicity, identifying phenotype changes associated to disease onset and progression, improving early diagnosis and prognosis . A very interesting and innovative application of metabolomics is sportomics . Metabolomics could improve efficacy precision and accuracy, opening the way to one-person trials, with high efficacy and low toxicity . Currently, a clear and well-defined correlation between each metabolite and the related clinical meaning is not available, even if we are working at the creation of specific atlas dealing with metabolites involvement in several pediatric and neonatal diseases and conditions .
Below, we report only few clinical insights obtained through metabolomics application. Heart Metabolomics seems promising in cardiovascular disease, the leading cause of death and a major cause of disability worldwide including myocardial ischemia, infarction and coronary heart disease. Metabolomics could help in the identification of biomarkers able to early detect risks of such diseases before clinical signs, allowing prevention and early intervention that could prevent fatal consequences . That gut microbiota–host crosstalk seems the gap between cardiovascular risk factors, diet, and cardiovascular residual risk, via translocation through the intestinal barrier . An extensive review on such topic was recently published . Metabolites involved in cardiovascular risk are often related to gut microbiota; i.e., trimethylamine N-oxide (TMAO), has recently been linked to atherosclerosis and thrombosis rate increase. TMAO levels seem to correlate with the risk of cardiovascular events in patients with prior ischemic stroke, via the increase of proinflammatory monocytes . As reviewed by Chalkias and co-workers, metabolomics could be also applied in cardiac arrest, allowing the detection of metabolites linked to cardiac metabolism that could result useful biomarkers in the assessment of an increased risk of cardiac arrest and potentially improving the prevention and treatment of such condition . Auditory organ Moreover, in the idiopathic sudden sensorineural hearing loss, a frequent emergency whose aetiology is still unknown, metabolomics seems helpful in the early prediction of clinical outcome and therapy response to steroids (avoiding overtreatment of non-responders). A recent study evaluated through 1 H-NMR the urinary metabolome of patients with idiopathic sudden sensorineural hearing loss, and analyzing it according to the clinical outcome after steroids. Among evaluated subjects, a group was composed by healthy controls, a group by patients who did not recover from hearing loss after steroids and finally, patients who recovered after treatment were included. Urinary metabolome resulted significantly different between responders and non-responders, whith B-Alanine, 3-hydroxybutyrate and TMAO higher, and citrate and creatinine lower in the second group . Kidney In the field of pediatric nephrourological diseases, in 2010, urinary metabolome (with 1 H-NMR) between n = 21 children affected by nephrouropathies (including renal dysplasia, vesico-ureteral reflux, urinary tract infection, and acute kidney injury), and n = 19 healthy children was evaluated. As result, samples belonging to these two groups showed a clear and significant separation. Thus, metabolomics seems a promising, non-invasive tool in nephrourological diseases . Moreover, it seems possible to apply metabolomics in the early prediction of Chronic Kidney Disease in healthy adults born of extremely low birth weight (ELBW) . Some papers were published in this field, investigating metabolomics in pediatric and adult nephrology . Moreover, the role of urinary neneutrophil gelatinase-associated lipocalin (uNGAL) and kidney injury molecule-1 (KIM-1) as predictors of kidney injury severity was studied . Cancer Promising results have been also obtained through metabolomics application in oncology , even in pediatrics. Such technology could help in disease characterization, monitoring and therapeutic management . Finally, metabolomics can point out markers of hypoxic metabolism in cancer cells .
Metabolomics seems promising in cardiovascular disease, the leading cause of death and a major cause of disability worldwide including myocardial ischemia, infarction and coronary heart disease. Metabolomics could help in the identification of biomarkers able to early detect risks of such diseases before clinical signs, allowing prevention and early intervention that could prevent fatal consequences . That gut microbiota–host crosstalk seems the gap between cardiovascular risk factors, diet, and cardiovascular residual risk, via translocation through the intestinal barrier . An extensive review on such topic was recently published . Metabolites involved in cardiovascular risk are often related to gut microbiota; i.e., trimethylamine N-oxide (TMAO), has recently been linked to atherosclerosis and thrombosis rate increase. TMAO levels seem to correlate with the risk of cardiovascular events in patients with prior ischemic stroke, via the increase of proinflammatory monocytes . As reviewed by Chalkias and co-workers, metabolomics could be also applied in cardiac arrest, allowing the detection of metabolites linked to cardiac metabolism that could result useful biomarkers in the assessment of an increased risk of cardiac arrest and potentially improving the prevention and treatment of such condition .
Moreover, in the idiopathic sudden sensorineural hearing loss, a frequent emergency whose aetiology is still unknown, metabolomics seems helpful in the early prediction of clinical outcome and therapy response to steroids (avoiding overtreatment of non-responders). A recent study evaluated through 1 H-NMR the urinary metabolome of patients with idiopathic sudden sensorineural hearing loss, and analyzing it according to the clinical outcome after steroids. Among evaluated subjects, a group was composed by healthy controls, a group by patients who did not recover from hearing loss after steroids and finally, patients who recovered after treatment were included. Urinary metabolome resulted significantly different between responders and non-responders, whith B-Alanine, 3-hydroxybutyrate and TMAO higher, and citrate and creatinine lower in the second group .
In the field of pediatric nephrourological diseases, in 2010, urinary metabolome (with 1 H-NMR) between n = 21 children affected by nephrouropathies (including renal dysplasia, vesico-ureteral reflux, urinary tract infection, and acute kidney injury), and n = 19 healthy children was evaluated. As result, samples belonging to these two groups showed a clear and significant separation. Thus, metabolomics seems a promising, non-invasive tool in nephrourological diseases . Moreover, it seems possible to apply metabolomics in the early prediction of Chronic Kidney Disease in healthy adults born of extremely low birth weight (ELBW) . Some papers were published in this field, investigating metabolomics in pediatric and adult nephrology . Moreover, the role of urinary neneutrophil gelatinase-associated lipocalin (uNGAL) and kidney injury molecule-1 (KIM-1) as predictors of kidney injury severity was studied .
Promising results have been also obtained through metabolomics application in oncology , even in pediatrics. Such technology could help in disease characterization, monitoring and therapeutic management . Finally, metabolomics can point out markers of hypoxic metabolism in cancer cells .
Fetal life and perinatal period are crucial phases for neonatal development. The triggers and the conditions to which the fetus is exposed represent essential factors influencing the development of the newborn. In fact, as highlighted by the concept of perinatal programming, when a developing organism is exposed to specific intrauterine conditions, including excessive or inadequate nutrition, puts into practice several adaptive mechanisms and responses potentially modifying its development trajectory; therefore, in such window of vulnerability (or opportunity), persistent short- and long-term effects on newborn phenotype are performed. Thus, intrauterine stimuli and factors that are present during the early perinatal life can affect fetal and neonatal development and lead to negative consequences . In the evaluation of pre- and peri-conceptional factors, preterm birth and labor, intrauterine growth restriction (IUGR), maternal gestational diabetes (GDM), pre-eclampsia (PE), fetal infections, exposition to hyperoxia during post-natal life, potentially related to epigenetic changes on DNA , metabolomics application could give great advantages, both in the comprehension, diagnosis and treatment of such conditions and represents a promising field to investigate . The early diagnosis of the prediction of several maternal complications represents a challenge due to the high complexity of these conditions, still partially understood . Moreover, recent evidence highlights the pivotal role played by the placenta during fetal life. Such structure can be defined a metabolic interface between the mother and the fetus, influencing neonatal maturation and metabolism . Several metabolomics studies are currently available on common pregnancy complications. Step forwards, in the field of obstetrics, could improve pregnancy management and delivery assistance, with positive effects on maternal and fetal health, the reduction of costs, cesarean sections and hospitalization. Thus, we report some examples of metabolomics application in Human cytomegalovirus (HCMV) congenital infection, maternal obesity and preterm labor. HCMV, a potentially fatal viral infection during pregnancy, seems to alter metabolic profile of amniotic fluid (AF), in relation to maternal and fetal response to such infection. As evidenced, through GC-MS analysis of AF, different profiles characterize pregnant women that transmitted HMCV infection to their fetuses ( n = 20), instead of mothers who acquired but not transmitted the virus ( n = 20) and healthy controls ( n = 23). Moreover, AF samples from mothers whose neonates resulted symptomatic at birth ( n = 9) were clearly separated from AF belonging to neonates who acquired HCMV infection but did not show clinical signs, especially regarding metabolites related to fatty acids biosynthesis . In conclusion, metabolomics could describe maternal and fetal status in congenital HCMV infection, allowing an early diagnosis and an accurate management , and, in our opinion, it could be also applied in the evaluation of breast milk acquired CMV infection . In the field of maternal obesity (predisposing to pregnancies and fetal complications and potentially impairing neonatal outcome), we performed the first metabolomics study evaluating placental samples in normal weight ( n = 20) and obese ( n = 18) pregnant women, often affected by GDM as comorbidity. Through GC-MS, significant differences in antioxidant metabolites, nucleotide production, lipid synthesis and energy production were detected. In detail, obese mothers’ placentas also showed a peculiar fatty acids profile, related to an increase in placental metabolism and potentially reflecting intrauterine changes responsible of later diseases, including metabolic syndrome and cardiovascular disease . Metabolic pattern linked to the onset of labor, potentially detected by metabolomics tool, were pointed out in another study. Urine samples ( n = 59) were collected from a group of full-term pregnant women before and after the onset of labor and their metabolic discriminating features were pointed out through GC-MS and 1 H-NMR. As result, 18 metabolites allowed the discrimination between urinary samples of women in labor and not in labor (NL). In detail, glycine, alanine, acetone, 3-hydroxybutiyric acid, 2,3,4-trihydroxybutyric acid and succinic acid, characterized the late phase of labor. Thus, metabolomics could also help in discriminating urine samples from women in labor, offering a precocious screening of the onset of such condition . Another obstetric complication is represented by the premature rupture of membranes (PROM), defined as the fetal membranes’ rupture prior than the onset of labor. It can determine an increased rate of infections (such as chorioamnionitis and endometritis) and other complications for the mother and the fetus itself (abnormal fetal presentation, neonatal sepsis, intra-ventricular hemorrhage). PROM can occur at any gestational age (GA) and is often related to premature birth. Thus, PROM diagnosis should be promptly recognized and early and adequately managed (avoiding unnecessary antibiotics), reducing potential risks. However, sensible biomarkers are still lacking. Urinary maternal metabolome with CG-MS, describing interesting pathways associated with PROM and preterm labor, was therefore investigated. A total of n = 38 urinary samples was collected out of a group of n = 38 full-term pregnant women, divided into three subgroups: In the first, n = 11 women without PROM and labor, in the second, n = 10 pregnant women with PROM without labor, in the third, n = 17 pregnant women with PROM and labor were included. As result, the reduction of 9 metabolites resulted significantly associated with PROM (galactose, uric acid, 3,4-dihydroxybutyric acid, galactitol, alanine, lysine, 4-hydroxyphenylacetic acid, serine, and hydroxyproline dipeptide). Moreover, 60 metabolites significantly varied between the second and the third group. Most of them were higher in the group with PROM and labor, while phosphate, lactose, and uric acid were higher in the group with PROM without labor; among the increased metabolites in this group, 3,4- dihydroxybutyric acid is an intermediate of fatty acids oxidation (increasing in infections to provide energy), glucuronic, gulonic, glucaric, and gluconic acids are related to oxidative stress, cis-Aconitic acid (intermediate of tricarboxylic acid cycle), faces the increase in energy demand . Regarding spontaneous preterm birth, a still partially characterized cause of neonatal mortality and short- and long-term morbidity, metabolomics seems promising in the identification of sensible biomarkers . Potential long-term effects of preterm birth were also evaluated through a metabolomics study in adult patients (mean age 24 years), in which different metabolic urinary profiles were observed in subjects who were born full-term if compared to those showing ELBW . Recently, in a preliminary investigation, a different urinary metabolome was detected in relation to birth modality, in a sample collected in the first hours of life, underlining how this factor could influence neonatal metabolism, organogenesis and determine long-term effects. Full-term neonates born by vaginal delivery, if compared with neonates born by cesarean section, showed higher levels of dicarboxylic acids and Krebs cycle-related metabolites in neonates, probably due to differences in fatty acid oxidation, thermoregulation at birth or energy metabolism. Moreover, bacterial-related metabolites also showed some variations, in relation to a different microbiota colonization according to delivery mode . Finally, also in research in the field of the great obstetrical syndromes (PE, GDM, and IUGR) metabolomics could give a great contribution, improving prevention, early diagnosis, and monitoring, as reported by many authors . PE, a hypertensive gestational disorder originating in the placenta and affecting about 5 to 7% of the pregnancies, can lead to several fetal or maternal complications. It can occur since the 20th week of pregnancy. PE is the association of hypertension, proteinuria and edema, contributing to placental impairment and fetal distress. Related effects can also impair long-term neonatal outcome, potentially influencing his metabolism until adulthood. Due to these reason, the early identification of women at risk of developing PE would be desirable and the biomarkers currently employed in risk prediction are weak outcome predictors. Among the metabolomics studies in PE, we report few interesting and promising results. Sander and colleagues detected significant changes in serum metabolome (third trimester) of PE pregnant women ( n = 32), versus healthy controls ( n = 5), and most varying metabolites were hydroxyhexacosanoic acid, diacylglycerols, glycerophosphoinositols, nicotinamide adenine dinucleotide metabolites, bile acids and products of amino acid metabolism . Moreover, the group of Liu, analyzed the eicosanoid content in serum of PE ( n = 10) and healthy pregnant women ( n = 10) through LC-MS; as result, levels of arachidonic acid metabolites and some of the lipoxygenase metabolites of eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA), were increased, while cytochrome P450 metabolites of EPA and DHA were decreased in women with PE. The values of lleukotriene B4, 14,15-dihydroxy-eicosatetraenoate, 16-hydroxydocosahexaenoic acid and 8,9-epoxy eicosatetraenoic acid resulted significant markers of PE occurrence and progression . According to another study, metabolomics could help in discriminating PE pregnancies associated to pre-term and full-term delivery, evaluating seriate serum maternal samples (12, 20, 28 and 36 weeks of GA) in women with PE delivering full-term ( n = 165) and pre-term ( n = 29). Among the obtained results, it emerged that 4-hydroxyglutamate could represent a novel first-trimester predictor of pre-term disease . According to another study, metabolomics placental profiles could also be useful in identifying PE with placental dysfunction and even associated IUGR . Recent metabolomics results in IUGR , even in relation to gut dysbiosis and evaluating neonatal urines , and in GDM are currently available in literature. In the future, the study of great obstetrical syndromes, old data, could be better performed through the use of new eyes, especially metabolomics. Research should be focused on the interactions between genomics, dysbiosis, environmental factors, including diet, hypertensive and inflammatory factors. In conclusion, a metabolomic approach in different areas of maternal and perinatal medicine could really help in a best characterization of such conditions and in the identification of novel biomarkers, even if the current findings still require further validation on larger cohorts . The most recent metabolomics studies on amniotic fluids were recently reviewed, describing metabolic interactions between the mother and the fetus in several pathophysiological conditions, highlighting how such technique can describe feto-placental metabolism .
The potential role of metabolomics in perinatal asphyxia is an intriguing topic, well describing the concepts of great inter-individual variability and the needs of personalized approaches. Perinatal asphyxia is one of the most frequent causes of neonatal death or impaired outcome, potentially leading to severe disability, cerebral palsy and poor neurodevelopment . Such topic was deeply investigated across several studies, speculating the potential occurrence of specific metabolic perturbations in the urine of asphyxiated newborns. In the first of these studies, urinary samples were collected and evaluated from n = 3 full-term male asphyxiated neonates. Those patients, although characterized by the same clinical picture of early seizures and laboratory features (Ph 6.8 and EB – 22 mEq/l), underwent three really different outcomes. One out of three died within 48 h of life, one underwent acute renal and hepatic failure and cerebral palsy at discharge, the last completely recovered and was discharged in good clinical conditions. Thus, the same treatment (hypothermia) could not result the optimal treatment approach in all the asphyxiated neonates, due to the great differences characterizing each individual. The urinary metabolome of these neonates were analyzed at birth and at the end of hypotermia, with 1 H-NMR and resulted significantly different, reflecting the three different neonatal outcomes. In detail, mediators such as glycine, valine, maleic acid, and sorbitol varied in relation to asphyxia, while glucose, aspartic acid, asparagine, ornithine, gluconic acid and L-lysine were mostly influenced by kidney damage . By another study, evaluating n = 14 neonates ( n = 6 cases, n = 8 controls) through the same platform, we demonstrated that asphyxia-related metabolic urinary variations were also evident 48 h after birth. Hypoxaemia and acidosis mostly determined variations in metabolites involved in energy demand, kidney damage and oxidative stress (lactate, glucose, TMAO, threonine, 3-hydroxysovalerate) . Finally, progressive modifications of urinary metabolome from n = 12 asphyxiated newborns were pointed out, compatible with the progression of clinical condition. Urinary samples were evaluated at birth, during and at the end of hypothermia, at a week and a month after birth through GC-MS and 1 H-NMR . In detail, taurine, hypotaurine, U1710, lactic acid, lysine, mannitol and ubiquinone showed the greatest variations . The urinary metabolome of survivors was clearly different from that of neonates that died in the first 8 days of life, and these differences were already present at birth. These modifies could indicate irreversible asphyxia-related perturbations, indicated by metabolic fingerprints, such as lactic acid, taurine and other metabolites . Moreover, energy deficiency, variations in the cycle of tricarboxylic acids (TCA) and the increase in lactic acid represent negative predictors, while their decrease during post-natal time may represent indicators of aerobic metabolism and homeostasis restoring in surviving infants with the best outcomes . Finally, lactic acid, myo-inositol, betaine increased while citrate, α-ketoglutarate, succinate, acetone, dimethylamine, glutamine, pyruvate, arginine and acetate decreased in asphyxiated newborns’ urinary profiles at 1 month of life . These data highlight that determining a real threshold for survivals characterizing each metabolite could be highly useful. In a pig model of perinatal asphyxia, urinary metabolomics was also applied to investigate the effect of different oxygen concentration (18, 21, 40, and 100%) administered during resuscitation. In such study, 21% of oxygen resulted associated to the best outcome and metabolic effect, highlighting that metabolomics could also help in monitoring the effects of therapeutic approaches and oxygen supplementation . These findings point out that each patient is characterized by a high inter-individual variability; maybe some patients die or develop side effects due to an overtreatment while others would benefit from a more aggressive treatment approach. In fact, the current medicine is based on detailed protocols, calibrated on the mean of patients. However, research is revealing that the mean of patients does not exist and the development of precision medicine could be beneficial for each single individual. If these results were confirmed, metabolomics could help in identifying early markers of perinatal asphyxia, describing the evolution of such condition over time and resulting highly related to neonatal prognosis.
Neonatal nutrition is a main relevance topic, being one of the most important factor influencing the early newborns development and affecting short- and long-term outcome, due to the power of breast milk (BM)-associated perinatal programming. In fact, in the first weeks of life, BM is able to change the fate of newborns’ metabolism . BM contains water (88%) nutrients (lipids, carbohydrates, proteins, vitamins, minerals), bioactive components as growth-factors (GFs), hormones, cytokines, chemokines, antimicrobial compounds like immunoglobulins (Ig), a specific microbiome and BM-related cells including epithelial, immune cells and multipotents stem cells (SCs) . BM represents the ideal biofluid for neonatal nutrition, especially if premature, able to modify its composition according neonatal needs, especially in terms of GA or lactation stage . Among BM components, BM oligosaccharides (HMO) can shape neonatal gut microbiome, influencing immune system development, protecting against infections and reducing necrotizing enterocolitis (NEC) rate . HMOs composition in BM highly depends on maternal genetic factors, since mothers can be divided in Secretors ( Se+ ) and non Sectetors ( Se- ) according to the expression of the enzyme α-1-2-fucosyltransferase (FUT2), codified by Se gene. Metabolomics studies evidenced a clear separation between BM of Se + and Se- mothers, with a higher content of fucosylated oligosaccharides, 2α-fucosyl-lactose, lacto-difucotetraose, lacto-N-fucopentaose and lacto-N-difucoesaose in the first group. Thus, neonates of Se- mothers could benefit from specific HMOs supplementation, to avoid NEC and other infections . Even BM microbes, also called maternal “lactobiome” seems to influence neonatal outcome . Thus, understanding BM composition and its effects is a central issue of modern research. A personalized nutrition based on the features of each newborns would be desirable. In this perspective, metabolomics represents an ideal tool to analyze BM and the composition of different types of commercial available formula milks (FM), allowing their improvement to resemble as possible BM composition. Moreover, metabolomics results also promising in the detection of drugs and contaminants in BM, helping in the determinations of its safety in specific conditions or maternal exposition of environmental toxicants . The first metabolomic study evaluating BM composition was performed in 2012 by our research group. BM from n = 20 mothers delivering neonates of GA 26–36 weeks, and from n = 3 full-term delivering mothers was collected from 1 to 13 weeks post-partum. In the same study, samples of FM were also analyzed. As result, a clear separation occurred in BM vs FM (samples analyzed through 1 H-NMR and GC-MS). In fact, a higher lactose concentration was found in BM, while maltose resulted higher in FM. Some differences also characterized FAs profile, such as oleic and linoleic acids that were higher in FM. Although on a small number of samples, a correlation with GA was highlighted, especially regarding carbohydrates; i.e., lactose increased during milk maturation . Interesting results emerged from the analysis of metabolic effects of different nutrition regimens on neonates. The urinary metabolome in the first week of life in three groups of newborns (GC-MS), divided in appropriate for gestational age (AGA), small for gestational age (SGA) and large for gestational age (LGA) was compared. This study highlighted the power of early nutrition on neonatal metabolic pathways. In fact, despite a clear separation showed by urinary metabolome of AGA group than LGA and SGA at birth, at 1 week of life, urinary samples were mostly influenced by nutrition, with a clear separation between the samples of breastfed neonates and those exclusively fed with FM . It was also demonstrated that such significant nutrition-related metabolic differences can persist up to 4 months of life. The metabolites showing the greatest variations (measured through 1H-NMR) were those related to energy metabolism, antioxidant action, neuro-modulation and brain development, surfactant synthesis; moreover, variations were detected in HMOs content and intestinal microbiome-produced metabolites. These findings highlight the effects of early nutrition on neonatal development . Recently, a unique study on BM collected from mothers delivering preterm multiples ( n = 19 couples and n = 5 triplets) from birth and up to 20th w of postnatal life pointed out a greater protein content in BM of preterm multiples than singletons matched for GA, despite a lower content of lactose. The higher protein content in BM for preterm multiples could face the nutritional and development needs of such vulnerable category . In a preliminary investigation comparing metabolomics profiles of n = 15 different kinds of FM including n = 6 organic –bio formulas vs BM, a significantly higher methionine content in organic bio-formulas (about 3 folds higher) than conventional FM was pointed out . This result is highly interesting taking into account that methionine is an epigenetic mediator, involved in methylation. In pathological conditions, BM could interfere with neonatal development and can impair the function of several organs . In this regard, we recently focused on Great Obstetrical Syndromes-GOS (PE, GDM and IUGR), pointing out that BM metabolome could be altered in such conditions, through unknown mechanisms including maternal deficiency of specific metabolites and potentially involving inflammatory triggers. Such altered BM composition will affect neonatal development, contributing to the adverse long-term outcomes in children born by mothers affected by GOS and therefore exposed to an altered intrauterine environment . In a rat model, it has been also demonstrated that maternal obesity and non-alcoholic fatty liver disease (NAFLD) influence offspring metabolism, predisposing to dysmetabolism, insulin resistance, obesity and NAFLD itself, potentially via BM (in fact, a higher content of leptin was detected in these mothers instead of healthy controls) . Finally, BM contains several populations of cells, including epithelial cells, immune cells and stem cells (SCs) named human breast-milk derived SCs (BMSCs) . These cells take part in mammary gland proliferation during pregnancy and lactation, end express different and specific markers according to lactation stage and GA at birth . The greatest potential of BMDSCs on neonatal outcome depends on the ability, after neonatal ingestion, to pass through neonatal gut into circulation, where they could survive and be transferred into brain and other neonatal organs, influencing their development . Since BMDSCs resulted able to differentiate into several cellular lines, including nervous cells and neural SCs , the transfer of BMDSCs through breastfeeding could improve the maturation of neonatal brain and other organs, especially in premature neonates .
Sepsis, potentially caused by viruses, bacteria or fungi, is a frequent cause of neonatal morbidity and mortality, especially if affecting the highly susceptible premature newborns. “Early onset” sepsis (EOS) occurs within 72 h of life, while “late onset” sepsis (LOS) between 72 h and 6 days of life. Although sepsis represents a life-threatening condition, current biomarkers lack in diagnostic accuracy. The early and sensible diagnosis of such condition, based on reliable and accurate mediators, could improve its management and prognosis. Currently, sepsis diagnosis, allowed by blood microbiological culture, is often delayed . Thus, metabolomics approach could provide new chances in the diagnosis of sepsis, in clarifying its pathogenic mechanisms and prognosis, through the definition of novel sensible biomarkers, as reviewed by Lee and colleagues . Metabolomics could reveal sepsis-related metabolic pathways, such as hypoxia, oxidative stress, and increased energy needs (influencing glucose levels and oxidative metabolism of fatty acids) . Moreover, preventive and therapeutic strategies could benefit from the whole comprehension of the interactions among the host and its microbiome, playing a pivotal role in sepsis progression . Some studies available in literature, briefly reviewed above, show promising result, even if they require further confirmation on larger samples. In the first of them, urinary samples collected at a single time-point from septic neonates (including both EOS and LOS) was compared to healthy controls ( 1 H-NMR and GC-MS), highlighting effects on energy metabolites (including the increase in glucose, maltose, lactate, acetate, ketone bodies intermediates, and effects on antioxidants) . Moreover, in another study, a different urinary profile in a preterm neonate affected by fungal sepsis instead of healthy controls (GC-MS) was pointed out. Among the observed results, some proteolysis-related amino acids suggesting a hypermetabolic and hypercatabolic state increased, and the metabolite D-serine resulted a good predictor of antifungal treatment response, reducing itself during therapy . In a study of Serafidis and co-workers, urinary metabolic profile evaluated through 1 H-NMR and LC-MS allowed a clear discrimination between septic and non-septic neonates, analyzing samples collected at the time of diagnosis, and after 3 and 10 days. Metabolite variations disappeared at the end of the symptoms, giving promising information for prognosis and therapy . The group of Stewart and co-workers demonstrated that metabolomics application could also give interesting results on fecal samples; in fact, they showed the role of gut microbiome is involved in the pathogenesis of LOS. By comparing a group of healthy neonates to LOS affected infants, Bifidobacteria (with protective beneficial effects) resulted higher in the first group, in conjunction with prebiotic oligosaccharides, raffinose, sucrose, and acetic acid. Interestingly, the same bacterial species isolated by blood culture were dominant in gut microbial community, as consequence of a bacterial translocation from the gut into circulation and arguing for the first time a dependence of neonatal sepsis on gut dysbiosis . Finally, metabolomics was also applied in pediatric sepsis, in a cohort of n = 60 septic pediatric patients (including n = 7 newborns); their serum metabolome was compared with n = 40 healthy pediatric controls and showed an increased in lactate, glucose, creatinine, 2-oxoisocaproate, 2-hydroxysovalerate and 2-hydroxybutyrate and lower levels of threonine, acetate, 2-aminobutyrate, and adipate . In conclusion, metabolomics could provide precocious and accurate sepsis marker in the perspective of an early diagnosis and a tailored management, potentially reflecting disease progression, therapy related efficacy, and toxicity .
The expression autism spectrum disorders (ASD) includes a group of neurodevelopmental disorders characterized by delayed or impaired language development and difficulties in social interactions, and repetitive and stereotyped behaviors. The pathogenesis of such condition, showing a high heterogeneity, includes the interaction among genetic factors, environmental risk factors, socioeconomic status, maternal and neonatal infections, prenatal nutrients, immune deregulation, maternal exposure to potentially toxic drugs, formula feeding, and epigenetic components (including DNA methylation), even if the exact mechanisms are not fully understood, up to now. Despite ASD rate rapid increase in recent years, their diagnosis is still largely based on clinic signs and sensible biomarkers are not available. Metabolomics recently emerged as promising tool for a better characterization of such diseases, allowing the individuation of sensible biomarkers, their monitoring and maybe the introduction of innovative treatments. In particular, the metabolites mostly associated to ASD seems those involved in amino acid metabolism, cholesterol metabolism, folate abnormalities, antioxidant status, nicotinic acid metabolism, and mitochondrial metabolism. A great role could also be played by some metabolites derived from the gut microbiota, potentially shaping ADS children behavior, metabolic patterns and immune response , tryptophan, vitamin B6, purine metabolic pathways, phenylalanine and tyrosine biosynthesis, intermediary compounds of the TCA cycle . Metabolic features of autistic children were evaluated in several studies; some of them are reviewed above. By two recent studies, the involvement of oxidative mechanisms and intestinal microbiome in ASD predisposition was detected, via the interaction with the gut-brain axis and to the lack of the intestinal mucosal barrier. Through the analysis of urinary samples from n = 21 ASD versus their n = 21 healthy siblings (aged between 4 and 17 years), different levels of oxidative metabolites, carbohydrate metabolism intermediates, bacterial-derived metabolites suggesting an increase in Clostridia spp . in the gut were detected. Moreover, aromatic amino acids precursors of neurotransmitters and key hormones for the nervous system such as catecholamine and serotonin also showed variations. These results evidence that diet is a relevant epigenetic factor, influencing gut microbiome, in ASD pathogenesis . In ASD children, the increase in Clostridium, Alistipes, Akkermansia, Caloramator, Sarcina spp ., and the reduction in Prevotella spp., E. siraeum, and Bifidobacterium spp. occurred, with consequent alterations in urinary levels of hippuric acid, p-hydroxyphenylacetic acid and 3-(3-hydroxyphenyl)-3-hydroxypropanoic acid, and propionic acid . Plasma metabolome of ASD children was also evaluated by Orozco and colleagues, who described interesting metabolic alterations related to the impairment in neurodevelopment, also taking into account the potential overlap between ASD and other causes of developmental delay, including Down Syndrome and idiopathic-developmental delay . Moreover, since ASD prevalence is higher in males, metabolomics approach was also applied to investigate the urinary pathways in males and female patients, analyzing the possible molecular causes of such gender difference and trying to find sensible biomarkers for the diagnosis of ASD potentially related to the subject’s gender. In detail, the authors evidenced a significant increase in the levels of adenine, 2-methylguanosine, creatinine, 7alpha-hydroxytestololactone and a decrease in creatine in females; thus, they identified creatinine:creatine ratio as potential marker of ASD in females . Metabolomics perturbations associated to ASD cannot be only detected in blood, urine or saliva, but even in cerebellum or cortex samples of affected patients. In the study of Kurochin et al., 1366 metabolites were compared in the prefrontal cortex grey matter of ASD patients and healthy controls, revealing different profiles and metabolic pathways and opening the way to different analysis strategies . Finally, the potential pathogenic effects of some drugs (including thalidomide and, as recently highlighted, acetaminophen) on ASD development should be further investigated . A clinical promising application of metabolomics in neuropsychiatric disorders was recently reviewed. The current evidence on Pediatric Acute-onset Neuropsychiatric Disorder (PANS), a clinical condition characterized by sudden obsessive-compulsive symptoms and a close dependence on infective triggers and potential post-infectious immune-mediated mechanisms, was analyzed. In this regard, a metabolomics approach was applied to investigate the case of a 10 year-old girl with PANS. Our evaluation of her urinary metabolome was performed before and after the treatment with clarithromycin, macrolides playing antimicrobial activity and potentially acting as immunomodulatory agent. During such pharmacological treatment, a great improvement of her clinical manifestations with the reduction of symptoms was observed, in addition to a clear modification in urinary metabolomics pathways, especially regarding metabolites related to protein biosynthesis, energy metabolisms, aminoacids involved in brain functions and microbial products also related gut colonization . Metabolomics seems also promising in providing metabolic information improving early recognition and potentially modifying the treatment and prognosis of inborn errors of metabolism (IEMs), as recently reviewed. Through a new approach based on the integration of metabolomics and genomic data, a set of metabolites selectively influenced by a specific gene inactivation in urine, blood or other biofluids could improve our knowledge of disease-related metabolites in various genetic condition . The study and application of metabolomics to neuropsychiatric disorders seems to be a promising tool for the immediate future, potentially identifying specific biomarkers, involved since the early phases of fetal life and influencing neurodevelopment from perinatal programming to adulthood .
Metabolomics has been extensively studied in Medicine, as evidenced by the presence of more than 28.600 related papers on PubMed, until now. In the last years, many articles and reviews have been published in the fields of obstetrics, perinatology, neonatology and pediatrics. Metabolomics resulted a highly promising tool in the early diagnosis of several fetal, perinatal, pediatric and adulthood conditions, through the detection of specific and sensible biomarkers. Moreover, metabolomics could help in monitoring the disease progression, in optimizing therapy and in the evaluation of related side effects, in the perspective of a tailored management. In the next years, we will move towards a personalized holistic approach in pediatrics. Even if none of the mentioned techniques was yet validated for current clinical application, it is reasonable to predict that available results could be confirmed on larger cohorts of patients and through a more close standardization of protocols and studies. In the future, we hope that metabolomics sensible sticks could be included in clinical practice, potentially accurate and cheap, to investigate rapidly neonatal biofluids and to introduce an innovative and personalized therapeutic management at the patient’s bed. However, it is possible to predict that metabolomics will change pediatrics in the immediate future, representing both an evolution and a revolution. We could say that metabolomics can bring from small metabolites to big ideas, and we hope it could provide more chances to sick neonates and children, highlighting what is good health and revealing how to maintain and defend it, improving well-being and preventing diseases. A dream? May be I am a dreamer, but I am not the only one.
|
HIV HGM biobank as a research platform for paediatric infectious diseases and COVID-19 pandemic | 9f48b37e-1697-4645-a789-76119cb610c3 | 9130977 | Pediatrics[mh] | In the year 2020, a new infectious respiratory disease emerged in Wuhan, Hubei province, China . An initial cluster of infections was linked to Huanan seafood market, potentially due to a human contact with animal species. Subsequently, human-to-human transmission occurred and the disease now termed, coronavirus disease 19 (COVID-19) rapidly spread within China . The first cases of COVID-19 occurred in December 2019 and Spain was one of the most affected countries during the first wave of COVID-19 (March to June) . Then it was devastated by a second wave of COVID-19 infections and all subsequent waves that have devastated the entire planet and caused the greatest medical problem experienced by humanity in recent decades. In the severe pandemic context biobanks have become even more essential than before . They are professional repositories of biological samples. In Spain, in 2004, the Hospital General Universitario Gregorio Marañón founded a national biobank, named HIV HGM BioBank, with the aim of contributing to advance understanding of different pathologies through the transfer, management, register, processing, cryopreservation and cession of biological material from patients, always for research purposes and under conditions that guarantee its usefulness in current studies and future research that may appear as knowledge evolves. This biobank is located in Hospital General Universitario Gregorio Marañón, Madrid, one of top 5 hospitals in Spain, and contains a large number of samples and represents a centralized approach to collecting, processing, and storing samples from all over the country. Along its 17 years of trajectory, this platform has focused its excellence on infectious diseases and, more specifically, on paediatric donors. Due to its experience, HIV HGM BioBank is now an international model of management in studies and clinical trials based on cohorts of infant samples. This HIV HGM BioBank holds whole blood and serum, plasma, urine, cerebrospinal fluid, feces and meconium, breastmilk, nasal aspirate, and different types of immune cells and DNA, all stored for research use . Our HIV HGM BioBank is engaged in the physical placement of samples and the full amount of work associated with these samples, and responsible for ethically and accurately data management, related to consent, privacy, and control. In this sense, in Spain, Royal Decree 1716/2011 of November 18, states the basic requirements on authorization and performance of biobanks for biomedical research and the management of human biological samples and their associated data in order to achieve excellence in quality and data integrity. According to this rule, biobanks shall: (i) guarantee the informed consent process; (ii) safeguard the right of the withdrawal of donors’ decisions; (iii) guarantee the right to privacy and respect for the will of the subjects participating in the studies; and (iv) assure the equal distribution of samples. . This biobank represents a novel approach, not only to HIV-1 infection, but also to a COVID-19 research, and specializes in paediatric population . This is why the HIV HGM BioBank is of general interest to basic and clinical research teams working on HIV-1 and COVID-19, and also to those groups who try to establish large networks focused on research on specific clinical problems in epidemiology, biology, routes of transmission and potential therapies . It is important to note that biobanks are biomedical, scientific, infrastructural development, and they represent a political, legal, ethical enterprise, being integrated by the regulation, medicine, law and society . The purpose of this Spanish BioBank is to set up the unified findings, and discussions on the design and the selection of equipment, the management development methods and staff training, on the standardization of methods for the collection, processing, shipping and storage of biomaterial of different origins as well as on methods and validations for quality control, creation and use of databases of information accompanying biospecimens. Biobanks have been crucial in the run towards a COVID-19 vaccine and/or treatment and a source of knowledge to understand the possible mechanisms that contribute to the appearance and spread of the most critical pandemic of the last decades . The personnel of Spanish HIV HGM BioBank are responsible for the reception, preservation and storage of multiple samples areas, with a high expertise in infectious diseases and excellence in working with infant samples . Taking into consideration the present pandemic context, the number of staff working in the Spanish HIV HGM BioBank has been increased. This increase is also a reflection of the critical situation that the COVID-19 pandemic has caused for those countries in which infectious diseases already represented a very important challenge in their containment and treatment, and serves as a potent reminder of the necessity to reinforce medical and public health capacities. Also the threat of future outbreaks should not be underestimated .
At this moment eighty-five hospitals spread across Spain use the HIV HGM BioBank, which have collections of human biomaterials of individual interest. All the HIV-1 samples collected by our HIV HGM BioBank represent the Spanish HIV-1 infection the same as with COVID-19 infection. Our objectives are: (i) the new structure and function of the HIV_COVID-19 HGM BioBank that releases very efficiently samples to different research project, not only in Spain, but also in other countries, (ii) the importance and contribution of a biobank specialized in paediatric research, even more in a pandemic context when a specific approach for this population is needed. Up to now, for more than 15 years of the existence, the HIV HGM BioBank (web site www.hivhgmbiobank.com ) has become not only a Spanish HIV HGM BioBank, but also, since February 2020, a Spanish Coronavirus HIV HGM BioBank. The function of this biobank is crucial for the development of new diagnostic and possible therapies for all persons with infectious diseases, such as HIV-1 and COVID-19 infections and HIV-1/COVID-19 coinfections. Biobank workflow is maintained in a strictly organized manner. This HIV-1 and COVID-19 BioBank applies standard operating procedures (SOPs) for samples required the preservation of viability, structural integrity, functionality and stability. The SOPs ensure correct implementation of essential biobanking components such as samples, associated databases, donors, ethical approvals and informed consents, acquisition, transport, preparation, analysis process faultlessness, proper storage, conditions and terms samples sharing and ensuring the maintenance of the material stored. One of the main goals of this HIV_COVID-19 HGM BioBank is to recognize the researcher’s needs. Therefore, a researcher must complete a sample release application with the aim to receive HIV-1, COVID-19, HIV/COVID-19 or infectious samples of donated materials from patients. Once the received application is approved by the members of the Scientific Committee of the Spanish HIV HGM BioBank, the researcher signs a Material Transfer Agreement (MTA) with the director of the biobank and the coordinator of the cohort. In return the researcher is obliged to send a scientific report with her/his work results every year and to index a reference in materials & methods and acknowledgements sections. The samples such as blood, serum, plasma, peripheral blood mononuclear cells (PBMC), pellet cells, DNA, RNA, umbilical cord blood, feces, meconium and breast milk are processed and stored in the Spanish HIV_COVID-19 HGM BioBank. It has been recognized that one of the main objectives is to process, store and provide different samples from HIV-1 and/or COVID-19, not only in adults patients, but also in newborns, neonates, infants, children, adolescents and elders to research projects. It is important to know that COVID-19 guidelines mandate the use of Biosafety Level 2 (BSL-2) rules for laboratories, including the use of protective equipment freezers, the use of Class II Biological Safety Cabinets, and proper disinfection routines. In other words, conditions under which the high Biosafety Level 3 (BSL-3) practices should be followed when working with culture specimens, with different cell lines. Over time, the operation of the HIV_COVID-19 HGM BioBank will become crucial for any research. As expected, the HIV_COVID-19 HGM BioBank has been set up according to a system of quality management based on the rules written in UNE EN-ISO 9001:2015 that covers the full spectrum of the HIV_COVID-19 HGM BioBank´s operations. This recently updated HIV_COVID-19 HGM BioBank is run by a scientific director and a data manager who are assisted by a Biomedical Research Scientific Committee. An independent Biomedical Research Ethics Committee (CEIm, Hospital General Universitario Gregorio Marañón reviews the agreements made with the different cohorts regarding the patients’ Informed Consent. ( https://www.iisgm.com/organizacion/comisiones/comite-de-etica-de-la-investigacion-con-medicamentos-ceim/ ). The current COVID-19 pandemic, patient sample collection, processing, and analyses are at the forefront of this emergency. COVID-19 presents some unique issues in the biobanking world. The overwhelming scope of this pandemic, with around 2 million cases globally as of mid-April, has assigned an outstanding role to biobanks as they work with samples essential for developing diagnostics and vaccines. The HIV_COVID-19 HGM BioBank consists of well-trained, professional personnel, adequate facilities, equipment, protective measures and biosafety to incorporate novel functions and deal with samples of a new virus associated with the infectious disease such as the COVID-19. This amplified function of the Spanish HIV HGM BioBank was launched at the beginning of the pandemic to obtain and collect samples from newborns, neonates, infants, children, adolescents, adults and elders infected or who had undergone the infection by COVID-19. The HIV_COVID-19 HGM BioBank has been created to offer different samples donated by patients with the objective to discover new findings in newborns, children, adolescents and adults infected by HIV-1, COVID-19, or HIV-1/COVID-19 . It is worth noting that biobanks have been identified as a biomedical scientific infrastructural development integrated into the preexisting form of regulation, medicine, law, and society. Strict compliance of ethical norms is always guaranteed. The HIV_COVID-19 HGM BioBank offers samples, technical and scientific advances, establishing interesting support to global research on the HIV-1 and COVID-19 infection diseases. In summary, the main function of this HIV_COVID-19 HGM BioBank is to collect and guard human samples strictly connected with their associated data, guaranteeing the highest quality, confidentiality of donors and complying with current ethical and legal regulations. In fact, our Spanish HIV HGM BioBank stores thousands of samples of healthy donors, HIV-1 individuals, COVID-19 individuals, and HIV-1_COVID-19 coinfected individuals.
Since its creation, HIV HGM BioBank has processed 472,618 aliquots from 54,106 samples of 18,979 donors. The 94.04% of this material comes from patients of infectious diseases. HIV HGM BioBank holds 21,806 aliquots from 5784 samples of 2065 paediatrics donors, of which the 68.47% are of infectious pathologies. Since the moment the pandemic started, our HIV HGM BioBank has received and processed 2901 aliquots from 1407 samples of 593 donors infected with COVID-19 virus. Thus, 27.79% of those material (806 aliquots) belongs to infant patients. This effort positions this platform as a reference in the research of infectious diseases and/or pediatric field. The relevance of counting with HIV HGM BioBank in the scientific community is described along several quality indicators: (a) samples, aliquots & cohorts at researcher’s disposal; (b) collaborations; (c) donated material; and (d) publications. Those elements are exposed as follows:
Biomaterial at scientific disposal The historical growth ratio for the number of samples preserved in HIV HGM BioBank is + 25% per year, + 10% in the last five years. For the number of aliquots processed each year the ratios are more variable, but HIV HGM BioBank has a + 9% global since 2004, with a record of + 53% and a − 17% minimum. The growth in number of collections managed by HIV HGM BioBank is + 14% per year historically, and + 8% per year for 2016–2020 (Fig. ). Collaborations and services In clinical development the global growth ratio is + 23%, + 17% in the last five years. For research projects with public or private foundation, the growth ratio is + 12% per year (Fig. ). Material donated Aliquot donation has developed a global growth ratio of + 28%, + 15% in the last 5 years. The forecasts of samples transfer is one of the main obstacles in biobanking management because it does not suit to the studies of previous demand, so its evolution is more erratic than other indicators. Despite this trait, HIV HGM BioBank has experienced a positive growth since 2004 in sample cession (Fig. ). Results promoted by HIV HGM BioBank Papers published thanks to the donation of HIV HGM samples or services, with a global growth ratio of + 27%, + 9% in 2016–2020 period (Fig. ). Due to this great expansion in the last five periods, HIV HGM BioBank expects a great development in short time cycles. Also, the specific legislation in biobank activities established in 2012 promotes the use of these facilities in research and clinical trials, improving and driving our requests, services and collaborations. Nowadays, only in the first trimester of 2021, three new clinical trials and 2 paediatric cohorts have been requested. The ratios of expected development are + 8% for projects, + 11% collections and + 13% of participation in clinical trials (Fig. ).
The historical growth ratio for the number of samples preserved in HIV HGM BioBank is + 25% per year, + 10% in the last five years. For the number of aliquots processed each year the ratios are more variable, but HIV HGM BioBank has a + 9% global since 2004, with a record of + 53% and a − 17% minimum. The growth in number of collections managed by HIV HGM BioBank is + 14% per year historically, and + 8% per year for 2016–2020 (Fig. ).
In clinical development the global growth ratio is + 23%, + 17% in the last five years. For research projects with public or private foundation, the growth ratio is + 12% per year (Fig. ).
Aliquot donation has developed a global growth ratio of + 28%, + 15% in the last 5 years. The forecasts of samples transfer is one of the main obstacles in biobanking management because it does not suit to the studies of previous demand, so its evolution is more erratic than other indicators. Despite this trait, HIV HGM BioBank has experienced a positive growth since 2004 in sample cession (Fig. ).
Papers published thanks to the donation of HIV HGM samples or services, with a global growth ratio of + 27%, + 9% in 2016–2020 period (Fig. ). Due to this great expansion in the last five periods, HIV HGM BioBank expects a great development in short time cycles. Also, the specific legislation in biobank activities established in 2012 promotes the use of these facilities in research and clinical trials, improving and driving our requests, services and collaborations. Nowadays, only in the first trimester of 2021, three new clinical trials and 2 paediatric cohorts have been requested. The ratios of expected development are + 8% for projects, + 11% collections and + 13% of participation in clinical trials (Fig. ).
Biomedical research and the development of personalized medicine go hand in hand with the development of management platforms for large sample collections and big data management, such as biobanks. If these structures were essential before the pandemic, they are even more now that we have discovered that, unfortunately, situations of serious global health crises are unpredictable and unavoidable. It is essential to learn from the experience that COVID-19 has offered us in terms of adapting our structures and knowledge to redirect resources to respond to health problems that society demands to solve urgently. In this sense, it is even more important to have specialized entities, not only in respond to major global challenges, but also in managing the same responses in the pediatric population, whose requirements are specific and sometimes very different from those of the adult population. In this sense, the HIV HGM Biobank has developed a strategy for rapidly adapting its infrastructures and know-how to the implementation of pediatric and adult collections of COVID19, as well as cohorts of patients co-infected by both pathologies. The possibility of enable to the scientific community pediatric and neonatal samples of excellent quality and a large amount of associated data is especially relevant, since the effects of SARS-COV2 infection in patients with an immature immune system must be studied in depth in the future since its repercussions are still unknown. In addition, some exceptions were established in the legal requirements for obtaining informed consent, authorized by research ethics committees. A waiver was granted from the need to obtain a signed informed consent during the period of confinement that during the first wave kept the entire population at home. During the first and hardest part of the pandemic, oral consent was valid for the deposit of biological samples in our biobank. This exemption ended in July 2020, and then the donors had to be re-contacted and a written consent formalized. This allowed a more agile and productive management of the deposit of material in the HIV HGM Biobank for these new collections of COVID19 samples. In this way, more material and data were accessed in less time, especially in a moment when achieving critical masses of samples is key to providing tools to the scientific community and being able to provide answers to society. This strategy will allow innumerable studies in the future that will help clarify the mechanisms of action and the possible routes of intervention to eradicate SARS-COV2 in infants and adults.
|
Strategies to Improve the Clinical Outcomes for Direct-to-Consumer Pharmacogenomic Tests | abff6880-3e00-4d76-a5fa-b5faf0ed1000 | 7999840 | Pharmacology[mh] | 1.1. Direct-to-Consumer Genetic Tests Since the completion of the Human Genome Project (HGP), DNA sequencing tests for health-related purposes have become common in medical laboratories . Additionally, the advent of high-throughput sequencing methods has made DNA analysis tests faster and easier. During early 2000, some genetic and genomic companies started offering genetic testing directly to individuals without the need for the prescription of physicians or other healthcare providers. Direct-to-consumer genetic tests (DTC-GT) are defined as genetic tests marketed directly to customers through print or visual media or the internet, or that can be bought online or in brick-and-mortar stores with no/least involvement of healthcare professionals in this process . DTC-GT allows customers to access their genomic interpreted data whenever and wherever they want. However, over time, after some critical evaluations, companies selling DTC-GT started engaging geneticists and medical professionals’ in the form of pre-and post-test consultation for consumers (advertisements, articles, brochures, personal contact, etc.) to provide advice on further actions to be taken after the genetic testing results. Most companies declare in their policies and consent forms for consumers that such tests should not be considered as diagnostic tests, but only as informative tests . Although the DTC-GT tests will be more useful in preventive areas than diagnosis issues, the test outcomes can demonstrate significant help for the future clinical assessments of individuals through providing the healthcare-related result before they visit a specialist. A DTC-GT report can bring attention to a specific condition in people, which may need more consideration and clinical confirmation or intervention by the clinicians . Today, DTC-GT services are available as kits for obtaining saliva or buccal swab samples which are non-invasive. The samples can then be sent to the company providing the service where DNA analysis is performed, usually in CAP and/or CLIA accredited laboratories. The companies mostly using array-based or sequencing platforms (targeted gene sequencing panels or broad range genomic tests such as whole-genome sequencing (WGS) or whole-exome sequencing (WES)) approaches for analyzing specific mutations or providing a comprehensive picture for genomic variants in a given sample. Each company applies its microarray or sequencing technologies (i.e., Illumina HumanOmniExpress-24 single nucleotide polymorphism (SNP) chip for 23andMe and WES for Genos) with post-processing involving imputation, and the interpreted genomic data then are returned to customers via the internet or mail after a couple of weeks or months . Some companies such as 23andMe (Sunnyvale, CA, USA), Color Genomics (Burlingame, CA, USA), etc. also provide raw genetic data to their customers, so the customers can use this raw genetic data for further processing and analysis through free online resources and tools such as Promethease, Live Wello, Genetic Genie, etc. with the help of a physician, clinical geneticist, genetic counselor, pharmacist or other trained genetic professionals. Currently, DTC-GT companies offer their services in two main categories which include medical and non-medical genetic tests. The medical genetic tests can be classified as carrier tests (e.g., hemoglobinopathies), disease susceptibility detection tests (e.g., Parkinson’s disease), pharmacogenomic tests (for the specific number of drug-related genes), life-style related tests (genetic analysis for complex diseases), and prenatal tests (PND & PGD). They can also be divided into tests for monogenic disorders, polygenic defects, multifactorial diseases, genome-wide testing (thousands of SNPs), and broad range tests (WGS & WES). Medical genetic testing services are the most common tests used by people and are the main reasons for the increasing growth of DTC-GT companies. Non-medical testing services, on the other hand, consist of testing for some traits and features in individuals, which are not necessarily related to disease or health, and are usually for “infotainment”. Examples of these include ancestry information, ear lobe attachment, and the flush reaction after drinking alcohol, etc. However, there is an argument that ancestry data should be included under medical information, because this information helps to determine whether a specific ethnic group has a predisposition to a particular genetic condition (e.g., Tay-Sachs disease in individuals of Ashkenazi Jewish ethnicity and lactose intolerance in people of East Asian, West African, and Arab descent) . This study aims to provide an overview of direct-to-consumer pharmacogenomic tests (DTC-PTs) as one of the health-related services for DTC-GT companies and discuss the strategies that might be beneficial for improving the market usability and clinical outcomes of such tests. 1.2. Pharmacogenomics and Its Integration in DTC Companies Pharmacogenomic (PGx) tests constitute one of the important genetic testing services of DTC-GT companies. PGx tests reveal genetic variations that can be linked to the efficacy and/or responses to drugs; therefore, most people are interested in finding out about their genome function concerning their drug intake. As a potential molecular risk factor, PGx variants may affect several medication processes and bring about the different outcomes of safety and efficacy for assigned treatment approaches. Studies have reported that almost all people have at least one actionable functional variant in their genes for drug pharmacokinetics and pharmacodynamics . Most pharmacovariants are categorized as polymorphisms through the human genome; therefore, they may show no discernable phenotype until the time for drug utilization by individuals. Hence, pre-emptive genotyping and providing the result (by DTC companies) would be extremely beneficial for the patients who refer to the clinic later. Indeed, drug-related gene scanning can provide the information before any prescription and clinical decision. Besides, the PGx test data can be used as a lifetime predictive tool for drug safety and efficacy. PGx profiling (not as DTC) is a routine test in some clinical laboratories and hospitals (e.g., Mayo Clinic, St. Jude Children’s Research Hospital, Vanderbilt University Medical Center, etc.) and soon it will become prevalent in many clinical centers through the updating of different provided guidelines . As the global need for PGx tests is increasingly acknowledged , more DTC-GT companies also will begin to provide PGx testing services in the near future. 1.3. Direct-to-Consumer Pharmacogenomic Tests (DTC-PT) Various companies are offering several different genetic tests and services, but some companies are offering only a few specific tests. Currently, PGx analysis of individuals as pre-emptive genetic profiling and screening is offered by just a few companies . Based on companies’ public pages and depending on test type, whether it is single, combined with other tests, or a whole genome test, the price ranges from USD 100 to USD 1000 in different centers. The various functional genetic variations (FGVs) in the genome are determined so that proper prescription and treatment decisions can be provided; this helps in realizing the dream of personalized and precision medicine (PPM). PGx tests were launched for the first time in the early 1990s, with the anticipation that they could be used as an approach for reducing many potential adverse drug reactions (ADRs) and the first FDA approval of such tests appeared in 2005 . Similarly to other DTC genetic tests, DTC-PTs also are evaluated and assessed by the FDA through monitoring the clinical and analytic validity in addition to consumer’s understanding and perception of the descriptive information for the tests and the related results without any professional healthcare intervention. The regulations for the test implementation are then managed and declared to the companies subsequently. As per the two pioneering companies in this field, 23andMe and Pathway Genomics, more than 91% of their customers showed FGVs . Today, different companies evaluate and profile different drug–gene pairs. Even the screening portfolio for a single company may vary in different countries. For example, 23andMe, the only DTC-PT company with FDA approval for three PGx markers without a physician’s prescription, has provided profiling for different genes in different countries before . This is because people with diverse ethnicities show different types of biomarkers for the same drug which could result in alternative responses and efficacy. Concerns have been raised by some civil society organizations and regulatory bodies such as the FDA about the lack of medical supervision for most of the DTC PGx tests, resulting in a reduction in the number of companies offering these services . While the industry was reshaped and down-sized by the FDA warning letters, the need to obtain clearance/approval for such tests placed was as the top priority issue for the offering companies. At present, a few companies are offering PGx tests, either directly to consumers (23andMe) or through a physician (e.g., Veritas) . However, because of the waiting time for an appointment with a healthcare provider to order the PGx tests, people are reluctant to spend time receiving test orders from the physicians. Hence, companies that offer PGx tests directly to consumers may become more popular and will become the most common mode of PGx testing in the future; especially when such tests can be organized as a pre-emptive genotyping approach for individuals. Such companies should provide additional information to both patients and physicians before the test. The interpreted data of tested PGx biomarkers and related literature alongside the test methodology should be provided by the companies on their websites so the essential scientific information will be available for customers before they order PGx tests . The results of PGx profiling by DTC companies may serve as an approach for increasing the efficiency of future prescribed drugs. Even though the test is performed only once, the results can be utilized for people’s whole lifetime. Below are some insights into the field which could improve the market usability and clinical outcomes for these tests. 1.4. Approaches to Improve the Clinical Outcomes of DTC-PT DTC-PT is a double-edged sword, because it can raise concerns about drug dosage adjustment if the results are misinterpreted but can truly help if handled properly. We observed the pitfalls of DTC-PT tests over time and propose different strategies to improve their clinical outcomes alongside the market usability. Here, we list some of the main challenges that should be addressed appropriately to improve the positive effects of the DTC-PT. The first and, maybe the most important of all, is the integration of physicians and other healthcare providers such as human/clinical geneticists and clinical pharmacologists in the test procedure because they can provide appropriate scientific and clinical information about the test itself and the results to consumers . This will improve the completeness and reliability of such companies. However, it has been stated that the final decision about ordering the tests should still be made by the customers themselves; the current PGx guidelines are just about the interpretation of test results and not about who should order the tests, or when . Companies can make information easily available to customers through advertisements, articles, brochures, personal contact, etc. However, most companies provide information about analytical and clinical validity and utility and test quality of DTC-PT through in vitro diagnostic (IVD) validated equipment (i.e., premarket approval code for approved devices if there is one) for the physicians, who ordered the tests. Such diagnostic tests also have been described earlier and are freely accessible to everyone . Then, all the companies which provide PGx profiling may consider and prepare relevant information for customers alongside specialized information for healthcare providers. After gaining public trust by providing the needed information to consumers, the companies’ efforts could be focused on personalizing the services. PGx variants may be highly dependent on the specific population; therefore, different alleles plus population-specific haplotype/diplotype for every pharmacogene must be considered in their tests. This will also make a huge impact on the clinical validity and utility of the tests . Currently, many companies use pre-prepared SNP array chips or different orthogonal PCR approaches as genotyping methods (23andMe, Genelex (Seattle, Washington, United States), etc.). Here, the different allele frequencies and linkage patterns between different ethnic groups must also be considered for PGx result analysis. For example, the frequency of poor metabolizer alleles for CYP2C19 is higher in East Asian ethnicities (14%) when compared to those with European (2%) and African (4%) ancestries. Even in those variants which were considered to have a relationship with specific drugs universally, it has been found that there is an effect of ethnicity. Some of the examples are warfarin and rs9923231 in the VKORC1 gene and abacavir and the HLA-B*57:01 allele . In such a scenario, the recommendation is the employment of hypothesis-free sequencing technologies (WES, WGS, or long-read sequencers) besides using any local variant datasets for obtained data interpretation. This might be necessary when there are no clear guidelines from reference organizations concerning the identified variants, only annotations. However, any type of existed reference data could be provided alongside the final result for the customers, who are recommended to consult with a clinician based on their test results. To incorporate the revealed FGVs into clinical practice, some important factors and information such as sample numbers, ethnic background, and efficacy rate on dosage modification must be considered by a referred physician . New approaches for companies to deal with these issues could include the integration of an interdisciplinary team consisting of different fields of study in companies’ properties; as such, team efficacy for patients’ safety has been reviewed before . The related team could comprise a clinical geneticist, laboratory geneticist, clinical pharmacologist, and a medical doctor in the test-providing group (scientific support section (SSS) in the company). Gathering such professional medical advisors maybe not widely available; therefore, the feasibility and economic implications of DTC companies for implementing the recommendation from SSS members alongside the tests for the customers could also be a matter of challenge for some corporations. However, it would increase the credibility of the PGx test service if those companies utilized the services of SSS before providing results to their customers. Moreover, the interpretation and follow-up recommendations can be provided by the SSS team. It will also be useful for the companies if the SSS members can engage with the scientific research community and scrutinize the provided publications to gain new insights into PGx tests . Other approaches to improve the quality of companies’ services include the preparation of a well-designed personalized electronic card containing PGx test results which can be accessed quickly and made available through a linked local FGV database . Finally, there are some general trends for providing optimal and comprehensive PGx test outcomes. The first is increasing the numbers of included pharmacovariants (either with a guideline or annotated) into the test by employing next-generation DNA sequencing or long-read sequencing technologies instead of current techniques, as well as SNP arrays for variant identification. For example, 23andMe mostly utilizes its SNP chip for genotyping 715,000 SNPs but their tests are incomplete because many new and/or previously reported informative variants for some main pharmacogenes have not been captured in their panel. For instance, the panel ignores some population-specific predictive PGx markers in HLA-B , IFNL4 , and TPMT genes. 23andMe however, reduced the number of pharmacogenes and involved variants significantly, as is mentioned in their related PGx portal . The screening and inclusion of new variants into the company’s medical and health-related tests also need a license from the relevant authorities. Furthermore, because several PGx markers can be found in intronic and regulatory elements of pharmacogenes and the presence of some insertion–deletions (InDels), copy number variations (CNVs), and pseudogenes in drug-related genes (e.g., CYP2D6 ), next-generation sequencing (NGS) and long read sequencer technologies would be the best choice for identifying such variants . Comprehensive NGS methods such as WES and WGS work as hypothesis-free approaches and will find most of the potential FGVs in drug genes. The incidental findings (IFs) and variants of unknown significance (VUS) could be ignored in the final result, because the DTC companies’ goals and general policies do not enter in research or diagnostic areas but only identifying those variants which have been offered for detection before. Nevertheless, the challenging variants may be followed for further analysis by the SSS teams in companies. The second approach for optimizing and improving DTC-PT services is having a list of most prescribed drugs locally and focusing on the related pharmacogenes. It will make the tests outline the personalized and precision medicine (PPM) area more than before. Probably, in this way, the result shows more annotated variants than those with a clear guideline. In this case, companies should add a disclaimer to their reports that there might be a limitation and changeable efficacy for a particular result. The third and last approach would be the participation of the company’s representatives in scientific events to obtain scientific credentials in the field. At present, most companies also seek customers’ informed consent to share the customers’ data. Such activities will fuel research and bring more credit and validity to companies . For instance, according to 23andMe, more than 80% of their customers agreed to use their genomic data in the medical research area. However, strict informed consent forms and clear and sufficient information on further activities by the companies must be provided before this . The approaches for improving the clinical outcomes of DTC-PTs are summarized in . 1.5. Ethical and Legal Considerations for DTC-PT DTC-PT by definition includes no healthcare supervision in the test procedure. In the last decade, DTC companies have made it easier for people around the world to access DTC PGx testing; however, the lack of clinical supervision has raised many concerns, especially when changes in drug ordering, dosage adjustment, and other treatment approaches are required, in addition to customers performing self-therapy. These concerns have made DTC-GT, and of course the PGx tests through the related companies, controversial since their advent in the market . Over time, many regulatory and legislative entities such as the FDA, EU parliament, and U.K. Human Genetics Commission have started to monitor and implement regulations and directives for such tests . In 2013, the FDA warned 23andMe to wait for pre-marketing assessments and approval laws for their tests. Based on the warning letter, the company ceased and desisted the PGx tests that were offered. However, in 2017, the FDA sent an approval letter to the company for including some specific personal genomics and health-related tests (involving PGx profiling). Currently, 23andMe is the only company which offers PGx tests that can be requested directly by the customers. However, the rules are different between countries. For example, in European countries, laws were enacted both at the national and EU levels. These regulations are described in detail in other publications . However, questions such as why the number of FDA-revised and approved pharmacogenetic biomarkers differs from those which could be offered by the DTC companies, remain unanswered regarding DTC-PT implementation through the offering centers. Additionally, consumers’ data storage and future utilization in other activities or shared with third parties would be a very important concern in customers’ privacy protection and confidentiality. While the recommendations and guidelines for DTC-GT health-related services are available in the statements of policymakers and observers as well as the European Society of Human Genetics (ESHG), Global Alliance for Genomics and Health (GA4GH), Nuffield Council on Bioethics (NCB), etc., the information on data storage times, sample disposal, and extra research activities still varies between different companies; unfortunately, some do not consistently meet the international guidelines on transparency, related to privacy and secondary use of customers’ data . Hence, the consumers’ privacy protections and expectations must be handled with care by the companies’ terms of use, laws, and regulations. Recommendations for this public controversy have previously been provided and highlighted, which may raise advanced discussions in the field . also lists some important considerations for companies.
Since the completion of the Human Genome Project (HGP), DNA sequencing tests for health-related purposes have become common in medical laboratories . Additionally, the advent of high-throughput sequencing methods has made DNA analysis tests faster and easier. During early 2000, some genetic and genomic companies started offering genetic testing directly to individuals without the need for the prescription of physicians or other healthcare providers. Direct-to-consumer genetic tests (DTC-GT) are defined as genetic tests marketed directly to customers through print or visual media or the internet, or that can be bought online or in brick-and-mortar stores with no/least involvement of healthcare professionals in this process . DTC-GT allows customers to access their genomic interpreted data whenever and wherever they want. However, over time, after some critical evaluations, companies selling DTC-GT started engaging geneticists and medical professionals’ in the form of pre-and post-test consultation for consumers (advertisements, articles, brochures, personal contact, etc.) to provide advice on further actions to be taken after the genetic testing results. Most companies declare in their policies and consent forms for consumers that such tests should not be considered as diagnostic tests, but only as informative tests . Although the DTC-GT tests will be more useful in preventive areas than diagnosis issues, the test outcomes can demonstrate significant help for the future clinical assessments of individuals through providing the healthcare-related result before they visit a specialist. A DTC-GT report can bring attention to a specific condition in people, which may need more consideration and clinical confirmation or intervention by the clinicians . Today, DTC-GT services are available as kits for obtaining saliva or buccal swab samples which are non-invasive. The samples can then be sent to the company providing the service where DNA analysis is performed, usually in CAP and/or CLIA accredited laboratories. The companies mostly using array-based or sequencing platforms (targeted gene sequencing panels or broad range genomic tests such as whole-genome sequencing (WGS) or whole-exome sequencing (WES)) approaches for analyzing specific mutations or providing a comprehensive picture for genomic variants in a given sample. Each company applies its microarray or sequencing technologies (i.e., Illumina HumanOmniExpress-24 single nucleotide polymorphism (SNP) chip for 23andMe and WES for Genos) with post-processing involving imputation, and the interpreted genomic data then are returned to customers via the internet or mail after a couple of weeks or months . Some companies such as 23andMe (Sunnyvale, CA, USA), Color Genomics (Burlingame, CA, USA), etc. also provide raw genetic data to their customers, so the customers can use this raw genetic data for further processing and analysis through free online resources and tools such as Promethease, Live Wello, Genetic Genie, etc. with the help of a physician, clinical geneticist, genetic counselor, pharmacist or other trained genetic professionals. Currently, DTC-GT companies offer their services in two main categories which include medical and non-medical genetic tests. The medical genetic tests can be classified as carrier tests (e.g., hemoglobinopathies), disease susceptibility detection tests (e.g., Parkinson’s disease), pharmacogenomic tests (for the specific number of drug-related genes), life-style related tests (genetic analysis for complex diseases), and prenatal tests (PND & PGD). They can also be divided into tests for monogenic disorders, polygenic defects, multifactorial diseases, genome-wide testing (thousands of SNPs), and broad range tests (WGS & WES). Medical genetic testing services are the most common tests used by people and are the main reasons for the increasing growth of DTC-GT companies. Non-medical testing services, on the other hand, consist of testing for some traits and features in individuals, which are not necessarily related to disease or health, and are usually for “infotainment”. Examples of these include ancestry information, ear lobe attachment, and the flush reaction after drinking alcohol, etc. However, there is an argument that ancestry data should be included under medical information, because this information helps to determine whether a specific ethnic group has a predisposition to a particular genetic condition (e.g., Tay-Sachs disease in individuals of Ashkenazi Jewish ethnicity and lactose intolerance in people of East Asian, West African, and Arab descent) . This study aims to provide an overview of direct-to-consumer pharmacogenomic tests (DTC-PTs) as one of the health-related services for DTC-GT companies and discuss the strategies that might be beneficial for improving the market usability and clinical outcomes of such tests.
Pharmacogenomic (PGx) tests constitute one of the important genetic testing services of DTC-GT companies. PGx tests reveal genetic variations that can be linked to the efficacy and/or responses to drugs; therefore, most people are interested in finding out about their genome function concerning their drug intake. As a potential molecular risk factor, PGx variants may affect several medication processes and bring about the different outcomes of safety and efficacy for assigned treatment approaches. Studies have reported that almost all people have at least one actionable functional variant in their genes for drug pharmacokinetics and pharmacodynamics . Most pharmacovariants are categorized as polymorphisms through the human genome; therefore, they may show no discernable phenotype until the time for drug utilization by individuals. Hence, pre-emptive genotyping and providing the result (by DTC companies) would be extremely beneficial for the patients who refer to the clinic later. Indeed, drug-related gene scanning can provide the information before any prescription and clinical decision. Besides, the PGx test data can be used as a lifetime predictive tool for drug safety and efficacy. PGx profiling (not as DTC) is a routine test in some clinical laboratories and hospitals (e.g., Mayo Clinic, St. Jude Children’s Research Hospital, Vanderbilt University Medical Center, etc.) and soon it will become prevalent in many clinical centers through the updating of different provided guidelines . As the global need for PGx tests is increasingly acknowledged , more DTC-GT companies also will begin to provide PGx testing services in the near future.
Various companies are offering several different genetic tests and services, but some companies are offering only a few specific tests. Currently, PGx analysis of individuals as pre-emptive genetic profiling and screening is offered by just a few companies . Based on companies’ public pages and depending on test type, whether it is single, combined with other tests, or a whole genome test, the price ranges from USD 100 to USD 1000 in different centers. The various functional genetic variations (FGVs) in the genome are determined so that proper prescription and treatment decisions can be provided; this helps in realizing the dream of personalized and precision medicine (PPM). PGx tests were launched for the first time in the early 1990s, with the anticipation that they could be used as an approach for reducing many potential adverse drug reactions (ADRs) and the first FDA approval of such tests appeared in 2005 . Similarly to other DTC genetic tests, DTC-PTs also are evaluated and assessed by the FDA through monitoring the clinical and analytic validity in addition to consumer’s understanding and perception of the descriptive information for the tests and the related results without any professional healthcare intervention. The regulations for the test implementation are then managed and declared to the companies subsequently. As per the two pioneering companies in this field, 23andMe and Pathway Genomics, more than 91% of their customers showed FGVs . Today, different companies evaluate and profile different drug–gene pairs. Even the screening portfolio for a single company may vary in different countries. For example, 23andMe, the only DTC-PT company with FDA approval for three PGx markers without a physician’s prescription, has provided profiling for different genes in different countries before . This is because people with diverse ethnicities show different types of biomarkers for the same drug which could result in alternative responses and efficacy. Concerns have been raised by some civil society organizations and regulatory bodies such as the FDA about the lack of medical supervision for most of the DTC PGx tests, resulting in a reduction in the number of companies offering these services . While the industry was reshaped and down-sized by the FDA warning letters, the need to obtain clearance/approval for such tests placed was as the top priority issue for the offering companies. At present, a few companies are offering PGx tests, either directly to consumers (23andMe) or through a physician (e.g., Veritas) . However, because of the waiting time for an appointment with a healthcare provider to order the PGx tests, people are reluctant to spend time receiving test orders from the physicians. Hence, companies that offer PGx tests directly to consumers may become more popular and will become the most common mode of PGx testing in the future; especially when such tests can be organized as a pre-emptive genotyping approach for individuals. Such companies should provide additional information to both patients and physicians before the test. The interpreted data of tested PGx biomarkers and related literature alongside the test methodology should be provided by the companies on their websites so the essential scientific information will be available for customers before they order PGx tests . The results of PGx profiling by DTC companies may serve as an approach for increasing the efficiency of future prescribed drugs. Even though the test is performed only once, the results can be utilized for people’s whole lifetime. Below are some insights into the field which could improve the market usability and clinical outcomes for these tests.
DTC-PT is a double-edged sword, because it can raise concerns about drug dosage adjustment if the results are misinterpreted but can truly help if handled properly. We observed the pitfalls of DTC-PT tests over time and propose different strategies to improve their clinical outcomes alongside the market usability. Here, we list some of the main challenges that should be addressed appropriately to improve the positive effects of the DTC-PT. The first and, maybe the most important of all, is the integration of physicians and other healthcare providers such as human/clinical geneticists and clinical pharmacologists in the test procedure because they can provide appropriate scientific and clinical information about the test itself and the results to consumers . This will improve the completeness and reliability of such companies. However, it has been stated that the final decision about ordering the tests should still be made by the customers themselves; the current PGx guidelines are just about the interpretation of test results and not about who should order the tests, or when . Companies can make information easily available to customers through advertisements, articles, brochures, personal contact, etc. However, most companies provide information about analytical and clinical validity and utility and test quality of DTC-PT through in vitro diagnostic (IVD) validated equipment (i.e., premarket approval code for approved devices if there is one) for the physicians, who ordered the tests. Such diagnostic tests also have been described earlier and are freely accessible to everyone . Then, all the companies which provide PGx profiling may consider and prepare relevant information for customers alongside specialized information for healthcare providers. After gaining public trust by providing the needed information to consumers, the companies’ efforts could be focused on personalizing the services. PGx variants may be highly dependent on the specific population; therefore, different alleles plus population-specific haplotype/diplotype for every pharmacogene must be considered in their tests. This will also make a huge impact on the clinical validity and utility of the tests . Currently, many companies use pre-prepared SNP array chips or different orthogonal PCR approaches as genotyping methods (23andMe, Genelex (Seattle, Washington, United States), etc.). Here, the different allele frequencies and linkage patterns between different ethnic groups must also be considered for PGx result analysis. For example, the frequency of poor metabolizer alleles for CYP2C19 is higher in East Asian ethnicities (14%) when compared to those with European (2%) and African (4%) ancestries. Even in those variants which were considered to have a relationship with specific drugs universally, it has been found that there is an effect of ethnicity. Some of the examples are warfarin and rs9923231 in the VKORC1 gene and abacavir and the HLA-B*57:01 allele . In such a scenario, the recommendation is the employment of hypothesis-free sequencing technologies (WES, WGS, or long-read sequencers) besides using any local variant datasets for obtained data interpretation. This might be necessary when there are no clear guidelines from reference organizations concerning the identified variants, only annotations. However, any type of existed reference data could be provided alongside the final result for the customers, who are recommended to consult with a clinician based on their test results. To incorporate the revealed FGVs into clinical practice, some important factors and information such as sample numbers, ethnic background, and efficacy rate on dosage modification must be considered by a referred physician . New approaches for companies to deal with these issues could include the integration of an interdisciplinary team consisting of different fields of study in companies’ properties; as such, team efficacy for patients’ safety has been reviewed before . The related team could comprise a clinical geneticist, laboratory geneticist, clinical pharmacologist, and a medical doctor in the test-providing group (scientific support section (SSS) in the company). Gathering such professional medical advisors maybe not widely available; therefore, the feasibility and economic implications of DTC companies for implementing the recommendation from SSS members alongside the tests for the customers could also be a matter of challenge for some corporations. However, it would increase the credibility of the PGx test service if those companies utilized the services of SSS before providing results to their customers. Moreover, the interpretation and follow-up recommendations can be provided by the SSS team. It will also be useful for the companies if the SSS members can engage with the scientific research community and scrutinize the provided publications to gain new insights into PGx tests . Other approaches to improve the quality of companies’ services include the preparation of a well-designed personalized electronic card containing PGx test results which can be accessed quickly and made available through a linked local FGV database . Finally, there are some general trends for providing optimal and comprehensive PGx test outcomes. The first is increasing the numbers of included pharmacovariants (either with a guideline or annotated) into the test by employing next-generation DNA sequencing or long-read sequencing technologies instead of current techniques, as well as SNP arrays for variant identification. For example, 23andMe mostly utilizes its SNP chip for genotyping 715,000 SNPs but their tests are incomplete because many new and/or previously reported informative variants for some main pharmacogenes have not been captured in their panel. For instance, the panel ignores some population-specific predictive PGx markers in HLA-B , IFNL4 , and TPMT genes. 23andMe however, reduced the number of pharmacogenes and involved variants significantly, as is mentioned in their related PGx portal . The screening and inclusion of new variants into the company’s medical and health-related tests also need a license from the relevant authorities. Furthermore, because several PGx markers can be found in intronic and regulatory elements of pharmacogenes and the presence of some insertion–deletions (InDels), copy number variations (CNVs), and pseudogenes in drug-related genes (e.g., CYP2D6 ), next-generation sequencing (NGS) and long read sequencer technologies would be the best choice for identifying such variants . Comprehensive NGS methods such as WES and WGS work as hypothesis-free approaches and will find most of the potential FGVs in drug genes. The incidental findings (IFs) and variants of unknown significance (VUS) could be ignored in the final result, because the DTC companies’ goals and general policies do not enter in research or diagnostic areas but only identifying those variants which have been offered for detection before. Nevertheless, the challenging variants may be followed for further analysis by the SSS teams in companies. The second approach for optimizing and improving DTC-PT services is having a list of most prescribed drugs locally and focusing on the related pharmacogenes. It will make the tests outline the personalized and precision medicine (PPM) area more than before. Probably, in this way, the result shows more annotated variants than those with a clear guideline. In this case, companies should add a disclaimer to their reports that there might be a limitation and changeable efficacy for a particular result. The third and last approach would be the participation of the company’s representatives in scientific events to obtain scientific credentials in the field. At present, most companies also seek customers’ informed consent to share the customers’ data. Such activities will fuel research and bring more credit and validity to companies . For instance, according to 23andMe, more than 80% of their customers agreed to use their genomic data in the medical research area. However, strict informed consent forms and clear and sufficient information on further activities by the companies must be provided before this . The approaches for improving the clinical outcomes of DTC-PTs are summarized in .
DTC-PT by definition includes no healthcare supervision in the test procedure. In the last decade, DTC companies have made it easier for people around the world to access DTC PGx testing; however, the lack of clinical supervision has raised many concerns, especially when changes in drug ordering, dosage adjustment, and other treatment approaches are required, in addition to customers performing self-therapy. These concerns have made DTC-GT, and of course the PGx tests through the related companies, controversial since their advent in the market . Over time, many regulatory and legislative entities such as the FDA, EU parliament, and U.K. Human Genetics Commission have started to monitor and implement regulations and directives for such tests . In 2013, the FDA warned 23andMe to wait for pre-marketing assessments and approval laws for their tests. Based on the warning letter, the company ceased and desisted the PGx tests that were offered. However, in 2017, the FDA sent an approval letter to the company for including some specific personal genomics and health-related tests (involving PGx profiling). Currently, 23andMe is the only company which offers PGx tests that can be requested directly by the customers. However, the rules are different between countries. For example, in European countries, laws were enacted both at the national and EU levels. These regulations are described in detail in other publications . However, questions such as why the number of FDA-revised and approved pharmacogenetic biomarkers differs from those which could be offered by the DTC companies, remain unanswered regarding DTC-PT implementation through the offering centers. Additionally, consumers’ data storage and future utilization in other activities or shared with third parties would be a very important concern in customers’ privacy protection and confidentiality. While the recommendations and guidelines for DTC-GT health-related services are available in the statements of policymakers and observers as well as the European Society of Human Genetics (ESHG), Global Alliance for Genomics and Health (GA4GH), Nuffield Council on Bioethics (NCB), etc., the information on data storage times, sample disposal, and extra research activities still varies between different companies; unfortunately, some do not consistently meet the international guidelines on transparency, related to privacy and secondary use of customers’ data . Hence, the consumers’ privacy protections and expectations must be handled with care by the companies’ terms of use, laws, and regulations. Recommendations for this public controversy have previously been provided and highlighted, which may raise advanced discussions in the field . also lists some important considerations for companies.
Today, PGx studies provide a lot of information from drug–gene pairs. Just a couple of years ago, fewer than 80 drugs had PGx tags, but now there are more than 260 unique drugs with PGx recommendations, available through organizations such as CPIC, DPWG, CPND, and the FDA . The labels include actionable PGx, testing required, testing recommended, and informative PGx. The latter, according to PharmGKB, means that particular variants or metabolizer phenotypes do not affect a drug’s efficacy, dosage, metabolism, or toxicity. However, the lack of sufficient education for physicians and inaccessibility to such tests alongside lack of insurance coverage are major issues for customers and are barriers to making PGx profiling a routine and mandatory test. Additionally, inconsistencies in clinical pharmacogenetic recommendations among major sources exist, which may slow the clinical implementation of such test results. The prevalence and type of these inconsistencies have been comprehensively analyzed before . Over time, factors such as pharmacogenetic information accessibility before physician order and the pre-symptom identification of PGx biomarkers helping to provide better personalized clinical decision-making thereby reducing drugs’ side effects and adverse reactions, etc., has led to the rise of DTC PGx testing . However, challenges such as clinicians’ lack of education to understand the results, the non-availability of a geneticist or genetic counselor in most clinics, the interpretation of results based on just the previous genome-wide association studies, no consideration of family and background risks and people’s lifestyle, ethical and legal issues, the utility and validity of tests, potential misinterpretations of results, possible wrong decisions by healthcare providers due to the incorrect interpretation of results, and a lack of adequate clinical evidence for most identified FGVs led to the downfall of DTC-PTs . However, after the FDA approval letter to a company including some PGx tests among other tests, PGx tests came back into the limelight and PGx genotyping started to rise again. Nowadays, there are some companies that perform DTC-PTs for their customers as one of the main options between other healthcare-related tests , although, there are still some scientific considerations that must be addressed by these companies. For example, many genetic effects and modifications such as the presence of rare recombination events in specific populations, dominant negative effects of mutations in genes in drug metabolic pathways, gene duplication occasions in populations, epigenetic signatures in people, epistasis occurrence, variable expressivity through intra- and inter-families, and incomplete penetrance effects for some genetic alterations receive no attention during PGx tests. Therefore, the provided results may not depict the true potential of pharmacogenetic profiling and functioning of individuals . Additionally, different specificity and sensitivity, mainly because of diverse genetic variations or allele/haplotype frequencies for the tested pharmacogenes, in addition to the presence of any linkage disequilibrium between the tested SNPs in population-specific panels and possible findings through comprehensive genotyping approaches such as WES and WGS, plus long-read sequencer outcomes, should be taken into account. If any companies would like to be the pioneer and/or frontier in providing the most accurate DTC-PT services among the others, such a genetic analysis must be handled via the SSS team before declaring the final result to customers or their healthcare providers . For FGVs with the guideline, the task is clear for incorporation into clinical practice, but for the annotated variants more research evidence is required. Here, companies may use different approaches for reaching this goal. For instance, 23andMe considered at least three papers for considering the variance in their PGx testing . The challenge is when the clinical relevance of the variant is proved in just one published document. However, during the provision of personalized treatment for patients, it is better to also consider a clinicians’ opinion, which might be the exact genetic alteration for ADR in the specific patient(s). Nevertheless, alongside all these issues, companies must always remember that they should not cause any unnecessary concerns or anxiety for people. There were no legal concerns about the DTC PGx tests before, which meant that many people were willing to perform genetic profiling for themselves. This indicates the consumers’ desire for such tests and is a reminder of the fact that if there were proper regulations and directives by the governmental and legislative bodies, both patients and doctors would use PGx data for personalized prescribing, especially for the high-risk gene–drug pairs. Nevertheless, some basic issues as well as the lack of evidence-based guidelines for the use of PGx testing, such as the potential liability if prescribers do not consult PGx test results for every medication prescribed, if non-affordable medication is indicated by the customers’ PGx test results, or if there is no access to the medication recommended. Today, there is an increasing trend for using bioinformatic tools and frameworks without any need for background knowledge of complicated programming languages and scriptwriting for Linux/Ubuntu systems or Python (SUSHI of ETH Zurich, VarSeq of Golden Helix, etc.) . Therefore, the utilization of high throughput sequencing technologies and data analysis and interpretation has become easier and more common for companies. Because of that, more DTC companies may use comprehensive genotyping approaches. Soon, PGx tests may be ordered by customers and final results will be available before visiting their doctors, making treatment decisions faster.
The authorities’ regulations and the companies’ trends for providing different DTC genetic services are changing rapidly. DTC-PT may show the potential impact for becoming an essential tool for providing drug-related genetic variation information. More support in the form of funding (NIH), education (Gene-Equip, NICE), and dataset preparation (Illumina, San Diego, California, United States) are expected soon . Advances in technology bring a broader range of gene–drug interactions to companies’ local panels. Web-based interpretation services and smartphone applications, as well as cost lowering for PGx tests, make them very common and accessible everywhere. Lifetime and free consultations of DTC-PT results will be offered by these companies. In this exciting area, improving the clinical related outcomes and market usability of PGx tests could be guaranteed by the SSS members of companies. Soon, we may witness the smart future of PPM, where the pre-emptive PGx tests apply as routine tests by a majority of the population.
|
Genomic Multicopy Loci Targeted by Current Forensic Quantitative PCR Assays | 403ace7d-f18b-4532-b787-cd3528ad86b5 | 11507060 | Forensic Medicine[mh] | Forensic short tandem repeat (STR) analysis requires optimal DNA amounts. Too low amounts will fall below the analytical sensitivity and entail stochastic effects, resulting in a loss of information; too high amounts will cause analytical artefacts. To adjust the optimal DNA amounts for the subsequent analysis and to identify promising or unsuitable samples, it is important to quantitate the DNA that has been extracted from trace material. In human forensics, the quantitation methods should specifically measure human (and not contaminating microbial, animal- or plant-derived) DNA. Furthermore they should be highly sensitive such that only a small proportion of a precious sample is consumed for the quantitation. For these reasons, nowadays, quantitative PCR (qPCR)-based methods are used. These offer the further advantage of being able to simultaneously analyze several target DNA loci in a multiplex analysis. By this means they can provide additional information, such as on DNA degradation and the presence of PCR inhibitors or male components, while at the same time consuming less sample and reducing the time of the analysis. In this review, the technical principles of qPCR methods currently used in forensics are briefly explained, followed by considerations on the rationales of the multiplex assay design. Finally, the DNA loci targeted by current qPCR assays are discussed. Quantitative PCR is a relative quantitation method based on comparing measurements of an unknown sample with measurements of standard samples of known DNA concentrations to calculate the concentration of the unknown. The principle consists in monitoring the accumulation of PCR products in real time and determining the number of PCR cycles required to reach a certain threshold (called the threshold cycle, C t , or quantification cycle, C q ), typically set in the exponential phase of the PCR . The higher the template DNA amount, the fewer PCR cycles are required to reach the amplification threshold and thus the lower the C t value will be. In its simplest form, product accumulation is measured after the elongation step of each PCR cycle with the help of a DNA-binding fluorescent dye present in the reaction mix, such as SYBR green . As such dyes cannot distinguish between different PCR products, SYBR green-based qPCR cannot be multiplexed, and the specificity has to be confirmed by a further method (such as melting curve analysis ). For these reasons, forensic qPCR assays are not based on DNA-binding fluorescent dyes. In most forensic qPCR assays, the detection of PCR products is accomplished using dual-labeled hydrolysis probes (also called TaqMan probes) that moreover confer an additional level of target specificity . These are short oligonucleotides present in the PCR reaction that bind to one strand of the target amplicon between the primer binding sequences (see a). On one end, the probes are covalently linked to a fluorophore, the fluorescence emission of which, however, is quenched by a second fluorophore (called quencher) that is attached at the other end of the probe. By its 5′–3′ exonuclease activity, the passing Taq polymerase degrades the probe, thus releasing the quencher from the fluorophore. The light emission by the flurophore is then monitored at the end of the elongation step. A further occasionally used method (called Plexor technology) is based on the specific base pairing between nucleotides not normally present in DNA . Here, one of the PCR primers contains a fluorophore-labeled unusual base at its 5′ end that will not base pair with standard nucleotides (see b). In addition to the four standard nucleotides, the PCR mix contains a second non-standard nucleotide that is covalently linked to a quencher and is specifically base pairing with the unusual nucleotide at the primer-derived 5′ ends of the amplified templates. By incorporating this second nucleotide at the 3′ end of the newly synthesized strand, the signal from the fluorophore at the 5′ end of the template is quenched, allowing for monitoring the PCR product accumulation based on the decreasing fluorescence signal after each elongation step. Modern forensic DNA quantitation assays are designed as multiplex qPCR assays that provide quantitative information on several important parameters, such as DNA amount, DNA integrity, the presence of PCR inhibitors, and the male component in DNA mixtures. Specific amplicons for these parameters are amplified in parallel and detected in separate color channels. In addition, some assays contain a passive reference dye to control the amount of reaction mix of each sample. Apart from saving time and sample amount, the multiplex design has the advantage that all parameters are assessed from the same sample fraction, thus increasing accuracy. In the following sections, the principles of the analysis of the different parameters are explained. 3.1. Sensitivity and Specificity Modern forensic DNA analysis is able to establish complete STR profiles from as little as 125 pg genomic DNA, corresponding to the DNA of nineteen diploid cells (or 19 diploid genome copies, each 6.6 pg nuclear DNA) . For even lower DNA amounts, modified protocols have been developed, and thus, it is possible to obtain information from just four to five cells (or genome copies) or even fewer . Correspondingly, qPCR methods need to be able to reliably quantitate in the picogram range. While PCR is in principle able to amplify single DNA molecules, if single-copy loci or (multicopy targets from a single DNA locus, as shown in a) were chosen for quantitation, any such analysis of low template DNA would be at risk of stochastic sampling errors , thus possibly underestimating the true DNA amount (see a for explanation). To be able to quantitate in the picogram range and to minimize stochastic sampling errors, current forensic qPCR assays analyze genomic loci that are present in many copies per genome and are uniformly distributed across several chromosomes. Thus, irrespective of the genome fraction ending up in the subsample taken for quantitation, the quantitation reflects the overall amount of DNA in the original sample (see b). The human genome contains many such multicopy loci that sufficiently differ in sequence from genomic loci of non-human species to allow for the design of human-specific and highly sensitive DNA quantitation assays . Species cross reactivity has yet to be empirically tested when establishing and validating forensic qPCR assays. Current forensic qPCR assays can accurately quantitate DNA of just 2.5–5 pg/µL , and their sensitivity, expressed as the limit of detection (LOD), is even below 1 pg/µL for the latest assays . Of note, DNA concentrations less than 1 pg/µL are indicative of only a few genome copies present in the sample, resulting in incomplete STR profiles due to stochastic sampling effects . However, at low DNA concentrations, the accuracy of the qPCR assays is low; hence, the measurement may underestimate the true amount of DNA in the sample that may still yield a complete STR profile . Moreover, DNA samples with quantitation results of less than 1 pg/µL bear the stochastic risk that the subsample taken for quantitation has removed copies of some of the STR loci to be analyzed, thus augmenting allelic imbalances or causing allele or locus drop-outs in the subsequent STR analysis (see b). Thus, for singular traces that are expected to yield very low amounts of DNA (such as touch DNA or single hair shafts ), it may be advisable to dispense with the quantitation in order to increase the chance of a successful STR analysis when applying low copy DNA methods . 3.2. DNA Integrity The term DNA integrity expresses the intactness of genomic DNA as required for a type of downstream analysis. In forensic STR analysis, information on DNA integrity can thus serve as a predictor of DNA typing success and can help identifying samples where alternative DNA markers or DNA typing methods might be required . DNA integrity is inversely related to DNA degradation, which is often discussed in terms of single or double strand breaks, caused by microbial or tissue-derived DNases, affecting the STR amplicons. However, environmentally caused chemical damage of the DNA, such as the oxidation, crosslinking, or hydrolysis of bases, will likewise affect STR analysis by impairing primer binding or strand elongation and is as well part of DNA degradation . DNA strand breaks or chemical damage is more likely to occur the longer an amplicon is. The idea to assess DNA integrity by qPCR consists in the quantitation of two amplicons of different lengths, one longer amplicon sensitive to degradation (called the degradation target) and a short one relatively unaffected by degradation (called the quantitation target) . As only intact DNA will be PCR-amplified, the ratio between the two quantitation results correlates with the DNA degradation and is expressed as degradation index (DI) (see ). The long amplicon of such a qPCR assay typically has a length in the range of the longer STR amplicons of the STR assays, and the short amplicon has a length below the smallest STR amplicon to be analyzed. The typical size range of STR amplicons is from about 100 to 350 bp. Thus, if the degradation target is affected, the longer STR amplicons will be affected as well, resulting in a loss of information. In first qPCR assays quantitating DNA degradation, two single-copy loci of different amplicon lengths were analyzed . Nowadays, to increase the analytical sensitivity of such assays and to avoid stochastic effects, the quantitation of two differently sized amplicons of multicopy loci is used . Due to their uniform distribution across several chromosomal locations, it is implicitly assumed that their degradation reflects the degradation also of the STR loci of interest and can thus be used to predict STR typing success. The two amplicons should not overlap in order to avoid unpredictable amplification effects due to interference, competition, or amplification of the shorter amplicon from the longer PCR product. This can be achieved by designing non-overlapping amplicons of the same locus or by using two different multicopy loci with roughly similar copy numbers. 3.3. PCR Inhibition Depending on the DNA extraction protocol and the source of DNA, forensic DNA samples may contain impurities that impair PCR by various molecular mechanisms . It is useful to know about the presence of such so-called PCR inhibitors as their presence may mislead the qPCR-based quantitation and likely impairs STR typing as well, such that the sample would not be taken further to STR analysis, or additional measures might be envisaged to overcome the inhibition . PCR inhibition affects longer amplicons more strongly than the shorter ones ; thus, the presence of inhibitors may more strongly affect the degradation target (see ) and misleadingly suggest DNA degradation. Attempts to deal with degradation by increasing the sample amount will thus introduce even more inhibitor and impair the STR analysis even further. To detect PCR inhibition, modern forensic qPCR assays contain a so-called internal PCR control (IPC) that is a synthetic template DNA of a known amount that is efficiently PCR-amplified by a dedicated primer pair present in the reaction mix. The amplification of the IPC is sensitive to PCR inhibitors; thus, a shift to higher C t values than to be expected for the IPC input is indicative of PCR inhibitors present in the sample. Modern commercial STR assays are rendered robust against PCR inhibitors and can deal with PCR-inhibiting impurities up to a certain concentration . Thus, the amplification of the IPC should be sensitive to PCR inhibitor concentrations above those tolerated by STR assays. A comparison of current qPCR assays has revealed that while they all were equally able to detect the presence of PCR inhibitors, for some assays the DNA quantitation results were affected by higher inhibitor concentrations . 3.4. Male Contributors In sexual assault cases, intimate swabs are analyzed that typically contain a mixture of victim DNA and perpetrator-derived DNA . In the majority of cases, the victims are female and the perpetrators are male. To see whether a DNA analysis might have a chance to reveal the male perpetrator by typing autosomal STRs or Y-chromosomal STRs (Y-STRs), it is useful to quantitate the proportion of male-derived DNA in the total extract . Moreover, for some forensic questions, such as on archeological or historic samples, Y-STR typing may yield useful information, e.g., on genealogy or family relations . To these ends, modern qPCR multiplexes target Y-chromosomal sequences, typically from Y-specific multicopy loci to increase the analytical sensitivity. Modern forensic DNA analysis is able to establish complete STR profiles from as little as 125 pg genomic DNA, corresponding to the DNA of nineteen diploid cells (or 19 diploid genome copies, each 6.6 pg nuclear DNA) . For even lower DNA amounts, modified protocols have been developed, and thus, it is possible to obtain information from just four to five cells (or genome copies) or even fewer . Correspondingly, qPCR methods need to be able to reliably quantitate in the picogram range. While PCR is in principle able to amplify single DNA molecules, if single-copy loci or (multicopy targets from a single DNA locus, as shown in a) were chosen for quantitation, any such analysis of low template DNA would be at risk of stochastic sampling errors , thus possibly underestimating the true DNA amount (see a for explanation). To be able to quantitate in the picogram range and to minimize stochastic sampling errors, current forensic qPCR assays analyze genomic loci that are present in many copies per genome and are uniformly distributed across several chromosomes. Thus, irrespective of the genome fraction ending up in the subsample taken for quantitation, the quantitation reflects the overall amount of DNA in the original sample (see b). The human genome contains many such multicopy loci that sufficiently differ in sequence from genomic loci of non-human species to allow for the design of human-specific and highly sensitive DNA quantitation assays . Species cross reactivity has yet to be empirically tested when establishing and validating forensic qPCR assays. Current forensic qPCR assays can accurately quantitate DNA of just 2.5–5 pg/µL , and their sensitivity, expressed as the limit of detection (LOD), is even below 1 pg/µL for the latest assays . Of note, DNA concentrations less than 1 pg/µL are indicative of only a few genome copies present in the sample, resulting in incomplete STR profiles due to stochastic sampling effects . However, at low DNA concentrations, the accuracy of the qPCR assays is low; hence, the measurement may underestimate the true amount of DNA in the sample that may still yield a complete STR profile . Moreover, DNA samples with quantitation results of less than 1 pg/µL bear the stochastic risk that the subsample taken for quantitation has removed copies of some of the STR loci to be analyzed, thus augmenting allelic imbalances or causing allele or locus drop-outs in the subsequent STR analysis (see b). Thus, for singular traces that are expected to yield very low amounts of DNA (such as touch DNA or single hair shafts ), it may be advisable to dispense with the quantitation in order to increase the chance of a successful STR analysis when applying low copy DNA methods . The term DNA integrity expresses the intactness of genomic DNA as required for a type of downstream analysis. In forensic STR analysis, information on DNA integrity can thus serve as a predictor of DNA typing success and can help identifying samples where alternative DNA markers or DNA typing methods might be required . DNA integrity is inversely related to DNA degradation, which is often discussed in terms of single or double strand breaks, caused by microbial or tissue-derived DNases, affecting the STR amplicons. However, environmentally caused chemical damage of the DNA, such as the oxidation, crosslinking, or hydrolysis of bases, will likewise affect STR analysis by impairing primer binding or strand elongation and is as well part of DNA degradation . DNA strand breaks or chemical damage is more likely to occur the longer an amplicon is. The idea to assess DNA integrity by qPCR consists in the quantitation of two amplicons of different lengths, one longer amplicon sensitive to degradation (called the degradation target) and a short one relatively unaffected by degradation (called the quantitation target) . As only intact DNA will be PCR-amplified, the ratio between the two quantitation results correlates with the DNA degradation and is expressed as degradation index (DI) (see ). The long amplicon of such a qPCR assay typically has a length in the range of the longer STR amplicons of the STR assays, and the short amplicon has a length below the smallest STR amplicon to be analyzed. The typical size range of STR amplicons is from about 100 to 350 bp. Thus, if the degradation target is affected, the longer STR amplicons will be affected as well, resulting in a loss of information. In first qPCR assays quantitating DNA degradation, two single-copy loci of different amplicon lengths were analyzed . Nowadays, to increase the analytical sensitivity of such assays and to avoid stochastic effects, the quantitation of two differently sized amplicons of multicopy loci is used . Due to their uniform distribution across several chromosomal locations, it is implicitly assumed that their degradation reflects the degradation also of the STR loci of interest and can thus be used to predict STR typing success. The two amplicons should not overlap in order to avoid unpredictable amplification effects due to interference, competition, or amplification of the shorter amplicon from the longer PCR product. This can be achieved by designing non-overlapping amplicons of the same locus or by using two different multicopy loci with roughly similar copy numbers. Depending on the DNA extraction protocol and the source of DNA, forensic DNA samples may contain impurities that impair PCR by various molecular mechanisms . It is useful to know about the presence of such so-called PCR inhibitors as their presence may mislead the qPCR-based quantitation and likely impairs STR typing as well, such that the sample would not be taken further to STR analysis, or additional measures might be envisaged to overcome the inhibition . PCR inhibition affects longer amplicons more strongly than the shorter ones ; thus, the presence of inhibitors may more strongly affect the degradation target (see ) and misleadingly suggest DNA degradation. Attempts to deal with degradation by increasing the sample amount will thus introduce even more inhibitor and impair the STR analysis even further. To detect PCR inhibition, modern forensic qPCR assays contain a so-called internal PCR control (IPC) that is a synthetic template DNA of a known amount that is efficiently PCR-amplified by a dedicated primer pair present in the reaction mix. The amplification of the IPC is sensitive to PCR inhibitors; thus, a shift to higher C t values than to be expected for the IPC input is indicative of PCR inhibitors present in the sample. Modern commercial STR assays are rendered robust against PCR inhibitors and can deal with PCR-inhibiting impurities up to a certain concentration . Thus, the amplification of the IPC should be sensitive to PCR inhibitor concentrations above those tolerated by STR assays. A comparison of current qPCR assays has revealed that while they all were equally able to detect the presence of PCR inhibitors, for some assays the DNA quantitation results were affected by higher inhibitor concentrations . In sexual assault cases, intimate swabs are analyzed that typically contain a mixture of victim DNA and perpetrator-derived DNA . In the majority of cases, the victims are female and the perpetrators are male. To see whether a DNA analysis might have a chance to reveal the male perpetrator by typing autosomal STRs or Y-chromosomal STRs (Y-STRs), it is useful to quantitate the proportion of male-derived DNA in the total extract . Moreover, for some forensic questions, such as on archeological or historic samples, Y-STR typing may yield useful information, e.g., on genealogy or family relations . To these ends, modern qPCR multiplexes target Y-chromosomal sequences, typically from Y-specific multicopy loci to increase the analytical sensitivity. Modern forensic qPCR assays analyze mostly multicopy loci that are present on the autosomes (and on the gonosomes) or specifically on the Y chromosome and by this means comply with the high sensitivity of modern STR kits. Recently, the diversity and copy number of multicopy DNA loci in various human populations has comprehensively been analyzed using data from the 1000 Genomes Project . Ideally, loci used for forensic qPCR assays should be present in dozens to thousands of copies, should be uniformly distributed across the genome (or the Y chromosome), should display little variation in copy number inter-individually or across populations, should show little sequence variation, should be human-specific, and should be robustly and specifically PCR-amplifiable. This section gives information on the multicopy target loci of current forensic qPCR assays. An overview is given in that also shows that for the latest forensic qPCR kits on the market, the identities, chromosomal localizations, and copy numbers of the target loci have not been disclosed. The tendency to keep information on target loci confidential will be critically discussed at the end of this section. 4.1. Transposable Elements Ideal candidates for qPCR-suited multicopy genes are retrotransposons, some of which are dispersed in high copy numbers throughout the genome and emerged only during primate or human evolution and thus are primate- or even human-specific . Retrotransposons are a subclass of transposons, genetic elements that are integrated in the genome and can, by enzymatically catalyzed mechanisms, “jump” to other chromosomal locations. Retrotransposons do so in a copy-paste fashion by being transcribed into an RNA that integrates by reverse transcription-based mechanisms at other places in the genome (reviewed in ). They can be classified as autonomous retrotransposons (LINEs, long interspersed elements) and non-autonomous retrotransposons to which short interspersed elements (SINEs) and SINE-VNTR-Alu elements (SVAs) belong. Various mechanisms evolved in organisms that counteract transposition. Yet, over time, by means of the copy-paste jumping mechanism, some retrotransposons have populated large parts of the genomes, thus contributing to evolution but in individuals also adversely affecting gene function and causing disease . L1 is the major LINE in primates. With a length of 6 kb and over 500,000 copies, L1 sequences make up roughly 17% of the human genome . LINEs encode the enzymatic machinery (e.g., reverse transcriptase) required for transposition. Non-autonomous retrotransposons do not encode these enzymes and jump with the help of the enzymes expressed by LINEs. The predominant SINEs in primates are the Alu elements (reviewed in ). Alu elements are non-coding and have sizes of about 300 bp and terminal poly(dA) tails with varying lengths that lead to length variations of the Alu elements . With about one million copies they make up roughly 10% of the human genome. Alu elements were acquired late during primate evolution and are thus specific for higher non-human primates and humans. Alu elements were among the first multicopy loci used in highly sensitive, human-specific forensic qPCR assays . There are three major Alu subfamilies that evolved in primates ( Alu J, Alu S, and Alu Y—evolutionary age in this order) and can be further subdivided into lineages. Of these, the Alu Yb lineage is the second largest young Alu Y group and has 1733 copies in the human genome that are distributed across all chromosomes . The AluYb8 is used in the commercial Innoquant HY kit, which also uses an SVA as the target locus (see and ). The SVAs consist of Alu -like sequences, followed by a variable number tandem repeat (VNTR) and a SINE-R-like sequence; they are the evolutionarily youngest group of retrotransposons and have only a few thousand copies per human genome, each with a size of about 2 kb . Differences in VNTR repeat numbers are responsible for inter-individual length variations. SVAs evolved in primates and are composed of six subfamilies, two of which are even human-specific . 4.2. Autosomal Multicopy Loci That Are Not Retrotransposons Multicopy loci other than retrotransposons have been established for forensic qPCR assays as well. Candidates for genetic loci that are present in high copy numbers are genes that encode non-translated RNAs, some of which are in high cellular demand to sustain cellular functioning. For the majority of these RNA genes the functional significance of their high copy numbers is not clear . One class of candidates for multicopy loci are genes for small nuclear RNAs (snRNAs), which are components of the spliceosomes involved in pre-mRNA splicing . The RNU2 locus is used in the commercial Plexor HY kit (see and ) and in the non-commercial NuMY assay . It encodes the U2 small nuclear RNA (snRNA). The RNU2 locus has been classified as a macrosatellite, i.e., a cluster of tandemly arranged repeat units that are longer than those of VNTR loci. Similar to VNTRs, macrosatellites display variability in repeat numbers within the population . The RNU2 gene cluster is located on chromosome 17 and consists of 20–40 RNU2 repeats, each with a length of 6.1 kb . Thus, RNU2 as a target for DNA quantification has the disadvantage of having far fewer copies than retrotransposons and of inter-individual variations in copy number. Furthermore, the localization at one chromosomal site increases the risk of stochastic sampling effects when analyzing low template DNA (see a), and it is not clear how far the degradation of the RNU2 locus reflects degradation at other chromosomal sites. Other autosomal non-coding multicopy target loci that have been suggested for forensic DNA quantitation are the 45S ribosomal RNA gene units that are clustered in tandem arrays in about 400 copies on five pairs of autosomes . Each 45S rRNA gene unit contains an array of 5.8S, 18S, and 28S rRNA genes that are separated by spacer sequences and are preceded by a common regulatory sequence. The units are separated by intergenic spacer sequences. While the functionally important rRNA genes are highly conserved across species, their regulatory and intergenic parts are less well conserved and have been used for designing a human-specific qPCR assay called RiboD assay. This assay targets two non-overlapping amplicons of 67 bp and 362 bp in the regulatory region to measure DNA amount and DNA integrity . So far, however, no multiplex assays have been established, and intraindividual and interindividual variations in the sequence and copy number of rRNA genes have been reported that might influence quantitation. An advantage of rRNA genes in forensic qPCR would be that their clusters reside on several chromosomes, thus minimizing stochastic sampling effects. Furthermore, the assay principle can readily be transferred to non-human species, thanks to the conservation of rRNA gene arrays across species . 4.3. Y-Chromosomal Multicopy Loci Among the protein-coding multicopy genes is Testis-Specific Protein Y-Encoded ( TSPY ), which is located in a tandem array on the short arm of the Y chromosome . TSPY has been shown to be present in 66 copies per genome ; however, variability in copy number between human individuals has been reported , with low copy numbers correlating with male infertility . Due to its Y-chromosomal localization and high copy number, a 133 bp amplicon of the fourth exon of TSPY has been suggested for qPCR-based detection of the Y chromosome and as an alternative to amelogenin for forensic sex typing . It is not clear whether this is the same sequence as the 133 bp TSPY amplicon targeted by the Plexor HY kit . The non-commercial NuMY assay analyzes a multicopy sequence on the Y chromosome that has been termed YRS but has not been further characterized . A genome search revealed 44 amplicons on the Y chromosome with the predicted lengths of 117 bp to which primers and probes were perfectly matching . The Innoquant HY kit does not disclose the two Y chromosomal multicopy target loci that are analyzed . From both loci, the amplicons have the same size, so the sequence is probably derived from a duplicated region on the Y chromosome. The PowerQuant kit targets two different multicopy loci on the Y chromosome that, however, have not been disclosed . Probably due to the small size difference of the amplicons, they are not suited for specifically analyzing the degradation of the Y chromosome. By contrast, the Investigator Quantiplex Pro kit targets two Y-specific amplicons with considerable length differences (see ) and thus allows for specifically quantitating the degradation of the Y chromosome . The two amplicons reside on the same multicopy locus, which, however, has not been disclosed. The quantitation of the degradation of the Y chromosome may help in predicting Y-STR typing success . 4.4. Non-Disclosed Target Loci As mentioned in , for two commercial qPCR assays the Y-chromosomal target loci have not been disclosed. This conforms to a general tendency since for the latest forensic multiplex qPCR kits on the market, none of the target loci are disclosed, and only the amplicon lengths are provided. In the respective validation studies and in the technical information provided by the manufacturers, only superficial information is given on the copy numbers and chromosomal localizations (see ). The manual of the Investigator Quantiplex kit names a proprietary sequence (4NS1C) and mentions its copy number and its localization on several autosomes . For both PowerQuant and Quantifiler Trio kits, localization on multiple autosomes and high copy numbers are mentioned in the validation studies, and a generic reference is given for the target loci that, however, does not specify the sequences, copy numbers, or the chromosomal localizations . Validation and evaluation studies and subsequent successful applications both in experiments and in case work may suffice to trust the qPCR kits in forensic settings and also to reproduce findings obtained using the kits as such. However, the kits then rather resemble black boxes that work “somehow”, and the analyst cannot explain why particular results (e.g., misleading or unexpected results) might have been obtained. For example, it is not clear why the kits display different sensitivities, different levels of accuracy, and different DI values when analyzing different source materials (see ). Moreover, no novel ideas on improvements or on applications in unrelated contexts are stimulated. For example, sequence information might be of interest for considering the usage of the loci in evolutionary studies or for combining the assays with additional amplification targets. As it is feasible to find out the DNA loci by sequencing the amplified products and performing database searches, it is incomprehensible that the identity and nature of the loci have not been revealed by the manufacturers. It appears surprising that initial publications on these kits have been accepted in peer-reviewed journals and that the lack of locus information escaped the reviewers’ attention. In the interest of scientific transparency it is to be hoped that manufacturers will find their way back to scientific standards and will eventually provide this missing information. Ideal candidates for qPCR-suited multicopy genes are retrotransposons, some of which are dispersed in high copy numbers throughout the genome and emerged only during primate or human evolution and thus are primate- or even human-specific . Retrotransposons are a subclass of transposons, genetic elements that are integrated in the genome and can, by enzymatically catalyzed mechanisms, “jump” to other chromosomal locations. Retrotransposons do so in a copy-paste fashion by being transcribed into an RNA that integrates by reverse transcription-based mechanisms at other places in the genome (reviewed in ). They can be classified as autonomous retrotransposons (LINEs, long interspersed elements) and non-autonomous retrotransposons to which short interspersed elements (SINEs) and SINE-VNTR-Alu elements (SVAs) belong. Various mechanisms evolved in organisms that counteract transposition. Yet, over time, by means of the copy-paste jumping mechanism, some retrotransposons have populated large parts of the genomes, thus contributing to evolution but in individuals also adversely affecting gene function and causing disease . L1 is the major LINE in primates. With a length of 6 kb and over 500,000 copies, L1 sequences make up roughly 17% of the human genome . LINEs encode the enzymatic machinery (e.g., reverse transcriptase) required for transposition. Non-autonomous retrotransposons do not encode these enzymes and jump with the help of the enzymes expressed by LINEs. The predominant SINEs in primates are the Alu elements (reviewed in ). Alu elements are non-coding and have sizes of about 300 bp and terminal poly(dA) tails with varying lengths that lead to length variations of the Alu elements . With about one million copies they make up roughly 10% of the human genome. Alu elements were acquired late during primate evolution and are thus specific for higher non-human primates and humans. Alu elements were among the first multicopy loci used in highly sensitive, human-specific forensic qPCR assays . There are three major Alu subfamilies that evolved in primates ( Alu J, Alu S, and Alu Y—evolutionary age in this order) and can be further subdivided into lineages. Of these, the Alu Yb lineage is the second largest young Alu Y group and has 1733 copies in the human genome that are distributed across all chromosomes . The AluYb8 is used in the commercial Innoquant HY kit, which also uses an SVA as the target locus (see and ). The SVAs consist of Alu -like sequences, followed by a variable number tandem repeat (VNTR) and a SINE-R-like sequence; they are the evolutionarily youngest group of retrotransposons and have only a few thousand copies per human genome, each with a size of about 2 kb . Differences in VNTR repeat numbers are responsible for inter-individual length variations. SVAs evolved in primates and are composed of six subfamilies, two of which are even human-specific . Multicopy loci other than retrotransposons have been established for forensic qPCR assays as well. Candidates for genetic loci that are present in high copy numbers are genes that encode non-translated RNAs, some of which are in high cellular demand to sustain cellular functioning. For the majority of these RNA genes the functional significance of their high copy numbers is not clear . One class of candidates for multicopy loci are genes for small nuclear RNAs (snRNAs), which are components of the spliceosomes involved in pre-mRNA splicing . The RNU2 locus is used in the commercial Plexor HY kit (see and ) and in the non-commercial NuMY assay . It encodes the U2 small nuclear RNA (snRNA). The RNU2 locus has been classified as a macrosatellite, i.e., a cluster of tandemly arranged repeat units that are longer than those of VNTR loci. Similar to VNTRs, macrosatellites display variability in repeat numbers within the population . The RNU2 gene cluster is located on chromosome 17 and consists of 20–40 RNU2 repeats, each with a length of 6.1 kb . Thus, RNU2 as a target for DNA quantification has the disadvantage of having far fewer copies than retrotransposons and of inter-individual variations in copy number. Furthermore, the localization at one chromosomal site increases the risk of stochastic sampling effects when analyzing low template DNA (see a), and it is not clear how far the degradation of the RNU2 locus reflects degradation at other chromosomal sites. Other autosomal non-coding multicopy target loci that have been suggested for forensic DNA quantitation are the 45S ribosomal RNA gene units that are clustered in tandem arrays in about 400 copies on five pairs of autosomes . Each 45S rRNA gene unit contains an array of 5.8S, 18S, and 28S rRNA genes that are separated by spacer sequences and are preceded by a common regulatory sequence. The units are separated by intergenic spacer sequences. While the functionally important rRNA genes are highly conserved across species, their regulatory and intergenic parts are less well conserved and have been used for designing a human-specific qPCR assay called RiboD assay. This assay targets two non-overlapping amplicons of 67 bp and 362 bp in the regulatory region to measure DNA amount and DNA integrity . So far, however, no multiplex assays have been established, and intraindividual and interindividual variations in the sequence and copy number of rRNA genes have been reported that might influence quantitation. An advantage of rRNA genes in forensic qPCR would be that their clusters reside on several chromosomes, thus minimizing stochastic sampling effects. Furthermore, the assay principle can readily be transferred to non-human species, thanks to the conservation of rRNA gene arrays across species . Among the protein-coding multicopy genes is Testis-Specific Protein Y-Encoded ( TSPY ), which is located in a tandem array on the short arm of the Y chromosome . TSPY has been shown to be present in 66 copies per genome ; however, variability in copy number between human individuals has been reported , with low copy numbers correlating with male infertility . Due to its Y-chromosomal localization and high copy number, a 133 bp amplicon of the fourth exon of TSPY has been suggested for qPCR-based detection of the Y chromosome and as an alternative to amelogenin for forensic sex typing . It is not clear whether this is the same sequence as the 133 bp TSPY amplicon targeted by the Plexor HY kit . The non-commercial NuMY assay analyzes a multicopy sequence on the Y chromosome that has been termed YRS but has not been further characterized . A genome search revealed 44 amplicons on the Y chromosome with the predicted lengths of 117 bp to which primers and probes were perfectly matching . The Innoquant HY kit does not disclose the two Y chromosomal multicopy target loci that are analyzed . From both loci, the amplicons have the same size, so the sequence is probably derived from a duplicated region on the Y chromosome. The PowerQuant kit targets two different multicopy loci on the Y chromosome that, however, have not been disclosed . Probably due to the small size difference of the amplicons, they are not suited for specifically analyzing the degradation of the Y chromosome. By contrast, the Investigator Quantiplex Pro kit targets two Y-specific amplicons with considerable length differences (see ) and thus allows for specifically quantitating the degradation of the Y chromosome . The two amplicons reside on the same multicopy locus, which, however, has not been disclosed. The quantitation of the degradation of the Y chromosome may help in predicting Y-STR typing success . As mentioned in , for two commercial qPCR assays the Y-chromosomal target loci have not been disclosed. This conforms to a general tendency since for the latest forensic multiplex qPCR kits on the market, none of the target loci are disclosed, and only the amplicon lengths are provided. In the respective validation studies and in the technical information provided by the manufacturers, only superficial information is given on the copy numbers and chromosomal localizations (see ). The manual of the Investigator Quantiplex kit names a proprietary sequence (4NS1C) and mentions its copy number and its localization on several autosomes . For both PowerQuant and Quantifiler Trio kits, localization on multiple autosomes and high copy numbers are mentioned in the validation studies, and a generic reference is given for the target loci that, however, does not specify the sequences, copy numbers, or the chromosomal localizations . Validation and evaluation studies and subsequent successful applications both in experiments and in case work may suffice to trust the qPCR kits in forensic settings and also to reproduce findings obtained using the kits as such. However, the kits then rather resemble black boxes that work “somehow”, and the analyst cannot explain why particular results (e.g., misleading or unexpected results) might have been obtained. For example, it is not clear why the kits display different sensitivities, different levels of accuracy, and different DI values when analyzing different source materials (see ). Moreover, no novel ideas on improvements or on applications in unrelated contexts are stimulated. For example, sequence information might be of interest for considering the usage of the loci in evolutionary studies or for combining the assays with additional amplification targets. As it is feasible to find out the DNA loci by sequencing the amplified products and performing database searches, it is incomprehensible that the identity and nature of the loci have not been revealed by the manufacturers. It appears surprising that initial publications on these kits have been accepted in peer-reviewed journals and that the lack of locus information escaped the reviewers’ attention. In the interest of scientific transparency it is to be hoped that manufacturers will find their way back to scientific standards and will eventually provide this missing information. From a theoretical point of view, the chromosomal distributions and copy numbers of target loci would be expected to impact the sensitivity and accuracy of qPCR assays, their performance in quantitating DNA degradation and the male component, and their sensitivity to PCR inhibitors. (The detection of PCR inhibitors is not related to the multicopy targets and has been briefly discussed above; see .) As mentioned, for the latest commercial qPCR assays, the locus information has so far not been disclosed, so any comparisons of their performance must remain descriptive and can at best consider the different lengths of the amplicons used for the quantitation of DNA degradation. A comparison of the four commercial kits targeting multicopy loci residing on several chromosomes (see ) revealed comparable sensitivities for all of them, with some differences in accuracies at very low DNA concentrations and in precision , which might be attributable to stochastic effects due to chromosomal localizations or copy numbers. As summarized in , the respective developmental validation studies of these kits described their ability to reproducibly quantitate DNA below 1 pg/µL , whereas for the Plexor HY kit, DNA concentrations of 1.9 pg/µL were required . The lower sensitivity of the Plexor HY might be related to the clustering of its multicopy target gene at a single chromosomal location, making it more prone to stochastic effects (see a). Because the developmental validation studies determined the sensitivities in slightly different ways, the exact values given are hardly comparable. For example, for the Innoquant HY kit the ability to detect 0.0781 pg/µL DNA for the autosomal targets has been reported , which might be attributable to the high copy number of the target loci (see ). However, the sensitivity was not determined according to the Minimum Information for Publication of Quantitative Real-Time PCR Experiments (MIQE) guidelines, which suggest stating the sensitivity as the limit of detection (LOD), i.e., the lowest DNA concentration detectable with 95% certainty . For the Plexor HY, PowerQuant, and Quantifiler Trio kits, the LODs were determined as the lowest concentration yielding quantitation results in all replicates of serially diluted genomic DNA, thus following MIQE guidelines. However, different numbers of replicates and different DNA concentrations were analyzed. For the Investigator Quantiplex Pro kit, it is unclear whether for the reported 0.015625 pg/μL DNA all replicates were detected. The four commercial qPCR kits indicate degradation by the DI values that are determined with the help of their respective autosomal and degradation targets. For the same degraded DNA sample, the obtained DI values differ between the kits, and consequently, the kits use different DI values to flag samples for moderate or severe DNA degradation (see ). When analyzing degraded DNA, the assay with the longest degradation target (Investigator Quantiplex Pro) showed the highest DI values, whereas the two assays with the smallest degradation targets (Innoquant HY and Quantifiler Trio) showed the smallest DI values . Likewise, another study analyzing degraded DNA from skeletal remains found the lowest DI values for Innoquant HY and Quantifiler Trio . A further study analyzing sheared DNA found higher DI values for the QuantiPlex kit than for the Quantifiler Trio kit . Since the quantitation targets of these kits have roughly similar sizes, these findings are consistent with longer amplicons being more sensitive to degradation. However, despite the almost similar sizes of the degradation and quantitation targets between the kits, a study analyzing DNA from UV-exposed DNA or fingerprints found the Innoquant HY kit to yield higher DI values than the Quantifiler Trio kit . Moreover, in a study by Morrison et al. (2020) comparing the Investigator Quantiplex Pro kit and the PowerQuant kit with sonicated genomic DNA, the former showed higher DI values at DNA concentrations of 250 pg/µL; however, the DI values at the DNA concentration of 25 pg/µL were higher with the latter kit . These two latter studies indicate that for low template DNA, other factors, such as differential sensitivities of the target loci or different copy numbers, might also influence the quantitation of degradation. Interestingly, in the study by Morrison et al. (2020), the analysis of the degradation of sonicated male genomic DNA with the Investigator Quantiplex Pro kit yielded higher DI values with the Y targets than with the autosomal targets . The almost same sizes of autosomal and Y-chromosomal degradation targets (see ) might suggest a higher sensitivity to DNA degradation of the Y-specific degradation target or of the Y chromosome in general. A recent study by Chierto et al. (2024) seems to point to the same direction. In this study, the Y-specific, longer amplicon of the PlexorHY kit was used as the degradation target to quantitate the autosomal DNA degradation of male DNA samples (with the shorter autosomal amplicon used as the quantitation target), and the results correlated well with the autosomal DNA degradation determined with the PowerQuant kit . Since the size difference of the two amplicons of the PlexorHY kit is much smaller than that of the two autosomal targets of the PowerQuant kit (see ), these findings seem to imply that the TSPY locus might be particularly sensitive to degradation and that Y-specific DNA degradation might serve as an indicator of autosomal DNA degradation in general. It should, however, be noted that the Y-chromosomal targets of the Investigator Quantiplex Pro kit and the PlexorHY kit are present in considerably lower copy numbers than the autosomal targets. Thus, the lower quantification results for the Y targets might simply result from their lower copy numbers because the number of intact copies possibly surviving degradation will be lower. When the ability to detect the male component in male–female DNA mixtures was compared between the PowerQuant and Investigator QuantiPlex Pro kits, both performed similarly . Another study comparing the detection of the male targets using DNA from skeletal remains showed comparable sensitivities for the Innoquant HY, Quantifiler Trio, and Investigator Quantiplex Pro kits, whereas the PowerQuant kit was less sensitive . As already noted in , the commercial kits differ in their sensitivities to PCR inhibitors . However, as the quantification targets of all kits have roughly the same sizes, their length cannot explain the different sensitivities to PCR inhibition. Taken together, the latest commercial qPCR kits are generally well suited for quantitating DNA amount and degradation even at low DNA concentrations, as well as for quantitating male components and detecting PCR inhibition. There are, however, subtle differences in performance, and the question remains as to how much these differences are relevant to STR typing success. Several studies have suggested that for optimal STR typing results, the best combination of a qPCR kit and a STR kit should be empirically determined . Forensic qPCR assays can help to predict the success of STR analysis and are thus of huge importance in the forensic DNA-analytical work flow. Since forensic qPCR assays specifically quantitate human DNA and use the same method, PCR, that will be used by the subsequent STR analysis, they specifically quantitate the DNA that is actually PCR-amplifiable and thus analyzable. In addition to the quantitation of DNA amounts, forensic qPCR assays provide information on DNA degradation and PCR inhibitors as well and help the analyst decide on the most appropriate analytical steps. The current forensic qPCR assays are adapted to the demands of current downstream STR analysis and are able to analyze DNA concentrations below the sensitivity limits of current STR kits. In keeping pace with further developments of forensic DNA analysis in terms of analytical questions, marker types, sensitivities, robustness, and amplicon lengths, the qPCR assays will likely be further improved in order to carry on giving the most useful information. To foster research on technical improvements and accelerate the emergence of novel ideas, it would be important to make information on the identity of target loci publicly available as per common scientific standards that have ensured scientific progress during the last decades. For several reasons, in the near future, digital PCR (dPCR) assays will likely be implemented in forensic DNA analysis. As an absolute quantitation method, unlike qPCR, dPCR is not requiring additional quantification standards (for review, see ). In dPCR, a defined fraction of the sample is evenly divided into a high number of partitions, such that each contains only a few or no template copies, and PCR-positive partitions are then determined by end-point measurement to calculate back the number of template copies of the original sample. For dPCR, it will be even more important to target multicopy loci that are present on all regions of all chromosomes because it is unpredictable which fraction of a genome (i.e., which chromosomal fragments) will end up in a particular partition. Thanks to end-point measurement, dPCR is only little influenced by variations in PCR efficiencies and is thus less prone to PCR inhibition, but for the same reason, dPCR is also less well suited to detect PCR inhibition . Furthermore, dPCR is able to quantitate minor components in mixtures with a higher sensitivity than qPCR . Since dPCR instruments are able to co-detect several fluorescence colors, and specific PCR products in the partitions can be detected based on fluorescence using TaqMan probes, the design of forensic multiplex dPCR assays is in principle possible . It will be interesting to compare the usefulness of future forensic dPCR assays with qPCR assays for predicting the success of forensic DNA typing. |
Palatal root endodontic microsurgery in maxillary molars using the palatal approach: a case study | ff9349f2-d6c9-41b6-9a07-331bb14144e8 | 11460143 | Dentistry[mh] | With the increased use of dental microscopy since the 1990s , endodontic microsurgery has become an effective treatment option for periapical lesions. Compared with traditional endodontic surgery, endodontic microsurgery offers the advantages of better identification of the root apex, less bone removal, and easier removal of pathogenic anatomical structures, thereby achieving better therapeutic effects . The buccal approach is commonly used in endodontic microsurgeries . However, it may not be suitable sometimes because of the potential for unnecessary removal of the buccal bone or root apex, distance between the palatal root tip and buccal bone plate, difficulty of manipulation of the soft tissue on the buccal side and possibility of perforation of the maxillary sinus mucosa, especially when the lesion is located surrounding the palatal root. The palatal approach may be a favorable option for palatal roots of maxillary molars. However, important anatomical structures such as the greater palatine artery (GPA), greater palatine foramen(GPF) and the complex maxillary sinus limit the surgical access . And any damage to these structures can result in compromised visibility during the procedure, postoperative complications, and even severe symptoms, such as syncope and excessive blood loss. Therefore, preoperative risk assessment is essential to avoid these complications. The use and analysis of Cone-beam computed tomography (CBCT) imaging is crucial , for providing important information including the thickness of the cortical bone on both sides of the root , the surrounding anatomical structures such as nerves, blood vessels, and sinus cavities, as well as the length and orientation of the root, which allows the surgeon to design the flap before surgery, determine the approach, the starting point and amount of bone removal, and even the angle of the root-end resection. Moreover, the resilient texture of the palatal mucoperiosteum makes flap elevation more laborious and increases the risk of inadvertent injury to surrounding vital anatomical structures. Owing to the obstruction of the sight line by the maxillary teeth, clinical techniques are required to achieve an optimal visual field during surgery. According to Lee et al., the use of a palatal approach in endodontic microsurgery for the palatal root of the maxillary first molar has a significantly high success rate and minimal occurrence of complications . Limited literature has been published regarding the comprehensive analysis and synthesis of endodontic microsurgeries performed on second molars through the palatal approach. From January to December 2022, nine cases of endodontic microsurgery via the palatal approach were reported at the Hangzhou Stomatological Hospital, China, comprising four maxillary first molars and five maxillary second molars. The clinical and radiographic findings at the 24-month follow-up revealed complete healing in all cases.
The characteristics of the nine patients are presented in Table . All patients had previously undergone root canal treatment or retreatment for symptomatic apical periodontitis, and the periapical lesions were mainly concentrated in the palatal root. This case report focuses on the right maxillary second molar of a 35-year-old man who presented with the main complaint of recurring occlusal pain for approximately 10 months. The pain was not alleviated after root canal treatment and retreatment, which had been performed 6 months previously. Clinical examination revealed an access cavity on the occlusal surface of the right maxillary second molar (tooth 17) filled with cement. Tooth 17 was tender on mild percussion, and no sinus tracts were found in the buccal or palatal mucosa (Fig. A). The periodontal probing depth was 2–3 mm circumferentially. Periapical radiography (eXpert DC, Gendex Dental Systems, Pennsylvania, USA) showed that tooth 17 had undergone root canal treatment and showed periapical radiolucency (Fig. B). Cone-beam computed tomography (NewTom Giano, CEFLASC S. C., Imola, Italy) revealed periapical radiolucency surrounding the apex of the palatal root and the corresponding mucosal hyperplasia of the maxillary sinus floor. The maxillary sinus floor is located between the buccal and palatal roots. The buccal cortical plate was intact 13–14 mm from the palatal root tip. The distance between the palatal root tip and the palatal plate was approximately 2 mm (Fig. C-E). The palatal foramen was situated distal to the second molar, 13 mm from the crest of the alveolar ridge (Fig. F). Based on the patient’s history and the clinical and radiographic findings, a diagnosis of previous root canal treatment for symptomatic apical periodontitis in tooth 17 was established. Because of the location of the periapical lesions and the maxillary sinus floor, endodontic microsurgery using the palatal approach was recommended and was selected by the patient. During palatal apical surgery of maxillary molars, the surgeon’s sight is hindered by the crown of the maxillary tooth. A line of sight was established on preoperative CBCT analysis to access the apical 3 mm region of the palatal root through the crown of tooth 21, with an angle closest to 90° relative to the long axis of the root, facilitating optimal visualization of the root section(Fig. G). The patient received 60 mg of loxoprofen sodium (NSAIDs) orally before surgery. The patient’s neck was supported with a neck pillow and the head was gently tilted to achieve a maxillary occlusion plane angle of approximately 100° from the ground. (Fig. A) The surgeon was in the 10 o’clock position relative to the patient while adjusting the microscope to provide a field of view from tooth 21 to the maxillary molars and palatal mucosa. A total of 3.4 mL of 4% articaine hydrochloride with 1:100,000 epinephrine was administered by infiltration. An additional 5 mL physiological saline was slowly injected under the palatal periosteum to separate it from the bone surface. A full-thickness flap was raised with a 10 mm vertical-release incision mesial to tooth 16 and an extended incision distal to tooth 17 (Fig. B). An intact cortical plate was observed (Fig. C). According to preoperative CBCT, a 3 mm apical region was cut using a helical high-speed (80000r/min) Lindemann bur (H162SXL.314.014, Brasseler, Germany) of 33 mm in length and 1.4 mm in diameter under a microscope (Zumax 2380, Suzhou, China) at 4x-8x magnification Commence 3 mm from the apex and progressively extend towards the apex to create a bone cavity with a diameter marginally exceeding 3 mm, thereby facilitating subsequent surgical procedures. And the palatal apex was positioned and exposed. The granulation tissue surrounding the palatal root was removed using a excavator (920–62 F, Devemed, Germany) and for sites close to the root furcation, a Gracey curette (121 − 020 #11/12, Hu-Friedy, USA) was selected, and a 3 mm root tip was resected nearly perpendicular to the long axis of the root using a Lindemann bur (H162SXL.314.014, Brasseler, Germany) at 8x magnification. (Fig. D). The hemostat used Racellet #3 pellets containing 0.55 mg epinephrine per pellet. An epinephrine cotton pellet was placed and slightly pressured at the bottom of the bone cavity, then cotton pellets was stuffed above and removed after 2 min. If bleeding was still present, the procedure was repeated. The resected root surface was stained with methylene blue and observed directly over the crown of tooth 21 at 14x-21x magnification. One canal with a gap between the gutta-percha and dentin was stained, and no fractures or cracks were observed (Fig. E). Root-end preparation using an ultrasonic tip (ASLD, Acteon, France) (Fig. F) was performed and Bioceramic iRoot BP Plus (Innovative BioCeramix, Vancouver, Canada) was used at 8x-14x magnification (Fig. G). The flap was closed with a 4 − 0 monofilament suture (Fig. H) after saline irrigation and postoperative radiography was performed (Fig. I). After operation, 0.2% chlorhexidine solution was given to the patient, and rinsed after 24 h, 1 min each time, twice a day for 1 week. The patient was asymptomatic 7 days after suture removal. At 3- and 24-month follow-ups, tooth 17 was asymptomatic, percussion-negative, and had no mobility. Periapical radiographs showed healing of the periapical lesions (Fig. A-B). CBCT imaging at the 24-month follow-up revealed complete healing after endodontic microsurgery and reformation of the periodontal space to a normal width (Fig. C-E). The submucosal hyperplasia of the maxillary sinus normalized. The patient was asyptomatic, and prick tests and swab tests were normal. According to Molven’s classification, it is categorized as type A, indicating complete healing .
According to Kim’s classification, the periapical lesions in these nine cases were all type B or C. The procedure was successfully performed and complete healing was achieved (type A or B, according to Molven’s classification) . Owing to the complex anatomical structures of the palate, a range of indications for the palatal approach can be considered. In cases where lesions primarily affect the palatal roots, it is possible to avoid damage to the GPA and GPF during surgery. Endodontic microsurgery via a palatal approach represents an effective and preferred treatment option for refractory periapical periodontitis of the maxillary molars following unsuccessful root canal therapy or re-therapy. Preoperative CBCT must be performed in conjunction with a clinical examination to evaluate the complexity of the procedure. The evaluation should encompass the following aspects. Length of palatal root The length of the palatal root should be measured using CBCT. The mean distances between the GPA and the enamelo-cemental junction of the maxillary first molar and second molar were 13 ± 1.4 mm and 13.9 ± 1 mm, respectively (mean ± standard deviation) . Therefore, if the length of the palatal root exceeds 15 mm, there is a high risk of damage to the GPA and GPF. The anatomical positioning of the GPA and GPF The GPF is typically situated on the palatal aspect of the third molar, or between the second and third molars . The descending palatine artery exits through this foramen along the pterygopalatine canal and transforms into the GPA . It is commonly found at the junction between the horizontal plate of the maxilla and the alveolar ridge . Coronal CBCT revealed a semicircular low-density shadow at this junction, which gradually transitioned into a ditch-like low-density shadow towards its distal end, representing the GPF (Fig. F). Intraoperative palpation can aid in determining tumor location. A “safe zone” is defined as being within 8.7 mm from the enamelo-cemental junction of the first molar and 10.9 mm from the second molar . Limiting the incision and bone removal procedures within this designated safe zone is recommended. Condition of oral aperture Patients should have a minimum mouth opening capacity of at least two fingers and be able to tolerate a duration of at least 1 h. A mouth opener can be used as an aid, if necessary. Perspective of observation In endodontic microsurgery, the direction of root-end resection should be as perpendicular as possible to the long axis of the roots to eliminate most of the complex anatomical structures. Direct observation of the surface of the root section is necessary to diagnose the causes of root canal treatment failure . However, palatal approaches are challenging because of obstructions from hard tissues such as the maxillary teeth (Fig. A-B). On preoperative CBCT images, it is possible to simulate the line of sight reaching the 3 mm area of the root apex from above different maxillary crowns. To obtain a direct view of the root section, the angle between this line and the long axis of the root should reach or exceed 90° (Fig. G). The optimal observation direction for the palatal root of the maxillary first molars may be between the canine and first premolar regions. However, for the maxillary second molars, observation between the first and second premolars may be the most suitable. Some skills and tips have been obtained for the palatal approach to endodontic microsurgery. Surgical Positioning For left maxillary molars, the operator is recommended to be in the 9 o’clock position relative to the patient. The 10 o’clock position is preferred for the right maxillary molars. The patient should be placed in a supine position with a neck pillow to ensure that the maxillary plane forms an angle of approximately 100° from the ground while rotating the head by 30°-45° towards one side. Additionally, taking advantage of the high arch in the palatal dome can significantly enhance visibility. Incision It is recommended that the vertical incision be equal in length to the palatal root. To avoid inadequate surgical exposure, the vertical incision should be mesial to the first premolar for the first molar and mesial to the maxillary first molar for the maxillary second molar. The crevicular incision should extend towards the distal part of the affected tooth, and an additional 3 mm extended incision distal to the affected tooth can help reduce the flap tension. Palate flap The palatal mucoperiosteum and bone tissue are intimately connected by Sharpey’s fibers, and the texture of the palatal mucosa is resilient . To address this issue, a subperiosteal injection of 5 ml normal saline along with an additional distal incision may be selected. Bone removal To prevent damage to the GPF and to optimize intraoperative visualization, the bone removal position should be located within 3 mm of the coronal direction of the palatal root apex. Gradual enlargement of the removal area is performed along its periphery until the root tip is exposed and a bone cavity with a diameter of 3–4 mm is identified. Root-end resection In some cases, the root section perpendicular to the long axis of the root could not be directly observed because of hard tissue obstruction. A “two-step root-end resection” method has been suggested. First, a significant bevel angle such as 45° was used, followed by staining, root-end preparation, and filling (Fig. C-D). Next, the root was resected at a 90° vertical surface (Fig. E). The palatal approach in cases of periapical lesions of maxillary molar palatal roots should be advocated. The tips of palatal endodontic microsurgery and the relationship between observation and bevel angle are rarely mentioned in previous literature. And above methods have been proposed to solve the difficulties of palatal endodontic microsurgery. Besides, The integration of digital guide plates and dynamic navigation technology with the palatal approach is poised to further enhance its efficacy.
The length of the palatal root should be measured using CBCT. The mean distances between the GPA and the enamelo-cemental junction of the maxillary first molar and second molar were 13 ± 1.4 mm and 13.9 ± 1 mm, respectively (mean ± standard deviation) . Therefore, if the length of the palatal root exceeds 15 mm, there is a high risk of damage to the GPA and GPF.
The GPF is typically situated on the palatal aspect of the third molar, or between the second and third molars . The descending palatine artery exits through this foramen along the pterygopalatine canal and transforms into the GPA . It is commonly found at the junction between the horizontal plate of the maxilla and the alveolar ridge . Coronal CBCT revealed a semicircular low-density shadow at this junction, which gradually transitioned into a ditch-like low-density shadow towards its distal end, representing the GPF (Fig. F). Intraoperative palpation can aid in determining tumor location. A “safe zone” is defined as being within 8.7 mm from the enamelo-cemental junction of the first molar and 10.9 mm from the second molar . Limiting the incision and bone removal procedures within this designated safe zone is recommended.
Patients should have a minimum mouth opening capacity of at least two fingers and be able to tolerate a duration of at least 1 h. A mouth opener can be used as an aid, if necessary.
In endodontic microsurgery, the direction of root-end resection should be as perpendicular as possible to the long axis of the roots to eliminate most of the complex anatomical structures. Direct observation of the surface of the root section is necessary to diagnose the causes of root canal treatment failure . However, palatal approaches are challenging because of obstructions from hard tissues such as the maxillary teeth (Fig. A-B). On preoperative CBCT images, it is possible to simulate the line of sight reaching the 3 mm area of the root apex from above different maxillary crowns. To obtain a direct view of the root section, the angle between this line and the long axis of the root should reach or exceed 90° (Fig. G). The optimal observation direction for the palatal root of the maxillary first molars may be between the canine and first premolar regions. However, for the maxillary second molars, observation between the first and second premolars may be the most suitable. Some skills and tips have been obtained for the palatal approach to endodontic microsurgery.
For left maxillary molars, the operator is recommended to be in the 9 o’clock position relative to the patient. The 10 o’clock position is preferred for the right maxillary molars. The patient should be placed in a supine position with a neck pillow to ensure that the maxillary plane forms an angle of approximately 100° from the ground while rotating the head by 30°-45° towards one side. Additionally, taking advantage of the high arch in the palatal dome can significantly enhance visibility.
It is recommended that the vertical incision be equal in length to the palatal root. To avoid inadequate surgical exposure, the vertical incision should be mesial to the first premolar for the first molar and mesial to the maxillary first molar for the maxillary second molar. The crevicular incision should extend towards the distal part of the affected tooth, and an additional 3 mm extended incision distal to the affected tooth can help reduce the flap tension.
The palatal mucoperiosteum and bone tissue are intimately connected by Sharpey’s fibers, and the texture of the palatal mucosa is resilient . To address this issue, a subperiosteal injection of 5 ml normal saline along with an additional distal incision may be selected.
To prevent damage to the GPF and to optimize intraoperative visualization, the bone removal position should be located within 3 mm of the coronal direction of the palatal root apex. Gradual enlargement of the removal area is performed along its periphery until the root tip is exposed and a bone cavity with a diameter of 3–4 mm is identified.
In some cases, the root section perpendicular to the long axis of the root could not be directly observed because of hard tissue obstruction. A “two-step root-end resection” method has been suggested. First, a significant bevel angle such as 45° was used, followed by staining, root-end preparation, and filling (Fig. C-D). Next, the root was resected at a 90° vertical surface (Fig. E). The palatal approach in cases of periapical lesions of maxillary molar palatal roots should be advocated. The tips of palatal endodontic microsurgery and the relationship between observation and bevel angle are rarely mentioned in previous literature. And above methods have been proposed to solve the difficulties of palatal endodontic microsurgery. Besides, The integration of digital guide plates and dynamic navigation technology with the palatal approach is poised to further enhance its efficacy.
|
Enhancing Medical Student Engagement Through Cinematic Clinical Narratives: Multimodal Generative AI–Based Mixed Methods Study | 1da83f3a-729f-4481-aaa8-48c48dc300d3 | 11751740 | Pharmacology[mh] | Background Student and trainee engagement is a critical factor in medical education, influencing outcomes such as academic achievement, overall well-being, satisfaction, and reduced burnout . High levels of engagement have been linked to increased motivation and better learning experiences, as active participation encourages deeper understanding and application of complex material . In contrast, traditional lecture-based learning often results in passive absorption of information, limiting student engagement and negatively affecting the ability to interact meaningfully with content . To address this, we developed a cinematic clinical narrative (CCN), an interactive multimedia learning experience designed to enhance student engagement by integrating cinematic storytelling and narrative-based learning techniques. This method builds upon the principles of cinemeducation, a teaching approach that uses film to create emotional connections and foster active learning . By using generative artificial intelligence (genAI) tools, we have further enhanced the learning experience and decreased the barrier to entry for instructors, making it more immersive and adaptable to current educational needs. GenAI has been recognized as a transformative tool in reshaping medical education, offering new opportunities for interactive, technology-driven learning environments that promote active student engagement . The target audience for our CCN comprises first-year medical students learning pharmacology related to the immune system. Medical students often face a knowledge gap in understanding complex pharmacological interactions and the intricacies of immune responses largely due to the difficulty of the material . Furthermore, there is speculated to be a skill gap in medical and other professional health science students in applying theoretical knowledge to clinical scenarios and the real problem of burnout due to many factors, one of which is the large amount of knowledge required to retain in a short amount of time . The CCN aims to address these issues by enhancing comprehension, clinical application skills, and empathy toward patients with autoimmune diseases. The CCN used a unique instructional approach by merging cinemeducation with multiple genAI platforms, tailored for first-year medical students in pharmacology. This method addresses the challenge of enhancing engagement and knowledge retention in complex subjects such as immune system pharmacology. Unlike traditional didactic teaching, our approach, supported by others advocating for innovative teaching strategies, uses storytelling to deepen understanding and empathy . Use of genAI in medical training, particularly in personalizing learning experiences and competencies for genAI-based tools, is also a current area of active research . This aligns with other researchers who highlight the importance of interactive and engaging content in medical education . Our project also leverages the effectiveness of narrative-based learning, which offers an experiential learning environment over conventional teaching methods and is more accurate to real-world situations . Medical students often struggle to engage with and retain complex pharmacological concepts, especially in preclinical education, where traditional teaching methods can lead to passive learning and poor knowledge retention. To address this challenge, we developed and implemented a novel instructional approach, CCNs, which leverages multimodal genAI tools to create immersive, engaging learning experiences. The aim of this study is to evaluate the effectiveness of these genAI-enhanced CCNs in increasing student engagement, interest, and knowledge retention in medical pharmacology concepts. We tested this intervention by assessing student interest using the Situational Interest Survey for Multimedia (SIS-M) and measuring examination performance on content covered by the CCNs. We hypothesize that students exposed to CCNs will report higher levels of engagement compared with traditional case-based learning and have passing examination grades on questions related to the CCN. Theoretical Framework The instructional method in the CCN uses contemporary educational theories emphasizing active, learner-centered approaches. Drawing inspiration from the Constructivist Learning Theory, which advocates for knowledge construction through experience , our approach uses an adaptation of cinemeducation to create an immersive learning environment . This also aligns with Mayer’s Cognitive Theory of Multimedia Learning, which suggests that learning is enhanced through multimodal presentations . Furthermore, our multimodal use of various genAI platforms for content development is informed by the Technological Pedagogical Content Knowledge (TPACK) framework , ensuring an effective integration of technology in teaching. This methodology responds to identified needs in medical education for more engaging and effective teaching strategies, bridging theory and practice in a novel and impactful way. Student and trainee engagement is a critical factor in medical education, influencing outcomes such as academic achievement, overall well-being, satisfaction, and reduced burnout . High levels of engagement have been linked to increased motivation and better learning experiences, as active participation encourages deeper understanding and application of complex material . In contrast, traditional lecture-based learning often results in passive absorption of information, limiting student engagement and negatively affecting the ability to interact meaningfully with content . To address this, we developed a cinematic clinical narrative (CCN), an interactive multimedia learning experience designed to enhance student engagement by integrating cinematic storytelling and narrative-based learning techniques. This method builds upon the principles of cinemeducation, a teaching approach that uses film to create emotional connections and foster active learning . By using generative artificial intelligence (genAI) tools, we have further enhanced the learning experience and decreased the barrier to entry for instructors, making it more immersive and adaptable to current educational needs. GenAI has been recognized as a transformative tool in reshaping medical education, offering new opportunities for interactive, technology-driven learning environments that promote active student engagement . The target audience for our CCN comprises first-year medical students learning pharmacology related to the immune system. Medical students often face a knowledge gap in understanding complex pharmacological interactions and the intricacies of immune responses largely due to the difficulty of the material . Furthermore, there is speculated to be a skill gap in medical and other professional health science students in applying theoretical knowledge to clinical scenarios and the real problem of burnout due to many factors, one of which is the large amount of knowledge required to retain in a short amount of time . The CCN aims to address these issues by enhancing comprehension, clinical application skills, and empathy toward patients with autoimmune diseases. The CCN used a unique instructional approach by merging cinemeducation with multiple genAI platforms, tailored for first-year medical students in pharmacology. This method addresses the challenge of enhancing engagement and knowledge retention in complex subjects such as immune system pharmacology. Unlike traditional didactic teaching, our approach, supported by others advocating for innovative teaching strategies, uses storytelling to deepen understanding and empathy . Use of genAI in medical training, particularly in personalizing learning experiences and competencies for genAI-based tools, is also a current area of active research . This aligns with other researchers who highlight the importance of interactive and engaging content in medical education . Our project also leverages the effectiveness of narrative-based learning, which offers an experiential learning environment over conventional teaching methods and is more accurate to real-world situations . Medical students often struggle to engage with and retain complex pharmacological concepts, especially in preclinical education, where traditional teaching methods can lead to passive learning and poor knowledge retention. To address this challenge, we developed and implemented a novel instructional approach, CCNs, which leverages multimodal genAI tools to create immersive, engaging learning experiences. The aim of this study is to evaluate the effectiveness of these genAI-enhanced CCNs in increasing student engagement, interest, and knowledge retention in medical pharmacology concepts. We tested this intervention by assessing student interest using the Situational Interest Survey for Multimedia (SIS-M) and measuring examination performance on content covered by the CCNs. We hypothesize that students exposed to CCNs will report higher levels of engagement compared with traditional case-based learning and have passing examination grades on questions related to the CCN. The instructional method in the CCN uses contemporary educational theories emphasizing active, learner-centered approaches. Drawing inspiration from the Constructivist Learning Theory, which advocates for knowledge construction through experience , our approach uses an adaptation of cinemeducation to create an immersive learning environment . This also aligns with Mayer’s Cognitive Theory of Multimedia Learning, which suggests that learning is enhanced through multimodal presentations . Furthermore, our multimodal use of various genAI platforms for content development is informed by the Technological Pedagogical Content Knowledge (TPACK) framework , ensuring an effective integration of technology in teaching. This methodology responds to identified needs in medical education for more engaging and effective teaching strategies, bridging theory and practice in a novel and impactful way. Participants and CCN Design Overview This study was conducted at the University of Idaho WWAMI Medical Education Program, which is part of a collaborative University of Washington School of Medicine program serving Washington, Wyoming, Alaska, Montana, and Idaho. The WWAMI program provides medical education to students across these states, offering them the opportunity to complete their first 2 preclinical years of medical school in their home states before transitioning to clinical training. The target learners for this study were first-year medical students in the WWAMI program enrolled in a 6-week foundational infections and immunity course, which included topics covering immune system pharmacology. Students in this course attend pharmacology lectures that culminate in clinical cases, allowing them to apply their newly acquired knowledge of medications to real-world patient scenarios. We decided to reimagine one of these cases into “Shattered Slippers,” a CCN that was presented as a fictional sequel to the movie “Another Cinderella Story” . This fictional sequel features the star from the original movie, Selena Gomez, which was purposeful, given her real-life battle with lupus and her experience receiving a kidney transplant. This choice not only provides a strong thematic link connecting the CCN to the source material but also serves to humanize and demystify the conditions under study. The development of “Shattered Slippers” used a suite of genAI platforms to create an immersive and engaging learning experience . The plot was crafted using GPT-4, known for its language understanding and generation capabilities. For visual imagery, Leonardo.ai and Stable Diffusion were used to generate high-quality, contextually relevant images. Narration was produced using Eleven Labs, ensuring a coherent and captivating storytelling experience. Furthermore, the theme song, integral to setting the tone of the educational module, was composed using the combined efforts of GPT-4 and Suno. These artificial intelligence (AI)–generated materials were all integrated into 2 PowerPoint presentations. Part I of the CCN was presented at the end of a 1-hour pharmacology lecture on immunomodulatory drugs with specific focus on nonsteroidal anti-inflammatory drugs, glucocorticoids, and innate immune system inhibitors. Part II of the CCN was presented 4 weeks later at the end of a 1-hour pharmacology lecture on immunomodulatory and transplant drugs with specific focus on cytokine inhibitors, cytotoxic drugs, and antimetabolites. Both lectures were presented in-person with >90% of students attending both lectures. The combined CCN is provided as a supplemental file . At the conclusion of the course, students were informed about Selena Gomez’s actual medical journey. This revelation effectively bridged the gap between the fictional narrative of “Shattered Slippers” and real-world medical scenarios, thereby enhancing the educational impact and relevance of the clinical cases discussed. Plot Development The process of developing the plot for “Shattered Slippers” began with a reimagining of a clinical case initially presented in the first-year medical school curriculum. This original case centered around a ballerina struggling with rheumatoid arthritis, where students were tasked with diagnosing the sources of her pain and inflammation and selecting suitable immunomodulatory medications. Using ChatGPT (GPT-4) , a large language model (LLM), we transformed this clinical scenario into a compelling narrative for “Shattered Slippers.” The sequential steps of the medical case were input into GPT-4, with instructions to adapt these into a fictional storyline ( and ). To enhance thematic resonance and real-world connection, the ballerina’s diagnosis in the plot was altered from rheumatoid arthritis to lupus, mirroring the real-life medical condition of Selena Gomez, who stars in the CCN. Further expanding the scope of the narrative, the plot incorporated a kidney transplant storyline. This addition served a dual purpose. First, it aligned with the second lecture on immunoregulatory pharmacology focusing on organ transplant pharmacology. Second, it resonated with Selena Gomez’s personal medical history, as she has undergone a kidney transplant. This incorporation not only ensured continuity with the educational objectives of the course but also added depth and authenticity to the fictional narrative, making it more engaging and relatable for the students. Image Generation In order to create a more immersive educational experience, fictional images were integrated into the “Shattered Slippers” case study. These images were generated using the Leonardo.ai platform , which harnesses the capabilities of the Stable Diffusion XL image–generating technology ( and ). In an effort to maintain transparency and distinguish between real and AI-generated content, all images depicting real people were marked with an “AI-generated image” icon. This icon, chosen for its symbolic significance, is the spinning top from the movie “Inception.” The selection of this particular icon was purposeful; it serves as a metaphor for the increasingly blurred lines between reality and artificial constructs, mirroring the movie’s thematic exploration of distinguishing reality from illusion. This concept was explained to the students prior to their engagement with the case, setting the stage for a thoughtful consideration of the role and impact of genAI in content creation. This iconography not only helped in identifying AI-generated images but also subtly underscored the advanced capabilities of genAI in creating hyperrealistic images. Narration Generation Enhancing the immersive aspect of the CCN, an audio narration was incorporated to accompany the text on the PowerPoint slides. This element was designed to emulate the experience of listening to a movie narrator, thereby bringing the story of “Shattered Slippers” to life in an auditory format. To achieve this, the finalized script of the plot was submitted to the Eleven Labs platform , which specializes in converting text into lifelike audio narration . Each of these audio narrations were incorporated into their corresponding PowerPoint slides. As each slide was presented during the course, the audio narration played automatically, further synchronizing the visual and auditory elements of the learning experience. This integration of audio narration with the visual content not only enriched the storytelling aspect of the module but also supported diverse learning styles, facilitating a more engaging and multisensory educational experience for the students. Theme Song Generation Although not directly educational, a theme song for “Shattered Slippers” was created to complete the immersive experience. The inclusion of a theme song aimed to add an additional layer of engagement and context to the fictional movie, contributing to a more comprehensive and cinematic learning environment. The lyrics for the theme song were generated using GPT-4 . Following the lyric generation, Suno Chirp Bot, a genAI tool for music composition , was used to create the melody and vocals for the theme song. This genAI-driven process allowed for a harmonious blend of lyrics and music, resulting in a fully rendered theme song . Once completed, the theme song was embedded into the PowerPoint presentation. This musical addition served as a capstone to the multisensory educational module, further enriching the student’s experience by providing a unique auditory element that complemented the visual and textual components of “Shattered Slippers.” Data Collection The “Shattered Slippers” CCN was integrated into 2 distinct pharmacology lectures, both of which focused on medications used in immune system modulation. The target audience for this CCN was a class of 40 first-year medical students (n=40). This approach aimed not only to enrich their understanding of immunomodulatory pharmacology but also to engage them in a unique and memorable learning experience. To evaluate student interest in the CCN as an educational tool, at the conclusion of the course, students were invited to participate in a feedback process using the SIS-M of which 18 students responded (n=18). The SIS-M was developed by Dr Tonia Dousay, a professor in instructional design and educational technology, to assess various constructs of situational interest in multimedia-based learning environments. Originally created for the educational field, the SIS-M focuses on adult learners and measures constructs such as triggered situational interest (initial engagement with multimedia), maintained interest, and value interest (perceived usefulness of the content). The survey was originally used to evaluate the effectiveness of multimedia in promoting engagement and motivation in higher education and adult learning settings and has recently been used in medical education research , making it an appropriate tool for assessing learner engagement in this study. This survey was used to capture their views and opinions on the “Shattered Slippers” case, providing insights into student engagement, interest, and the overall impact of the CCN on their learning experience. The survey includes items to rank on a 1‐5 scale (1=strongly disagree, 5=strongly agree), a question asking for preference of clinical case format, and an open-ended question asking, “Why do you think this is your preference.” The CHERRIES report for this survey is supplied . Data Analysis The research team used Microsoft Excel for the analysis of the SIS-M survey results. The average class pharmacology examination grades (n=40) from questions covered by the “Shattered Slippers” case study (n=2) were analyzed for achievement data. These included a multiple-choice question, selected by the course lead (not the study author) from a pool of questions that tested pharmacology content covered in each pharmacology lecture. The questions were administered during the students’ weekly examinations, scheduled for the week immediately following the presentation of the material. Importantly, these questions were modeled after USMLE-style step 1 board questions, which assess students’ ability to apply their pharmacological knowledge in a clinical context. Using this format provides a rigorous and standardized measure of student understanding of the material, ensuring that the assessment reflects the type of knowledge and critical thinking required for success on future board examinations. The SIS-M survey’s analysis focused on various dimensions of situational interest: triggered interest, maintained-value (MV), maintained interest, and maintained-feeling (MF). Thematic analysis was conducted using ChatGPT (GPT4o and o1-preview) and Claude 3.5 Sonnet. This involved generating initial codes and identifying themes, followed by the researcher combining and refining these themes for overlap and relevancy between the 3 LLMs . Prompt engineering techniques used included Persona Prompting , Zero-Shot Chain of Thought (CoT) , and Self-Criticism . The Zero-Shot Chain of Thought prompting was not used with the ChatGPT o1-preview model, as it has built-in Tree-of-Thought functionality in every output. The initial prompt was the following: Act like a brilliant medical education researcher. I am doing a study on aCinematic Clinical Narrative(CCN) which is an educational tool that combines clinical case studies with storytelling techniques typically seen in movies or TV shows. By embedding medical information within a compelling fictional storyline, CCNs help medical students retain complex medical concepts in an engaging, memorable way. The CCN in the study was called“Shattered Slippers,”was a fictional sequel to the movie“Another Cinderella Story,”and stars Selena Gomez. It covered the topics of immunomodulatory medications for treating lupus, and kidney transplants. I surveyed the participants on their preference of the CCN over traditional clinical cases and asked them to explain their preference. Please perform a thematic analysis on the below participant responses marked between <response> </response>. Let’s work this out in a step by step way to be sure we have the right answer. <response> Participant responses here </response> This was then followed by the following Self-Criticism prompt: “Please reflect on your previous answer for any errors.” Ethical Considerations This educational research was approved as exempt by the institutional review board of the University of Idaho (21-223). As the CCN incorporated references to real celebrities and included AI-generated images of actual people, we consulted legal counsel to ensure compliance. The counsel advised that, given the educational context and the clear labeling of images as AI-generated rather than real, the usage was permissible. Furthermore, we end the CCN with a brief description of the real-life health struggles of the celebrities, which is all public information. However, since this remains a legally gray area, we recommend exercising caution in future projects that use similar techniques. The SIS-M was conducted anonymously to ensure the confidentiality of participants’ responses. No identifying information was collected, allowing students to provide honest feedback without concern for personal attribution. This study was conducted at the University of Idaho WWAMI Medical Education Program, which is part of a collaborative University of Washington School of Medicine program serving Washington, Wyoming, Alaska, Montana, and Idaho. The WWAMI program provides medical education to students across these states, offering them the opportunity to complete their first 2 preclinical years of medical school in their home states before transitioning to clinical training. The target learners for this study were first-year medical students in the WWAMI program enrolled in a 6-week foundational infections and immunity course, which included topics covering immune system pharmacology. Students in this course attend pharmacology lectures that culminate in clinical cases, allowing them to apply their newly acquired knowledge of medications to real-world patient scenarios. We decided to reimagine one of these cases into “Shattered Slippers,” a CCN that was presented as a fictional sequel to the movie “Another Cinderella Story” . This fictional sequel features the star from the original movie, Selena Gomez, which was purposeful, given her real-life battle with lupus and her experience receiving a kidney transplant. This choice not only provides a strong thematic link connecting the CCN to the source material but also serves to humanize and demystify the conditions under study. The development of “Shattered Slippers” used a suite of genAI platforms to create an immersive and engaging learning experience . The plot was crafted using GPT-4, known for its language understanding and generation capabilities. For visual imagery, Leonardo.ai and Stable Diffusion were used to generate high-quality, contextually relevant images. Narration was produced using Eleven Labs, ensuring a coherent and captivating storytelling experience. Furthermore, the theme song, integral to setting the tone of the educational module, was composed using the combined efforts of GPT-4 and Suno. These artificial intelligence (AI)–generated materials were all integrated into 2 PowerPoint presentations. Part I of the CCN was presented at the end of a 1-hour pharmacology lecture on immunomodulatory drugs with specific focus on nonsteroidal anti-inflammatory drugs, glucocorticoids, and innate immune system inhibitors. Part II of the CCN was presented 4 weeks later at the end of a 1-hour pharmacology lecture on immunomodulatory and transplant drugs with specific focus on cytokine inhibitors, cytotoxic drugs, and antimetabolites. Both lectures were presented in-person with >90% of students attending both lectures. The combined CCN is provided as a supplemental file . At the conclusion of the course, students were informed about Selena Gomez’s actual medical journey. This revelation effectively bridged the gap between the fictional narrative of “Shattered Slippers” and real-world medical scenarios, thereby enhancing the educational impact and relevance of the clinical cases discussed. The process of developing the plot for “Shattered Slippers” began with a reimagining of a clinical case initially presented in the first-year medical school curriculum. This original case centered around a ballerina struggling with rheumatoid arthritis, where students were tasked with diagnosing the sources of her pain and inflammation and selecting suitable immunomodulatory medications. Using ChatGPT (GPT-4) , a large language model (LLM), we transformed this clinical scenario into a compelling narrative for “Shattered Slippers.” The sequential steps of the medical case were input into GPT-4, with instructions to adapt these into a fictional storyline ( and ). To enhance thematic resonance and real-world connection, the ballerina’s diagnosis in the plot was altered from rheumatoid arthritis to lupus, mirroring the real-life medical condition of Selena Gomez, who stars in the CCN. Further expanding the scope of the narrative, the plot incorporated a kidney transplant storyline. This addition served a dual purpose. First, it aligned with the second lecture on immunoregulatory pharmacology focusing on organ transplant pharmacology. Second, it resonated with Selena Gomez’s personal medical history, as she has undergone a kidney transplant. This incorporation not only ensured continuity with the educational objectives of the course but also added depth and authenticity to the fictional narrative, making it more engaging and relatable for the students. In order to create a more immersive educational experience, fictional images were integrated into the “Shattered Slippers” case study. These images were generated using the Leonardo.ai platform , which harnesses the capabilities of the Stable Diffusion XL image–generating technology ( and ). In an effort to maintain transparency and distinguish between real and AI-generated content, all images depicting real people were marked with an “AI-generated image” icon. This icon, chosen for its symbolic significance, is the spinning top from the movie “Inception.” The selection of this particular icon was purposeful; it serves as a metaphor for the increasingly blurred lines between reality and artificial constructs, mirroring the movie’s thematic exploration of distinguishing reality from illusion. This concept was explained to the students prior to their engagement with the case, setting the stage for a thoughtful consideration of the role and impact of genAI in content creation. This iconography not only helped in identifying AI-generated images but also subtly underscored the advanced capabilities of genAI in creating hyperrealistic images. Enhancing the immersive aspect of the CCN, an audio narration was incorporated to accompany the text on the PowerPoint slides. This element was designed to emulate the experience of listening to a movie narrator, thereby bringing the story of “Shattered Slippers” to life in an auditory format. To achieve this, the finalized script of the plot was submitted to the Eleven Labs platform , which specializes in converting text into lifelike audio narration . Each of these audio narrations were incorporated into their corresponding PowerPoint slides. As each slide was presented during the course, the audio narration played automatically, further synchronizing the visual and auditory elements of the learning experience. This integration of audio narration with the visual content not only enriched the storytelling aspect of the module but also supported diverse learning styles, facilitating a more engaging and multisensory educational experience for the students. Although not directly educational, a theme song for “Shattered Slippers” was created to complete the immersive experience. The inclusion of a theme song aimed to add an additional layer of engagement and context to the fictional movie, contributing to a more comprehensive and cinematic learning environment. The lyrics for the theme song were generated using GPT-4 . Following the lyric generation, Suno Chirp Bot, a genAI tool for music composition , was used to create the melody and vocals for the theme song. This genAI-driven process allowed for a harmonious blend of lyrics and music, resulting in a fully rendered theme song . Once completed, the theme song was embedded into the PowerPoint presentation. This musical addition served as a capstone to the multisensory educational module, further enriching the student’s experience by providing a unique auditory element that complemented the visual and textual components of “Shattered Slippers.” The “Shattered Slippers” CCN was integrated into 2 distinct pharmacology lectures, both of which focused on medications used in immune system modulation. The target audience for this CCN was a class of 40 first-year medical students (n=40). This approach aimed not only to enrich their understanding of immunomodulatory pharmacology but also to engage them in a unique and memorable learning experience. To evaluate student interest in the CCN as an educational tool, at the conclusion of the course, students were invited to participate in a feedback process using the SIS-M of which 18 students responded (n=18). The SIS-M was developed by Dr Tonia Dousay, a professor in instructional design and educational technology, to assess various constructs of situational interest in multimedia-based learning environments. Originally created for the educational field, the SIS-M focuses on adult learners and measures constructs such as triggered situational interest (initial engagement with multimedia), maintained interest, and value interest (perceived usefulness of the content). The survey was originally used to evaluate the effectiveness of multimedia in promoting engagement and motivation in higher education and adult learning settings and has recently been used in medical education research , making it an appropriate tool for assessing learner engagement in this study. This survey was used to capture their views and opinions on the “Shattered Slippers” case, providing insights into student engagement, interest, and the overall impact of the CCN on their learning experience. The survey includes items to rank on a 1‐5 scale (1=strongly disagree, 5=strongly agree), a question asking for preference of clinical case format, and an open-ended question asking, “Why do you think this is your preference.” The CHERRIES report for this survey is supplied . The research team used Microsoft Excel for the analysis of the SIS-M survey results. The average class pharmacology examination grades (n=40) from questions covered by the “Shattered Slippers” case study (n=2) were analyzed for achievement data. These included a multiple-choice question, selected by the course lead (not the study author) from a pool of questions that tested pharmacology content covered in each pharmacology lecture. The questions were administered during the students’ weekly examinations, scheduled for the week immediately following the presentation of the material. Importantly, these questions were modeled after USMLE-style step 1 board questions, which assess students’ ability to apply their pharmacological knowledge in a clinical context. Using this format provides a rigorous and standardized measure of student understanding of the material, ensuring that the assessment reflects the type of knowledge and critical thinking required for success on future board examinations. The SIS-M survey’s analysis focused on various dimensions of situational interest: triggered interest, maintained-value (MV), maintained interest, and maintained-feeling (MF). Thematic analysis was conducted using ChatGPT (GPT4o and o1-preview) and Claude 3.5 Sonnet. This involved generating initial codes and identifying themes, followed by the researcher combining and refining these themes for overlap and relevancy between the 3 LLMs . Prompt engineering techniques used included Persona Prompting , Zero-Shot Chain of Thought (CoT) , and Self-Criticism . The Zero-Shot Chain of Thought prompting was not used with the ChatGPT o1-preview model, as it has built-in Tree-of-Thought functionality in every output. The initial prompt was the following: Act like a brilliant medical education researcher. I am doing a study on aCinematic Clinical Narrative(CCN) which is an educational tool that combines clinical case studies with storytelling techniques typically seen in movies or TV shows. By embedding medical information within a compelling fictional storyline, CCNs help medical students retain complex medical concepts in an engaging, memorable way. The CCN in the study was called“Shattered Slippers,”was a fictional sequel to the movie“Another Cinderella Story,”and stars Selena Gomez. It covered the topics of immunomodulatory medications for treating lupus, and kidney transplants. I surveyed the participants on their preference of the CCN over traditional clinical cases and asked them to explain their preference. Please perform a thematic analysis on the below participant responses marked between <response> </response>. Let’s work this out in a step by step way to be sure we have the right answer. <response> Participant responses here </response> This was then followed by the following Self-Criticism prompt: “Please reflect on your previous answer for any errors.” This educational research was approved as exempt by the institutional review board of the University of Idaho (21-223). As the CCN incorporated references to real celebrities and included AI-generated images of actual people, we consulted legal counsel to ensure compliance. The counsel advised that, given the educational context and the clear labeling of images as AI-generated rather than real, the usage was permissible. Furthermore, we end the CCN with a brief description of the real-life health struggles of the celebrities, which is all public information. However, since this remains a legally gray area, we recommend exercising caution in future projects that use similar techniques. The SIS-M was conducted anonymously to ensure the confidentiality of participants’ responses. No identifying information was collected, allowing students to provide honest feedback without concern for personal attribution. The quantitative assessment of the “Shattered Slippers” CCN using the SIS-M is summarized in . The results indicated high levels in participants’ interest with the “Shattered Slippers” CCN, with the majority of students (14/18) indicating a preference for the CCN over traditionally presented clinical cases, only 1 student preferring the traditional approach, and 3 expressing no preference . Participants indicated a high average triggered situational interest in the CCN (mean 4.58, SD 0.53), as well as high maintained interest scores indicated by the students (mean 4.40, SD 0.53). The results for MF interest indicated high MF in students receiving the CCN (mean 4.38, SD 0.51). A feeling of educational value by the participants was supported by high scores for MV interest (mean 4.42, SD 0.54). Bridging quantitative data with qualitative insights, the survey conducted among participants also provided an open-ended question for students to reflect on their opinion of the CCN. Thematic analysis of the responses revealed the following: Enhanced engagement through storytelling and entertainment : The combination of storytelling and entertainment in the CCN heightened student engagement, making the learning process more enjoyable and effective compared with traditional methods. Improved memorability and recall of medical concepts : The CCN’s engaging narrative and multimedia elements enhanced memory retention, making complex medical information more accessible and memorable. Relatability through pop culture and personal connection : Leveraging familiar pop culture icons such as Selena Gomez helped students form a personal connection with the material, enhancing engagement and motivation to learn. Preference for interactive and detailed learning : Some students value interactive learning environments and detailed information, suggesting that while the CCN is engaging, it could be further enhanced by incorporating active learning elements and comprehensive content. Suggestions for improvement : Attention to technical elements, such as the use of genAI voice narration, could improve the overall effectiveness and reception of the CCN. The thematic analysis reveals that the CCN “Shattered Slippers” was preferred over traditional case studies due to its engaging storytelling, enhanced memorability, and relatability through pop culture references. While students appreciated the innovative approach, some expressed a desire for more interactive learning methods and provided suggestions for technical improvements. Incorporating these insights can further refine the CCN as a valuable tool in medical education. In addition to the survey feedback from the SIS-M, the success of the “Shattered Slippers” CCN was further demonstrated academically. Students displayed strong comprehension and knowledge of the material covered, achieving an average score of 88% on examination questions pertaining to the case study content. This high performance underscores the effectiveness of the CCN as a teaching tool, suggesting that it may also be useful in promoting academic performance as well as student preference and interest. Principal Findings The “Shattered Slippers” CCN supports the pedagogical value of integrating innovative genAI-driven methods and culturally resonant themes into medical education. Our study shows the capacity of this approach to not only enhance student interest but also promote their understanding and retention of complex subject matter. Furthermore, it adds very little to no extra time to the lecture material, as it basically reskins the existing material into a more cinematic experience. This is particularly important, as many new active learning teaching methodologies either extend the amount of time students spend with the material or cause instructors to remove large amounts of material in order to incorporate novel active learning activities. We considered it ethical to clearly mark AI-generated images of real individuals to avoid confusion but did not deem it necessary to label AI-generated material such as text or audio that was not mimicking a real-world person. As genAI models continue to improve in generating realistic images and cloned voices, it will become increasingly important to label AI-generated materials that mimic real-world individuals to prevent confusion with reality and avoid potential legal issues. This study shows the importance of engaging students beyond conventional didactic methods, suggesting that the inclusion of elements such as plot development, multimedia, and popular culture can make learning more relatable and impactful. The feedback from the SIS-M supports that this approach can effectively address the initial problem of student disengagement and the need for more effective educational strategies as identified in the introduction. The process of creating CCNs with genAI tools is highly efficient and cost-effective. Designing the case outline took about a day, while plot and narration generation were completed in seconds using GPT-4 and Eleven Labs. Image and theme song generation took under an hour each, with slight delays due to iterative refinement. Overall, the time investment was minimal compared with traditional methods. The required technical skills are basic, involving familiarity with genAI platforms for text, image, and audio generation and standard project management skills to integrate these elements into a PowerPoint slide deck. In terms of cost, the only expense was a US $20 per month subscription to ChatGPT; other platforms were used on free tiers. This low cost, combined with fast production times, makes migrating to this format highly accessible and efficient for educators, offering significant time and cost savings compared with traditional content creation methods of this caliber. Future directions of this work will explore how similar immersive educational experiences can be scaled and adapted for diverse student populations and learning environments. The versatility of genAI-enhanced CCNs extends beyond pharmacology, offering potential applications in other areas such as anatomy, pathology, and clinical skills. This pedagogical strategy can be adapted to various medical disciplines, making abstract topics more engaging and accessible to diverse learners. It also asks questions on how educational policies might evolve to integrate this type of AI-generated material into curricula systematically. As genAI becomes more integral to education, policies must address both the ethical use of genAI and the need for genAI literacy among educators and students. Personalized, genAI-driven learning experiences could revolutionize how content is delivered, providing flexibility and tailored learning opportunities. There is an opportunity to explore interdisciplinary collaborations, merging medical education with fields such as AI, storytelling, and multimedia design. These collaborations could further refine educational tools and help bridge the gap between traditional learning and modern health care technologies, fostering genAI literacy in future medical professionals. This promising pilot study shows potential for scalability and broad applicability of genAI-enhanced CCNs. The strategy offers a model for transforming how complex medical topics are taught, providing a scalable, engaging solution that can be adapted across different medical content areas to meet evolving educational needs. Limitations Our project has limitations in terms of cultural adaptability due to its reliance on specific cultural references and celebrity figures, which may not resonate with all audiences. Furthermore, the use of genAI technologies presents challenges in environments with varying levels of technological resources and differing instructor familiarity with these platforms. While the skills required to effectively use genAI can vary depending on the model, these challenges are mitigated by the increasing availability of more user-friendly genAI platforms. These platforms are simplifying AI integration in educational contexts, expanding the potential for their broader application. For instance, prompt engineering, which is crucial for optimizing output from LLMs, is becoming less essential with newer versions such as ChatGPT’s o1-preview model, which incorporates many of these strategies into the system itself. This reduces the need for advanced user expertise and lowers the barrier to efficient LLM use. Another limitation of our study is the process of validity checking for AI-generated content. Although the materials were reviewed by medical professionals, including physicians, PhDs, and PharmDs, to ensure accuracy, the use of genAI introduces potential risks in content reliability, especially as AI-generated content may produce subtle inaccuracies or lack the nuanced context that a human expert might provide. Future implementations of this approach would benefit from a formalized validation process to ensure that the clinical and educational integrity of AI-generated materials is maintained. The evaluation methodology, focusing on immediate reactions via the SIS-M, provides a single time point of the resource’s impact but does not capture the longevity of knowledge retention or the applicability of the learned material in clinical settings. Furthermore, the study included a limited sample size, with only 18 respondents to the SIS-M survey, which may not provide a comprehensive view of the broader student population. Future research could explore longitudinal studies to measure the lasting educational benefits of such methodologies with a larger participant population. Furthermore, our study lacked a control or comparison group, a common challenge in medical education research. All students in the study were exposed only to the CCN case, and without a traditional case-based learning comparison, it is difficult to isolate the exact impact of the CCN on student performance. While we acknowledge that a control group could provide valuable insights, the integration of such comparisons is often logistically difficult in medical school settings. Future studies could address this by designing more controlled experimental conditions or through the use of quasi-experimental designs to better understand the differential effects of various educational interventions on learning. Conclusions The “Shattered Slippers” CCN demonstrates the effectiveness of combining cinemeducation with genAI in medical education. This approach enhanced student engagement, promoted knowledge retention, and offered a novel perspective on complex pharmacological clinical cases. The application and positive student feedback suggest that this multimodal genAI approach to educational content creation has potential for broader application in medical education. Our project also highlights the need for continuous innovation and adaptation in teaching methodologies to meet the evolving demands of health care education. Future research and development in this area could further transform medical education, making it more engaging, effective, and aligned with modern technological advancements. The “Shattered Slippers” CCN supports the pedagogical value of integrating innovative genAI-driven methods and culturally resonant themes into medical education. Our study shows the capacity of this approach to not only enhance student interest but also promote their understanding and retention of complex subject matter. Furthermore, it adds very little to no extra time to the lecture material, as it basically reskins the existing material into a more cinematic experience. This is particularly important, as many new active learning teaching methodologies either extend the amount of time students spend with the material or cause instructors to remove large amounts of material in order to incorporate novel active learning activities. We considered it ethical to clearly mark AI-generated images of real individuals to avoid confusion but did not deem it necessary to label AI-generated material such as text or audio that was not mimicking a real-world person. As genAI models continue to improve in generating realistic images and cloned voices, it will become increasingly important to label AI-generated materials that mimic real-world individuals to prevent confusion with reality and avoid potential legal issues. This study shows the importance of engaging students beyond conventional didactic methods, suggesting that the inclusion of elements such as plot development, multimedia, and popular culture can make learning more relatable and impactful. The feedback from the SIS-M supports that this approach can effectively address the initial problem of student disengagement and the need for more effective educational strategies as identified in the introduction. The process of creating CCNs with genAI tools is highly efficient and cost-effective. Designing the case outline took about a day, while plot and narration generation were completed in seconds using GPT-4 and Eleven Labs. Image and theme song generation took under an hour each, with slight delays due to iterative refinement. Overall, the time investment was minimal compared with traditional methods. The required technical skills are basic, involving familiarity with genAI platforms for text, image, and audio generation and standard project management skills to integrate these elements into a PowerPoint slide deck. In terms of cost, the only expense was a US $20 per month subscription to ChatGPT; other platforms were used on free tiers. This low cost, combined with fast production times, makes migrating to this format highly accessible and efficient for educators, offering significant time and cost savings compared with traditional content creation methods of this caliber. Future directions of this work will explore how similar immersive educational experiences can be scaled and adapted for diverse student populations and learning environments. The versatility of genAI-enhanced CCNs extends beyond pharmacology, offering potential applications in other areas such as anatomy, pathology, and clinical skills. This pedagogical strategy can be adapted to various medical disciplines, making abstract topics more engaging and accessible to diverse learners. It also asks questions on how educational policies might evolve to integrate this type of AI-generated material into curricula systematically. As genAI becomes more integral to education, policies must address both the ethical use of genAI and the need for genAI literacy among educators and students. Personalized, genAI-driven learning experiences could revolutionize how content is delivered, providing flexibility and tailored learning opportunities. There is an opportunity to explore interdisciplinary collaborations, merging medical education with fields such as AI, storytelling, and multimedia design. These collaborations could further refine educational tools and help bridge the gap between traditional learning and modern health care technologies, fostering genAI literacy in future medical professionals. This promising pilot study shows potential for scalability and broad applicability of genAI-enhanced CCNs. The strategy offers a model for transforming how complex medical topics are taught, providing a scalable, engaging solution that can be adapted across different medical content areas to meet evolving educational needs. Our project has limitations in terms of cultural adaptability due to its reliance on specific cultural references and celebrity figures, which may not resonate with all audiences. Furthermore, the use of genAI technologies presents challenges in environments with varying levels of technological resources and differing instructor familiarity with these platforms. While the skills required to effectively use genAI can vary depending on the model, these challenges are mitigated by the increasing availability of more user-friendly genAI platforms. These platforms are simplifying AI integration in educational contexts, expanding the potential for their broader application. For instance, prompt engineering, which is crucial for optimizing output from LLMs, is becoming less essential with newer versions such as ChatGPT’s o1-preview model, which incorporates many of these strategies into the system itself. This reduces the need for advanced user expertise and lowers the barrier to efficient LLM use. Another limitation of our study is the process of validity checking for AI-generated content. Although the materials were reviewed by medical professionals, including physicians, PhDs, and PharmDs, to ensure accuracy, the use of genAI introduces potential risks in content reliability, especially as AI-generated content may produce subtle inaccuracies or lack the nuanced context that a human expert might provide. Future implementations of this approach would benefit from a formalized validation process to ensure that the clinical and educational integrity of AI-generated materials is maintained. The evaluation methodology, focusing on immediate reactions via the SIS-M, provides a single time point of the resource’s impact but does not capture the longevity of knowledge retention or the applicability of the learned material in clinical settings. Furthermore, the study included a limited sample size, with only 18 respondents to the SIS-M survey, which may not provide a comprehensive view of the broader student population. Future research could explore longitudinal studies to measure the lasting educational benefits of such methodologies with a larger participant population. Furthermore, our study lacked a control or comparison group, a common challenge in medical education research. All students in the study were exposed only to the CCN case, and without a traditional case-based learning comparison, it is difficult to isolate the exact impact of the CCN on student performance. While we acknowledge that a control group could provide valuable insights, the integration of such comparisons is often logistically difficult in medical school settings. Future studies could address this by designing more controlled experimental conditions or through the use of quasi-experimental designs to better understand the differential effects of various educational interventions on learning. The “Shattered Slippers” CCN demonstrates the effectiveness of combining cinemeducation with genAI in medical education. This approach enhanced student engagement, promoted knowledge retention, and offered a novel perspective on complex pharmacological clinical cases. The application and positive student feedback suggest that this multimodal genAI approach to educational content creation has potential for broader application in medical education. Our project also highlights the need for continuous innovation and adaptation in teaching methodologies to meet the evolving demands of health care education. Future research and development in this area could further transform medical education, making it more engaging, effective, and aligned with modern technological advancements. 10.2196/63865 Multimedia Appendix 1 Shattered Slippers: cinematic clinical narrative. 10.2196/63865 Multimedia Appendix 2 Shattered Slippers full presentation. 10.2196/63865 Multimedia Appendix 3 ChatGPT plot generation. 10.2196/63865 Multimedia Appendix 4 Leonardo.ai image generation. 10.2196/63865 Multimedia Appendix 5 Eleven Labs narration generation and audio clips. 10.2196/63865 Multimedia Appendix 6 ChatGPT and Suno Chirp Bot theme song generation and audio clip. 10.2196/63865 Multimedia Appendix 7 Situations Interest Survey of Multimedia CHERRIES (Checklist for Reporting Results of Internet E-Surveys) report. |
Development of a PHBV nanoparticle as a peptide vehicle for NOD1 activation | 9beb292d-2722-4456-a9e8-0468fbba838a | 8174487 | Pharmacology[mh] | Introduction Innate immunity is the first line of defense in a host that promotes resistance against pathogens. It consists of the immediate response of constant and non-specific intensity that requires different types of mechanisms and elements in action, which prevents the entry of pathogens into the body and their proliferation. Among those elements are external chemical, physical, and biological barriers (skin, mucous membranes, body secretion enzymes such as lysozyme, lactoperoxidase, intestinal, vaginal microbiota, and pH) (O'Hara & Shanahan, ). When this first defensive barrier is overcome, the activation of internal innate mechanisms (soluble factors and cellular components) is triggered, avoiding the pathogen's establishment, development, and action. In that case, immune cells such as monocytes, neutrophils, and macrophages are responsible for phagocytizing the pathogen at the infection site (Stossel, ). Moreover, those immune cells release cytokines (TNF-α, IL-1β, IL-6, IL-12) to generate an inflammatory response, while other humoral components, such as the complement, antimicrobial peptides (defensins), and the acute phase proteins participate in the process of recognition and elimination of the pathogen. Pathogen recognition receptors (PRR) are responsible for detecting different microorganisms by recognizing conserved structures, known as pathogen-associated molecular patterns (PAMPs). PAMPs, which initiate an innate immune response during infection (Motta et al., ), compromise members of the nucleotide-linked oligomerization domain type receptors (NLR) family. NLR family consists of 22 cytoplasmic pathogen sensors, which are characterized by having a trimeric structure. NOD1 and NOD2 receptors belong to an NLR subfamily, both containing LRR and NOD domains, and differ only in one or two copies of the caspase activation recruitment domains (CARDs). Both NOD1 and NOD2 exert a fundamental role in the defense against bacterial infections and regulate the host inflammatory response (Charlotte et al., ). NOD2 agonists are highly pyrogenic; thereby, their use in humans has been restricted (Monie, ). Since it is not the case for NOD1 agonists, these represent an attractive focus to obtain a more controlled response. Inohara et al. proposed that NOD1 and NOD2 receptors are intracellular sensors of bacterial lipopolysaccharide (LPS). However, in subsequent studies, they concluded that NOD receptors detect the peptidoglycan fragments from the bacterial cell wall (Chamaillard et al., ; Inohara et al., ). One of the NOD1 agonists is the dipeptide γ- d -glutamyl- meso -diaminopimelic acid (iE-DAP), which is found in the peptidoglycan of Gram-negative bacilli and particular Gram-positive bacteria such as Bacillus subtilis and Listeria monocytogenes . iE-DAP has been shown to stimulate the immune system by triggering a signaling cascade that leads to NF-κB activation and inflammatory cytokines production (Chamaillard et al., ; Girardin et al., ). The innate immune system's action begins with the activation of NOD1 by the binding of its bacterial ligand (iE-DAP) to the LRR domain. This event initiates a process of auto-oligomerization in its central zone, then interacts with its protein-2 adapter molecule (RIP2, receptor-interacting protein 2), a serine/threonine kinase that binds to the CARD domain of the NOD1 receptor (Girardin et al., ; Kobayashi et al., ; Hasegawa et al., ). Recruitment of RIP2 at this site of interaction is essential, as it communicates the signaling cascade's initiation through NF-κB or via the mitogen-activated protein kinase (MAPK) (Ogura et al., ; Kobayashi et al., ). The use of an effective transport and release system of adjuvants, such as NPs based on biodegradable and biocompatible polymers, would increase the efficiency of poorly soluble or highly toxic compounds, reducing, in turn, their side effects in addition to protecting the compound through encapsulation (Kumari et al., ). There are a variety of polymers that can be used; the best known is poly(lactic- co -glycolic acid) (PLGA) and poly(lactic acid) (PLA). These polymers have been approved for humans by the Food and Drug Administration (FDA) (Panyam & Labhasetwar, ). Remarkably, the poly(3-hydroxybutyrate- co -3-hydroxyvalerate) (PHBV) has been shown to generate a prolonged and controlled release of the encapsulated compounds (Vilos & Velasquez, ). PHBV has been extensively studied as a biomaterial in microparticle-based drug transport systems (Sendil et al., ; Vilos & Velasquez, ; Vilos et al., ). It has the advantage of being biodegradable, non-toxic, and with a low production cost compared to the other polymers such as PLGA (Slater et al., ). It has been observed that PHBV nanoparticles allow controlled release of the encapsulated content at short times, being a strategic option to be used as modulators in the activation of the innate immune system in vitro (Peñaloza et al., ). Efficient internalization of these nanoparticles would be under endocytic mechanisms, where their incorporation, traffic, destination, and cell degradation would be dependent on physicochemical properties of NPs (Sahay et al., ; Yameen et al., ). Due to the fact that the iE-DAP agonist is not permeable to the membrane, we propose that PHBV nanoparticles might deliver the iE-DAP agonist into the cell cytoplasm in a sustained manner over time, allowing the activation of the NOD1 receptor reflected by the secretion of pro-inflammatory cytokines.
Materials and methods 2.1. Materials Poly(3-hydroxybutyric acid- co -hydroxyvaleric acid) (PHBV) 12% w/w poly-3-hydroxyvalerate (PHV); polyvinyl alcohol (PVA) (average mol wt. 30,000–70,000); Escherichia coli O55 lipopolysaccharide: B5; TWEEN 20 and 0.45 µm Millipore Filters were purchased from Sigma-Aldrich (St. Louis, MO). Trypan Blue stain 0.4%; antibiotic–antifungal (100×); bovine fetal serum; trypsin-EDTA 1× and Nile Red 552/636 were purchased from Gibco by Life Technologies (Carlsbad, CA). Hoechst 33342 was purchased from Invitrogen (Carlsbad, CA). EDTA was purchased from Calbiochem (San Diego, CA). PBS was purchased from Winkler (Taipei City, Taiwan). Dichloromethane, sodium bicarbonate, sulfuric acid, hydrogen peroxide, Triton X-100, and Amicon Ultra centrifugation filters were purchased from Merck (v). CytoTox 96 Non-Radioactive LDH Cytotoxicity Assay was purchased from Promega (Madison, WI). IL-6 Human ELISA MAX Deluxe Set, Biolegend (San Diego, CA); TNF-α Human ELISA MAX Deluxe Set, Biolegend, and Nunc MaxiSorp ELISA Plate were purchased from Biolegend (San Diego, CA). NOD1 Agonist γ- d -Glu-mDAP (iE-DAP) were purchased from Invivogen (San Diego, CA). 2.2. Preparation of iE-DAP–loaded PHBV NPs PHBV NPs were developed using a water-oil-water double emulsion method (Vilos et al., ). Briefly, in one vial, 1 ml of PHBV (3 mg/ml) with 399 µl of MiliQ water + 1 µl of the iE-DAP ligand (5000 µg/ml) or 100 µl of RN (Red Nile) fluorophore (1000 µg/ml) was added. The first emulsion (w1/o1) was prepared by sonication (Sonic Vibra Cell, Equilab) for 40 s. The water-in-oil emulsion was further emulsified by sonication for 30 s under the same conditions in 2 ml of an aqueous solution of 5 mg/ml PVA (w2). The mix was deposited in 100 ml precipitated glass by adding 10 ml of water and subjected to orbital agitation at 300 rpm overnight (Multistirrer Magnetic Stirrer, Arquimed, Santiago, Chile). Solvent evaporation was performed at room temperature for 12 h. Formulated NPs were washed three times with 50 ml of MilliQ water using Amicon Ultra-4 centrifuge filters with a molecular weight of 100 kDa and then centrifuged again at 3000 rpm for 15 min, then particles were suspended in 1000 µl of MilliQ water and stored at 4 °C for fresh use or at −20 °C for a subsequent lyophilization under vacuum at −80° C for 8 h (Coimbra et al., ; Vilos et al., ). 2.3. PHBV NPs characterization by dynamic light scattering (DLS) Each batch of formulated NPs was suspended in 1 ml of PBS 1× pH 7.4, and nanoparticle size (nm), polydispersity coefficient (PDI), and zeta potential (mV) were determined by the light scattering technique using a Nano-ZS Zetasizer (Malvern Instruments Ltd., Malvern, UK). 2.4. Encapsulation efficiency of iE-DAP in PHBV Aliquots of 1 ml were taken from the supernatants of each batch. This volume was filtered into a vial to be determined by the UPLC Acquity system (Waters, Milford, MA) free iE-DAP agonist concentration. In parallel as a complementary methodology, 6 mg of PHBV NPs encapsulating iE-DAP were weighed and dissolved in a mixture of 10% (v/v) DCM/MeOH, and it was sonicated at an intensity of 30% for 5 min to degrade the NPs and release the encapsulated content. Subsequently, the suspension was centrifuged at 9000 rpm for 10 min, the supernatant was filtered and collected in a vial to quantify by UPLC the amount of iE-DAP ligand encapsulates in NPs of PHBV. The encapsulation efficiency (% EE) was calculated as follows: (1) % E E = Real i E − DAP mass obtained b y UPLC 10 μ g initial concentration o f i E − DAP × 100 *Real iE-DAP mass obtained by UPLC: mass of the dipeptide iE-DAP encapsulated in the NPs-PHBV and quantified by UPLC as described. *10 μg initial concentration of iE-DAP: mass of the initial iE-DAP dipeptide to encapsulate in the synthesis of NPs-PHBV. 2.5. Transmission electron microscopy (TEM) NPs structure was also characterized using transmission electron microscopy. One drop of the NP sample was placed onto an ultra-thin Lacey carbon-coated 400-mesh copper grid and allowed to dry at room temperature for 10 min prior to image acquisition, ensuring no more than 1 min of electron beam exposure to the sample. TEM images were acquired using an LVEM5 electron microscope (Delong Instrument, Montreal, Quebec, Canada) at a nominal operating voltage of 5 kV. The small volume of the vacuum chamber in the LVEM5 microscope facilitates rapid sample visualization within 3 min before observation. The low voltage used delivers high contrast in soft materials (up to 20-fold) compared with high-voltage electron microscopes, which use accelerating voltages of approximately 100 kV; this procedure facilitates the emission of staining procedures and allows the direct visualization of biological samples. Digital images were captured using a Retiga 4000 R camera (QImaging, Inc., Tucson, AZ) at its maximal resolution. 2.6. Release profile of the iE-DAP over time The in vitro kinetics release study of iE-DAP since PHVB NPs, 6 mg of iE-DAP–loaded PHBV NPs were suspended into 1 ml of PBS 1×, then incubated at 37 °C under constant agitation. At the times of 0, 1, 2, 3, 4, 6, 8, 10, and 12 h, it was centrifuged at 9000 rpm for 5 min, and 50 µl aliquots were removed, filtered, collected in a vial, and subsequently quantified by UPLC to determine the amount of iE-DAP agonist released over time. As a control, empty NPs were used. 2.7. Stability of NPs over time Different batches of synthesized NPs were stored for 4 weeks at 4 °C, then at different times of 0, 1, 2, 3, and 4 weeks their physicochemical properties (size, potential Z, and PDI) were evaluated by DLS. 2.8. Cell culture Raw 264.7 cell line (murine macrophages, ATCC TIB-71™) was grown in TR6002 bottles (Trueline, Hebron, IL) by adding 15 ml of RPMI-1640 medium supplemented with 10% v/v fetal serum bovine (FBS), 1 mM sodium pyruvate, and 1% v/v penicillin–streptomycin–amphotericin B. Cells were incubated at 37 °C and 5% CO 2 , changing the culture medium every 2–3 days and propagated when they reached between 80 and 90% confluence. 2.9. PHBV internalization assay To quantify the internalization rate of NPs by cells, dye-loaded PHBV NPs were formulated by using the same conditions of iE-DAP-loaded PHBV NPs. Nile Red is an uncharged hydrophobic molecule that functions as a fluorescent probe for intracellular lipids and hydrophobic proteins' domains. During the nanoparticles formulation process, Nile Red was dissolved together with the polymeric solution. To avoid counting nanoparticles that could have remained attached to the cell surface, we did at least four washes using PBS prior to the quantification using cytometry. We have used this protocol successfully in previous articles (Peñaloza et al., ) and following recommendations from other sources (Fernando et al., ; Snipstad et al., ). In the NPs uptake experiments, cells were incubated with 10, 50, 100, 500, and 1000 µg/ml of dye-loaded NPs, and analyzed at 1, 2, 4, 6, 24, and 48 h. Then, the supernatant was discarded, and cells were washed three times with PBS. Finally, cells were resuspended in 500 µl of PBS, and samples were analyzed by flow cytometry (BD Accuri™ C6, BD Biosciences, Franklin Lakes, NJ). Ten thousand events were determined using the FL2 detection filter. 2.10. LDH cytotoxicity assay Empty PHBV NPs were weighed and reconstituted with culture medium (1 mg/ml). Then, they were sonicated for 30 s with an amplitude of 30% for its homogenization. Afterward, cells were incubated with 10, 50, 100, 500, and 1000 µg/ml of NPs for 24 or 48 h at 37 °C and 5% CO 2 . 45 min before completing the final incubation time, 100 µl of 10× lysis buffer was added to the positive control, and supernatants from cells were taken and centrifuged at 14,000 rpm for 5 min. About 50 µl aliquots of centrifuged and supernatants were transferred and mixed with 50 µl of mixed substrate for 30 min at room temperature. Then, 50 µl of stop buffer was added to stop the enzymatic reaction, and the absorbance at 490 nm was read in a spectrophotometry reader (Synergy H1 Hybrid Reader, Biotek®, Winooski, VT). Absorbances were used to calculate the percentage of cytotoxicity of each sample by the following equation: (2) % cytotoxicity = experimental − negative control positive control − negative control × 100 experimental: absorbance of the experimental sample; *negative control: culture medium absorbance; *positive control: absorbance with lysis buffer. 2.11. Cellular viability by annexin V assay Detection of phosphatidylserine translocated outside the cell membrane is a critical step in apoptosis. Phosphatidylserine can be labeled by fluorescein-annexin isothiocyanate V allowing the detection of dead cells by apoptosis (Moretti et al., ). To determine cell viability, cells were incubated with different concentrations (10, 50, 100, 500, and 1000 µg/ml) of empty PHBV NPs at different times (1, 2, 4, 6, 24, and 48 h). Phosphatidylserine residues translocated outside the cell membrane were detected by Annexin V conjugated with APC (BD Pharmingen™, San Diego, CA). Then, cells (5 × 10 5 ) were washed with annexin V binding buffer (10 mM HEPES, pH 7.4; 280 nM NaCl; 5 mM CaCl 2 ) and resuspended in the binding buffer; then Annexin V conjugated to APC (1 µg/ml) was added, mixed and allowed to incubate at 37 °C for 15 m. Samples were then analyzed by flow cytometry, and 10,000 events were evaluated using the FL4 detection filter (BD Accuri™ C6, BD Biosciences, San Diego, CA). 2.12. Immunofluorescence assay The translocation of p65 subunit of cytoplasmic NF-κB into the nucleus was evaluated by fluorescence microscopy. Raw 264.7 cells were grown in coverslips in 24-well culture plates at a concentration of 5 × 10 5 cells per well with different controls and treatments for 24 at 37 °C and 5% CO 2 . After that, cells were washed with PBS and fixed with 4% paraformaldehyde at room temperature for 10 min and then cells were permeabilized with 0.2% Triton X-100 for 10 min, prior to being washed twice with 0.2% PBS-BSA. About 1 µg/ml of rabbit polyclonal antibody was incubated against p65 subunit for 1 h at room temperature, then incubated the secondary antibody IgG conjugated with Alexa-fluor 488. Cells were also incubated with Hoechst 33342® (1 µg/ml) as a nuclear marker for 1 h at room temperature. Finally, 1 µl of Fluoromount-G™ was added to mount samples on coverslips. Samples were then observed by fluorescence microscopy (LSM 510, Carl Zeiss, Jena, Germany), and pictures were analyzed by Dimension cellSens v1.7.1 software (Olympus Corp., Tokyo, Japan). The blue to red color of the nucleus was exchanged through an image post-editing to visualize a better contrast with merge. To analyze the levels of nuclear translocation of the p65 subunit, immunofluorescence images were quantified through the Fiji ImageJ software (Wessel & Hanson, ). 2.13. ELISA Experimental cultures carried out as the immunofluorescence assays, supernatants were collected and stored at −80 °C. Subsequently, the concentration of pro-inflammatory cytokines secreted into these supernatants was evaluated. ELISA assay was used to detect the protein expression of IL-6 and TNF-α following the manufacturer's instructions with mouse IL-6 and TNF-α ELISA kits (MAX Deluxe Set, Biolegend, San Diego, CA). The reaction was stopped by adding 100 µl of stop buffer (NH 2 SO 4 ) (yellow coloring). The colorimetric reaction was read at 450 nm in a spectrophotometry reader (Synergy H1 Hybrid Reader, Biotek®, Winooski, VT). 2.14. Statistical analysis All data were analyzed using GraphPad Prism version 5.03 software (GraphPad Software, La Jolla, CA). Results were analyzed using one-way ANOVA with subsequent Bonferroni test. ns = not significant, * p < .05, ** p < .01, *** p < .001, **** p < .0001.
Materials Poly(3-hydroxybutyric acid- co -hydroxyvaleric acid) (PHBV) 12% w/w poly-3-hydroxyvalerate (PHV); polyvinyl alcohol (PVA) (average mol wt. 30,000–70,000); Escherichia coli O55 lipopolysaccharide: B5; TWEEN 20 and 0.45 µm Millipore Filters were purchased from Sigma-Aldrich (St. Louis, MO). Trypan Blue stain 0.4%; antibiotic–antifungal (100×); bovine fetal serum; trypsin-EDTA 1× and Nile Red 552/636 were purchased from Gibco by Life Technologies (Carlsbad, CA). Hoechst 33342 was purchased from Invitrogen (Carlsbad, CA). EDTA was purchased from Calbiochem (San Diego, CA). PBS was purchased from Winkler (Taipei City, Taiwan). Dichloromethane, sodium bicarbonate, sulfuric acid, hydrogen peroxide, Triton X-100, and Amicon Ultra centrifugation filters were purchased from Merck (v). CytoTox 96 Non-Radioactive LDH Cytotoxicity Assay was purchased from Promega (Madison, WI). IL-6 Human ELISA MAX Deluxe Set, Biolegend (San Diego, CA); TNF-α Human ELISA MAX Deluxe Set, Biolegend, and Nunc MaxiSorp ELISA Plate were purchased from Biolegend (San Diego, CA). NOD1 Agonist γ- d -Glu-mDAP (iE-DAP) were purchased from Invivogen (San Diego, CA).
Preparation of iE-DAP–loaded PHBV NPs PHBV NPs were developed using a water-oil-water double emulsion method (Vilos et al., ). Briefly, in one vial, 1 ml of PHBV (3 mg/ml) with 399 µl of MiliQ water + 1 µl of the iE-DAP ligand (5000 µg/ml) or 100 µl of RN (Red Nile) fluorophore (1000 µg/ml) was added. The first emulsion (w1/o1) was prepared by sonication (Sonic Vibra Cell, Equilab) for 40 s. The water-in-oil emulsion was further emulsified by sonication for 30 s under the same conditions in 2 ml of an aqueous solution of 5 mg/ml PVA (w2). The mix was deposited in 100 ml precipitated glass by adding 10 ml of water and subjected to orbital agitation at 300 rpm overnight (Multistirrer Magnetic Stirrer, Arquimed, Santiago, Chile). Solvent evaporation was performed at room temperature for 12 h. Formulated NPs were washed three times with 50 ml of MilliQ water using Amicon Ultra-4 centrifuge filters with a molecular weight of 100 kDa and then centrifuged again at 3000 rpm for 15 min, then particles were suspended in 1000 µl of MilliQ water and stored at 4 °C for fresh use or at −20 °C for a subsequent lyophilization under vacuum at −80° C for 8 h (Coimbra et al., ; Vilos et al., ).
PHBV NPs characterization by dynamic light scattering (DLS) Each batch of formulated NPs was suspended in 1 ml of PBS 1× pH 7.4, and nanoparticle size (nm), polydispersity coefficient (PDI), and zeta potential (mV) were determined by the light scattering technique using a Nano-ZS Zetasizer (Malvern Instruments Ltd., Malvern, UK).
Encapsulation efficiency of iE-DAP in PHBV Aliquots of 1 ml were taken from the supernatants of each batch. This volume was filtered into a vial to be determined by the UPLC Acquity system (Waters, Milford, MA) free iE-DAP agonist concentration. In parallel as a complementary methodology, 6 mg of PHBV NPs encapsulating iE-DAP were weighed and dissolved in a mixture of 10% (v/v) DCM/MeOH, and it was sonicated at an intensity of 30% for 5 min to degrade the NPs and release the encapsulated content. Subsequently, the suspension was centrifuged at 9000 rpm for 10 min, the supernatant was filtered and collected in a vial to quantify by UPLC the amount of iE-DAP ligand encapsulates in NPs of PHBV. The encapsulation efficiency (% EE) was calculated as follows: (1) % E E = Real i E − DAP mass obtained b y UPLC 10 μ g initial concentration o f i E − DAP × 100 *Real iE-DAP mass obtained by UPLC: mass of the dipeptide iE-DAP encapsulated in the NPs-PHBV and quantified by UPLC as described. *10 μg initial concentration of iE-DAP: mass of the initial iE-DAP dipeptide to encapsulate in the synthesis of NPs-PHBV.
Transmission electron microscopy (TEM) NPs structure was also characterized using transmission electron microscopy. One drop of the NP sample was placed onto an ultra-thin Lacey carbon-coated 400-mesh copper grid and allowed to dry at room temperature for 10 min prior to image acquisition, ensuring no more than 1 min of electron beam exposure to the sample. TEM images were acquired using an LVEM5 electron microscope (Delong Instrument, Montreal, Quebec, Canada) at a nominal operating voltage of 5 kV. The small volume of the vacuum chamber in the LVEM5 microscope facilitates rapid sample visualization within 3 min before observation. The low voltage used delivers high contrast in soft materials (up to 20-fold) compared with high-voltage electron microscopes, which use accelerating voltages of approximately 100 kV; this procedure facilitates the emission of staining procedures and allows the direct visualization of biological samples. Digital images were captured using a Retiga 4000 R camera (QImaging, Inc., Tucson, AZ) at its maximal resolution.
Release profile of the iE-DAP over time The in vitro kinetics release study of iE-DAP since PHVB NPs, 6 mg of iE-DAP–loaded PHBV NPs were suspended into 1 ml of PBS 1×, then incubated at 37 °C under constant agitation. At the times of 0, 1, 2, 3, 4, 6, 8, 10, and 12 h, it was centrifuged at 9000 rpm for 5 min, and 50 µl aliquots were removed, filtered, collected in a vial, and subsequently quantified by UPLC to determine the amount of iE-DAP agonist released over time. As a control, empty NPs were used.
Stability of NPs over time Different batches of synthesized NPs were stored for 4 weeks at 4 °C, then at different times of 0, 1, 2, 3, and 4 weeks their physicochemical properties (size, potential Z, and PDI) were evaluated by DLS.
Cell culture Raw 264.7 cell line (murine macrophages, ATCC TIB-71™) was grown in TR6002 bottles (Trueline, Hebron, IL) by adding 15 ml of RPMI-1640 medium supplemented with 10% v/v fetal serum bovine (FBS), 1 mM sodium pyruvate, and 1% v/v penicillin–streptomycin–amphotericin B. Cells were incubated at 37 °C and 5% CO 2 , changing the culture medium every 2–3 days and propagated when they reached between 80 and 90% confluence.
PHBV internalization assay To quantify the internalization rate of NPs by cells, dye-loaded PHBV NPs were formulated by using the same conditions of iE-DAP-loaded PHBV NPs. Nile Red is an uncharged hydrophobic molecule that functions as a fluorescent probe for intracellular lipids and hydrophobic proteins' domains. During the nanoparticles formulation process, Nile Red was dissolved together with the polymeric solution. To avoid counting nanoparticles that could have remained attached to the cell surface, we did at least four washes using PBS prior to the quantification using cytometry. We have used this protocol successfully in previous articles (Peñaloza et al., ) and following recommendations from other sources (Fernando et al., ; Snipstad et al., ). In the NPs uptake experiments, cells were incubated with 10, 50, 100, 500, and 1000 µg/ml of dye-loaded NPs, and analyzed at 1, 2, 4, 6, 24, and 48 h. Then, the supernatant was discarded, and cells were washed three times with PBS. Finally, cells were resuspended in 500 µl of PBS, and samples were analyzed by flow cytometry (BD Accuri™ C6, BD Biosciences, Franklin Lakes, NJ). Ten thousand events were determined using the FL2 detection filter.
LDH cytotoxicity assay Empty PHBV NPs were weighed and reconstituted with culture medium (1 mg/ml). Then, they were sonicated for 30 s with an amplitude of 30% for its homogenization. Afterward, cells were incubated with 10, 50, 100, 500, and 1000 µg/ml of NPs for 24 or 48 h at 37 °C and 5% CO 2 . 45 min before completing the final incubation time, 100 µl of 10× lysis buffer was added to the positive control, and supernatants from cells were taken and centrifuged at 14,000 rpm for 5 min. About 50 µl aliquots of centrifuged and supernatants were transferred and mixed with 50 µl of mixed substrate for 30 min at room temperature. Then, 50 µl of stop buffer was added to stop the enzymatic reaction, and the absorbance at 490 nm was read in a spectrophotometry reader (Synergy H1 Hybrid Reader, Biotek®, Winooski, VT). Absorbances were used to calculate the percentage of cytotoxicity of each sample by the following equation: (2) % cytotoxicity = experimental − negative control positive control − negative control × 100 experimental: absorbance of the experimental sample; *negative control: culture medium absorbance; *positive control: absorbance with lysis buffer.
Cellular viability by annexin V assay Detection of phosphatidylserine translocated outside the cell membrane is a critical step in apoptosis. Phosphatidylserine can be labeled by fluorescein-annexin isothiocyanate V allowing the detection of dead cells by apoptosis (Moretti et al., ). To determine cell viability, cells were incubated with different concentrations (10, 50, 100, 500, and 1000 µg/ml) of empty PHBV NPs at different times (1, 2, 4, 6, 24, and 48 h). Phosphatidylserine residues translocated outside the cell membrane were detected by Annexin V conjugated with APC (BD Pharmingen™, San Diego, CA). Then, cells (5 × 10 5 ) were washed with annexin V binding buffer (10 mM HEPES, pH 7.4; 280 nM NaCl; 5 mM CaCl 2 ) and resuspended in the binding buffer; then Annexin V conjugated to APC (1 µg/ml) was added, mixed and allowed to incubate at 37 °C for 15 m. Samples were then analyzed by flow cytometry, and 10,000 events were evaluated using the FL4 detection filter (BD Accuri™ C6, BD Biosciences, San Diego, CA).
Immunofluorescence assay The translocation of p65 subunit of cytoplasmic NF-κB into the nucleus was evaluated by fluorescence microscopy. Raw 264.7 cells were grown in coverslips in 24-well culture plates at a concentration of 5 × 10 5 cells per well with different controls and treatments for 24 at 37 °C and 5% CO 2 . After that, cells were washed with PBS and fixed with 4% paraformaldehyde at room temperature for 10 min and then cells were permeabilized with 0.2% Triton X-100 for 10 min, prior to being washed twice with 0.2% PBS-BSA. About 1 µg/ml of rabbit polyclonal antibody was incubated against p65 subunit for 1 h at room temperature, then incubated the secondary antibody IgG conjugated with Alexa-fluor 488. Cells were also incubated with Hoechst 33342® (1 µg/ml) as a nuclear marker for 1 h at room temperature. Finally, 1 µl of Fluoromount-G™ was added to mount samples on coverslips. Samples were then observed by fluorescence microscopy (LSM 510, Carl Zeiss, Jena, Germany), and pictures were analyzed by Dimension cellSens v1.7.1 software (Olympus Corp., Tokyo, Japan). The blue to red color of the nucleus was exchanged through an image post-editing to visualize a better contrast with merge. To analyze the levels of nuclear translocation of the p65 subunit, immunofluorescence images were quantified through the Fiji ImageJ software (Wessel & Hanson, ).
ELISA Experimental cultures carried out as the immunofluorescence assays, supernatants were collected and stored at −80 °C. Subsequently, the concentration of pro-inflammatory cytokines secreted into these supernatants was evaluated. ELISA assay was used to detect the protein expression of IL-6 and TNF-α following the manufacturer's instructions with mouse IL-6 and TNF-α ELISA kits (MAX Deluxe Set, Biolegend, San Diego, CA). The reaction was stopped by adding 100 µl of stop buffer (NH 2 SO 4 ) (yellow coloring). The colorimetric reaction was read at 450 nm in a spectrophotometry reader (Synergy H1 Hybrid Reader, Biotek®, Winooski, VT).
Statistical analysis All data were analyzed using GraphPad Prism version 5.03 software (GraphPad Software, La Jolla, CA). Results were analyzed using one-way ANOVA with subsequent Bonferroni test. ns = not significant, * p < .05, ** p < .01, *** p < .001, **** p < .0001.
Results 3.1. Generation and characterization of iE-DAP-loaded PHBV NPs NPs loaded with iE-DAP agonist formulation were obtained by the double emulsion method. To this end, an aqueous solution of iE-DAP was emulsified in PHBV previously dissolved in dichloromethane (DCM). A second emulsification step was performed by adding PVA as a surfactant, followed by sonication. The resulting solution was stirred to allow the evaporation of DCM. iE-DAP ligand encapsulation did not exert significant changes in the physicochemical properties of the NPs compared to the void ones . Instead, low PDI values indicated that the particle size distribution was homogeneous between 198 and 215 nm and had a negative Z potential of −6.5 to −12.5 mV, as shown in . With minimal differences in size, PDI, and Z potential among all synthesized NP batches, double emulsion protocol was reproducible. Shape and morphology were also analyzed by transmission electronic microscopy . 3.2. Assessment of the encapsulation efficiency of iE-DAP peptides in PHBV NPs Two methods were performed to calculate the amount of iE-DAP agonists encapsulated in NPs. The first method consisted of UPLC quantification of free agonist molecules that were not encapsulated during the synthesis, remaining in the supernatant. The second method consisted of degrading PHBV-iE-DAP NPs by using organic solvents (DCM/MeOH), which allows the release of agonist molecules to be quantified. The encapsulation efficiency was 70.6 ± 7.2%, and the amount of iE-DAP ligand mass per nanoparticle weight was equivalent to 7.6 ± 0.7 µg/mg of PHBV, respectively . 3.3. Stability of iE-DAP-loaded PHBV-NPs To preserve NPs under storage conditions for subsequent tests, the colloidal stability of the iE-DAP-loaded PHBV-NPs in PBS solution was determined for 4 weeks at 4 °C. Degradation of synthesized NPs can be observed both by changes in their size and surface charge distribution. As shown in , results indicate no variations in the size and surface charge of the NPs in 4 weeks of storage, meaning that the compound is stable over time. 3.4. Release profile of the iE-DAP over time The release profile of encapsulated iE-DAP agonist molecules from NPs was evaluated for 12 h in PBS at 37 °C. As shown in , the agonist release profile behaves as a sigmoid curve, where more than 20% (1.4 µg) of the total amount of iE-DAP agonist is exchanged from the NPs during the first 2 h, and its maximum exchange occurs after 8 h. PHBV-NPs allow a constant release of iE-DAP, which would validate their use as drug delivery systems. 3.5. Cell internalization of PHVB NPs As the immunostimulatory strategy proposed here consists of NPs encapsulating iE-DAP agonists, the cellular uptake of NPs by murine RAW 264.7 macrophages was evaluated. To this end, PHBV NPs loaded with Red Nile fluorophore were synthesized and used to visualize the amount of internalized NPs inside the cell. Then, the physicochemical properties of the synthesized PHBV-RN NPs were characterized, which did not change compared to empty NPs, as shown in . PHBV-RN NPs showed a negative surface charge of −6.5 mV and a size of 214.4 nm with a polydispersity value of 0.28, indicating a homogeneous size distribution similar to that previously reported (Peñaloza et al., ). Then, PHBV-RN NPs were added at different concentrations (10, 50, 100, 500, and 1000 µg/ml) and at different times (1, 2, 4, 6, 12, and 24 h) to the cells to be analyzed later by flow cytometry. shows that the internalization of the NPs is dependent on concentration and time. It is important to remark that results obtained by both experiments (concentration and time) showed congruence, where MFI (mean fluorescent intensity) were consistent when cells were incubated with a final concentration of 100 µg/ml PHBV-RN for 24 h. 3.6. Assessment of the potential cytotoxicity induced by NPs The LDH assay was used to observe the potential cytotoxic effect of the iE-DAP-loaded NPs, as well as the concentration range in which they could be harmful to cells. In this assay, the cell membrane's integrity was evaluated, measuring the release of the LDH enzyme as an indirect indicator of cytotoxicity. The results indicate an average cytotoxicity level of 12.6% at 24 h and 8.1% at 48 h, respectively . This test showed that even at high concentrations of incubated NPs, no cytotoxic effect was shown ( p > .05), suggesting that these NPs do not harmfully interact with the cell membrane, regardless of the concentration. Treatments were standardized compared to the control (baseline LDH exchange). 3.7. Cellular toxicity of NPs As different concentrations of PHBV NPs generate a discrete cytotoxic effect on RAW 264.7 cells. The next step was to determine whether these NPs could be triggering some apoptosis process. To this end, we evaluated the integrity of the cell membrane by annexin V, an apoptosis early marker. Annexin V preferentially binds to phospholipids that are translocated from the inside to the outside of the cell membrane in the early stages of apoptosis. Results obtained by flow cytometry indicate no significant changes in apoptosis levels concerning the negative control with only culture medium (except at a concentration of 1000 µg/ml) . Furthermore, at a chosen concentration of 100 µg/ml, empty NPs do not show significant changes in apoptosis levels over time when measured at different times (1, 2, 4, 6, 24, and 48 h) . As a positive control, a concentration of 1 µM H 2 O 2 was used as an oxidizing agent capable of inducing apoptosis by over 50% in RAW 264.7 cells (Piao et al., ). 3.7. Evaluation of the activation of NF-κB factor by iE-DAP-loaded PHBV NPs NF-κB factor has been reported to be activated under LPS stimulation, causing its nuclear translocation and activating the transcription of pro-inflammatory genes (Baldwin, ). To verify whether iE-DAP-loaded NPs can activate the NF-κB factor, immunofluorescence assays were carried out by using an anti-p65 antibody to identify the p65 NF-κB component, which is regularly located in the cytosol of unstimulated cells (control). As shown in , when 100 µg/ml of iE-DAP-loaded NPs for 24 h are incubated in RAW 264.7 cells (stable stock stored for 3–5 months at −20 °C, data not shown), it was possible to observe a significant increase in nuclear accumulation of p65 compared with an equivalent dose of the iE-DAP agonist in its soluble form. Empty NPs were also evaluated, showing a minimal accumulation in the nucleus . 3.8. Evaluation of cytokines secretion elicited by iE-DAP-loaded PHBV NPs Finally, to explore the effect as a possible adjuvant, the secretion of two pro-inflammatory cytokines (IL-6 and TNF-α) was quantified by ELISA. Results in indicate that the iE-DAP agonist effect was improved by being encapsulated into PHBV NPs compared to its free form. Furthermore, it was possible to observe an increase of up to 5 times at 24 h and 2 times at 48 h of IL-6 levels, while TNF-α levels showed an increase of up to 4 times at 24 h and 16 times at 48 h compared to iE-DAP free form ( p < .01).
Generation and characterization of iE-DAP-loaded PHBV NPs NPs loaded with iE-DAP agonist formulation were obtained by the double emulsion method. To this end, an aqueous solution of iE-DAP was emulsified in PHBV previously dissolved in dichloromethane (DCM). A second emulsification step was performed by adding PVA as a surfactant, followed by sonication. The resulting solution was stirred to allow the evaporation of DCM. iE-DAP ligand encapsulation did not exert significant changes in the physicochemical properties of the NPs compared to the void ones . Instead, low PDI values indicated that the particle size distribution was homogeneous between 198 and 215 nm and had a negative Z potential of −6.5 to −12.5 mV, as shown in . With minimal differences in size, PDI, and Z potential among all synthesized NP batches, double emulsion protocol was reproducible. Shape and morphology were also analyzed by transmission electronic microscopy .
Assessment of the encapsulation efficiency of iE-DAP peptides in PHBV NPs Two methods were performed to calculate the amount of iE-DAP agonists encapsulated in NPs. The first method consisted of UPLC quantification of free agonist molecules that were not encapsulated during the synthesis, remaining in the supernatant. The second method consisted of degrading PHBV-iE-DAP NPs by using organic solvents (DCM/MeOH), which allows the release of agonist molecules to be quantified. The encapsulation efficiency was 70.6 ± 7.2%, and the amount of iE-DAP ligand mass per nanoparticle weight was equivalent to 7.6 ± 0.7 µg/mg of PHBV, respectively .
Stability of iE-DAP-loaded PHBV-NPs To preserve NPs under storage conditions for subsequent tests, the colloidal stability of the iE-DAP-loaded PHBV-NPs in PBS solution was determined for 4 weeks at 4 °C. Degradation of synthesized NPs can be observed both by changes in their size and surface charge distribution. As shown in , results indicate no variations in the size and surface charge of the NPs in 4 weeks of storage, meaning that the compound is stable over time.
Release profile of the iE-DAP over time The release profile of encapsulated iE-DAP agonist molecules from NPs was evaluated for 12 h in PBS at 37 °C. As shown in , the agonist release profile behaves as a sigmoid curve, where more than 20% (1.4 µg) of the total amount of iE-DAP agonist is exchanged from the NPs during the first 2 h, and its maximum exchange occurs after 8 h. PHBV-NPs allow a constant release of iE-DAP, which would validate their use as drug delivery systems.
Cell internalization of PHVB NPs As the immunostimulatory strategy proposed here consists of NPs encapsulating iE-DAP agonists, the cellular uptake of NPs by murine RAW 264.7 macrophages was evaluated. To this end, PHBV NPs loaded with Red Nile fluorophore were synthesized and used to visualize the amount of internalized NPs inside the cell. Then, the physicochemical properties of the synthesized PHBV-RN NPs were characterized, which did not change compared to empty NPs, as shown in . PHBV-RN NPs showed a negative surface charge of −6.5 mV and a size of 214.4 nm with a polydispersity value of 0.28, indicating a homogeneous size distribution similar to that previously reported (Peñaloza et al., ). Then, PHBV-RN NPs were added at different concentrations (10, 50, 100, 500, and 1000 µg/ml) and at different times (1, 2, 4, 6, 12, and 24 h) to the cells to be analyzed later by flow cytometry. shows that the internalization of the NPs is dependent on concentration and time. It is important to remark that results obtained by both experiments (concentration and time) showed congruence, where MFI (mean fluorescent intensity) were consistent when cells were incubated with a final concentration of 100 µg/ml PHBV-RN for 24 h.
Assessment of the potential cytotoxicity induced by NPs The LDH assay was used to observe the potential cytotoxic effect of the iE-DAP-loaded NPs, as well as the concentration range in which they could be harmful to cells. In this assay, the cell membrane's integrity was evaluated, measuring the release of the LDH enzyme as an indirect indicator of cytotoxicity. The results indicate an average cytotoxicity level of 12.6% at 24 h and 8.1% at 48 h, respectively . This test showed that even at high concentrations of incubated NPs, no cytotoxic effect was shown ( p > .05), suggesting that these NPs do not harmfully interact with the cell membrane, regardless of the concentration. Treatments were standardized compared to the control (baseline LDH exchange).
Cellular toxicity of NPs As different concentrations of PHBV NPs generate a discrete cytotoxic effect on RAW 264.7 cells. The next step was to determine whether these NPs could be triggering some apoptosis process. To this end, we evaluated the integrity of the cell membrane by annexin V, an apoptosis early marker. Annexin V preferentially binds to phospholipids that are translocated from the inside to the outside of the cell membrane in the early stages of apoptosis. Results obtained by flow cytometry indicate no significant changes in apoptosis levels concerning the negative control with only culture medium (except at a concentration of 1000 µg/ml) . Furthermore, at a chosen concentration of 100 µg/ml, empty NPs do not show significant changes in apoptosis levels over time when measured at different times (1, 2, 4, 6, 24, and 48 h) . As a positive control, a concentration of 1 µM H 2 O 2 was used as an oxidizing agent capable of inducing apoptosis by over 50% in RAW 264.7 cells (Piao et al., ).
Evaluation of the activation of NF-κB factor by iE-DAP-loaded PHBV NPs NF-κB factor has been reported to be activated under LPS stimulation, causing its nuclear translocation and activating the transcription of pro-inflammatory genes (Baldwin, ). To verify whether iE-DAP-loaded NPs can activate the NF-κB factor, immunofluorescence assays were carried out by using an anti-p65 antibody to identify the p65 NF-κB component, which is regularly located in the cytosol of unstimulated cells (control). As shown in , when 100 µg/ml of iE-DAP-loaded NPs for 24 h are incubated in RAW 264.7 cells (stable stock stored for 3–5 months at −20 °C, data not shown), it was possible to observe a significant increase in nuclear accumulation of p65 compared with an equivalent dose of the iE-DAP agonist in its soluble form. Empty NPs were also evaluated, showing a minimal accumulation in the nucleus .
Evaluation of cytokines secretion elicited by iE-DAP-loaded PHBV NPs Finally, to explore the effect as a possible adjuvant, the secretion of two pro-inflammatory cytokines (IL-6 and TNF-α) was quantified by ELISA. Results in indicate that the iE-DAP agonist effect was improved by being encapsulated into PHBV NPs compared to its free form. Furthermore, it was possible to observe an increase of up to 5 times at 24 h and 2 times at 48 h of IL-6 levels, while TNF-α levels showed an increase of up to 4 times at 24 h and 16 times at 48 h compared to iE-DAP free form ( p < .01).
Discussion NPs have shown great potential for delivery system due to their multiple properties, such as the sustained supply of low molecular weight antigens toward a biological target for prolonged periods, to promote the protection and stabilization of the compound encapsulated by premature degradation (Hu et al., ; Vilos & Velasquez, ; Kumari et al., ). In this work, we evaluate the use of PHBV NPs as a delivery system for NOD1 receptor activation. Here we showed that the intracellular fate of the NPs after endocytosis could favor the cytoplasmic delivery of the encapsulated agonist, increasing the activation efficiency of the NOD1 receptor, improving in this way its action compared to the soluble form of the agonist. PHBV polymers have recently started to be used similarly to well-known PLGA and PLA NPs, having the same physicochemical properties as the latter one and above all, due to its advantage of being an efficient alternative at low cost (Zhao et al., ). As illustrates, the nanoprecipitation process allows obtaining a high encapsulation performance of the agonist and reproducibility, having minimal differences in size, PdI, and Z potential between the synthesized batches. The high loading ability of the NPs to encapsulate iE-DAP peptides (70%) can be explained by the hydrophobic character of the agonists (partition coefficient octanol/water Log P = 1.22), which allows them to be encapsulated with high affinity by a hydrophobic nanoparticle such as PHBV (Zhao et al., ). In contrast, for hydrophilic compounds, such as the Ceftiofur antibiotic, PHBV has shown only 39.5% of encapsulation efficiency (Vilos et al., ). Stable encapsulation of the molecules inside NPs is essential to use them in therapies and prevent the degradation of the encapsulated content. To this end, we measured the release of the agonist molecules in the PHBV NPs as a function of time . Although kinetics release in PHBV NPs has not been fully characterized, release mechanisms by other NPs has been reported (Xu et al., ). We could observe that our NPs behave as a triphasic type kinetic release profile, beginning with an erosion and rapid initial release of the agonist attached to the surface of the NPs, then the slow degradation of the polymer through the diffusion of water in the polymeric matrix and finally a sustained exchange over time (Soppimath et al., ; Hu et al., ). The peptide's full release occurs after 8 h under in vitro conditions; these values are comparable to those previously obtained by Babos et al. and Pavot et al. . However, these values are contrasted by Silva et al. , with different immunostimulant agents encapsulated in NPs, by presenting a slow cumulative release around 15 days. This behavior can be explained by the size and polymer concentration of the formulation, which influences the release rate of the compounds and degradation of the NPs (Tod et al., ). Additionally, stability assessments of the NPs indicate that size and surface charge did not vary over 4 weeks, which suggests that they contain an optimal size for a passive release of molecules as future use in in vivo applications (Peer et al., ). On the other hand, one of the advantages for biomedical use of these NPs is their negative surface charge granted by the carboxyl functional groups (Gupta et al., ). Those negatively charged groups avoid excessive protein binding and, therefore, prevent the stimulation of the innate immune system, resulting in a longer circulatory half-life. Besides, the negative charge minimizes non-specific binding with the cell surface and prevents aggregation itself (Nel et al., ). There is a direct relationship between NPs physicochemical properties and an efficient capture by immune cells. While NPs among 500 nm in size are uptake by phagocytosis (Foster et al., ), it was reported that NPs with a size around 100–350 nm are internalized by clathrin-mediated endocytosis (Gratton et al., ; Peñaloza et al., ). Given that the size of PHBV NPs is around 198–215 nm , it would probably be internalized through this process. After that, they might be found in intracellular compartments such as endosomes and lysosomes (Benfer & Kissel, ). These compartments are characterized by having a lower luminal pH of 6–6.5 in early endosomes and 4.5–5.2 in lysosomes (Yameen et al., ; Peñaloza et al., ). Our laboratory's previous results showed colocalization of these FITC-loaded NPs with late endosomes at 30 min and in lysosomes at 2 h, distinguishing the release of the FITC fluorophore in the human neutrophil cytoplasm of HL-60 and murine macrophages RAW 264.7 in 4 and 6 h, respectively (data not shown). The biological activity of iE-DAP-loaded NPs and the cytotoxicity were determined . Interestingly, after evaluating different concentrations and exposure times, NPs only generated about 12% of cytotoxicity, a percentage that is considered biocompatible for biological systems (Nel et al., ). However, tests with annexin V showed that at concentrations (>1 mg/ml) of NPs, the cells would have significant apoptosis levels in 20% . This phenomenon could be caused by PHBV polymer remains after being degraded in endosomal compartments by acidic pH. One of its components, hydroxybutyric acid, is a ketone body that can be used as a favorable metabolic alternative substrate with protective effects against apoptosis mechanisms in PC12 cells (Cheng et al., ). Nevertheless, by contrast, the release of the alcohol groups of the PVA surfactant used in the NPs as a stabilizing agent could increase the intracellular oxidative stress and explain these apoptosis levels. For this reason, we decided to use a final concentration of 100 µg/ml that was non-cytotoxic at the concentration and time range evaluated (Ma et al., ; Silva et al., ). In its free form, IE-DAP was able to induce the activation of NF-κB and the consequent increase in pro-inflammatory cytokines , despite being impermeable to the membrane. This behavior could be due to cell entry mechanisms. A study in human epithelial cells HEK293 reported that a clathrin-mediated endocytosis process takes up NOD1 ligands such as iE-DAP and PHT1 (SLC15A4) in early endosomes favors their passage from the cytosol into cells (Lee et al., ). Another study on macrophages derived from mouse bone marrow also demonstrated the importance of PHT1 and showed that PEPT2 (SLC15A2) participates in the internalization process of NOD ligands (Hu et al., ). Immunofluorescence results indicate that iE-DAP-loaded PHBV NPs induced NF-κB activation after translocation in the nucleus . This type of activation was observed to a lesser extent using the agonist in its free form and NPs empty than the control. NPs would improve peptide immunostimulatory properties, probably due to reduced degradation of the peptide in the culture medium or enhanced cell entry of the PHBV-iE-DAP NPs, transporting its content more efficiently to NOD1 intracellular receptors. Furthermore, we showed here that the secreted amounts of IL-6 and TNF-α pro-inflammatory cytokines are higher by iE DAP-loaded NPs than the peptide in the free form . Those values are consistent with the work already reported (Wischke et al., ; Pavot et al., ). Nevertheless, the secreted levels of TNF-α by empty PHBV NPs at 48 h are significant, indicating that they could also be acting as an inducer. Results of Pavot et al. exposing empty PLA NPs in human MoDCs did not produce a significant secretion of TNF-α. They showed that the degradation of PVA would be contributing to the formation of oxidative stress, activation of apoptosis and secretion of TNF-α pro-inflammatory cytokine (Cheng et al., ). On the other hand, our laboratory results suggested phagocytosis as the PHBV NPs internalization mechanism in immune cells. To date, a specific receptor, or a possible mechanism of activation of the cytokine TNF-α through PHBV NPs that reflects this secretion is still unknown. Despite this evidence, this pleiotropic cytokine can act both autocrine and paracrine way through TNFR2 membrane receptors in immune cells, explaining its effect in the activation and release levels observed at 48 h (Gane et al., ). Taken together, these behaviors could lead to a constant overproduction of pro-inflammatory cytokines, generating a hyperinflammatory state at the systemic level, causing damage and dysfunction of tissues and organs (Cartwright et al., ). This result suggests that there must be a delicate balance between the innate immune response, which must be enough to eliminate the pathogen and the negative feedback system to prevent pathological inflammation. Therefore, subsequent investigations could determine the precise range of peptide release from NPs as a function of time. This profile can be modified using PHBV polymers with low molecular weight or PHBV with a higher valerate composition to retain compounds with higher affinity by hydrophobic compounds (Göpferich, ). Moreover, the saturation point of the peptide loading of NPs should also be determined to increase or decrease its encapsulation and maintain adequate concentration levels over time for therapeutic purposes. Some studies have shown that NOD1 agonists can be used prophylactically as a complement of antibiotics, which substantially improves survival in microbial sepsis models (Mine et al., ).
Conclusions In the present work, we proposed a new formulation composed of a NOD1 agonist encapsulated into biocompatible PHBV NPs that might control the immune response at cellular level. Our experimental results showed that PHBV NPs could encapsulate iE-DAP peptide without altering its immunogenic properties. Moreover, we demonstrate that the iE-DAP agonist effect was improved by encapsulation compared to its soluble form, by promoting an activation of NOD1 receptor in macrophages.
|
DNA methylation profiling as a model for discovery and precision diagnostics in neuro-oncology | e33a021c-c564-4bff-a0a5-3f2c7a4fac5a | 8561128 | Pathology[mh] | A variety of methods exist to quantify DNA methylation, each with various applications in clinical oncology. Restriction endonuclease digestion, affinity enrichment methods such as methylated DNA immunoprecipitation (MeDIP), and electrochemical assays are examples used to interrogate genome-wide methylation (see Laird, 2010 for a comprehensive review). Bisulfite sequencing of genomic DNA, the most commonly used method, converts unmethylated cytosine to uracil after exposure to sodium bisulfite. Methylated cytosine (5-mC), however, is resistant to this reaction and can be detected in subsequent sequencing steps. This method provides single base resolution of CpG methylation and, when applied at a genome-wide scale (whole-genome bisulfite sequencing, or WGBS), can quantify the methylation state of the approximately 29 million CpGs in the human genome. While WGBS provides an unprecedented scale and resolution of the human methylome, it is currently prohibitively expensive and computationally intensive. Of the approximately 11,000 cancer samples catalogued in The Cancer Genome Atlas (TCGA), only 39 have been profiled with WGBS. A comprehensive and cost-effective alternative is the complementary use of bisulfite-converted DNA and microarray technology. Illumina microarrays for CpG methylation are based on the hybridization of fragmented whole-genome amplification products to oligonucleotide bead arrays which contain oligomers linked to specific CpGs. , This approach is based on their Infinium genotyping assay and, when first introduced in 2008, provided coverage of over 27,000 CpG sites (HumanMethylation27 DNA Analysis BeadChip, or HM27). This has since been followed by the HumanMethylation450 (HM450, >450,000 CpGs, “450k”) and HumanMethylationEPIC (EPIC, >850,000 CpGs, “850k”) arrays released in 2011 and 2016, respectively. The HM450 array, of which EPIC covers >90%, contains probes pre-selected for biologically meaningful regions of the genome including RefSeq genes (98.9%), CpG islands (96%) and adjacent regions (e.g. shores, shelves), FANTOM4 promoters, predicted enhancers, MHC regions, and DNase hypersensitivity sites. Thus, while the Infinium BeadChip arrays (HM450, EPIC) cover only ~1–3% of the human DNA methylome, it provides comprehensive coverage of cancer-relevant CpG sites at a comparatively lower cost and higher throughput than WGBS, as well as being compatible with formalin-fixed tissue and demonstrating robustness to pre-analytic factors. provides an overview of the current diagnostic applications of DNA methylation profiling and its role in surgical neuropathology. Recently, a novel adaption of long-read DNA sequencing using the Oxford Nanopore MinION sequencer has enabled accurate profiling of DNA modifications such as CpG methylation. The technique infers the methylation status using a modified hidden Markov model (HMM) to learn ionic current distributions and can achieve an accuracy up to 95% in identifying 5-mC. , The method does not require bisulfite treatment and can be performed in a significantly shorter amount of time (~hours) compared to methylation-based assays (~days). A recent study successfully extended this technology to classifying CNS tumors using the Heidelberg brain tumor classifier (classifier details discussed later). Notably, the relatively quick turnaround time may also allow for intraoperative tumor classification. To translate methylation array data into practical use, a variety of analytic and computational steps need to be performed. The proprietary Illumina file format (*.IDAT) contains the raw signal intensities which can be processed using the R programming language (with or without R Studio, an easy-to-use graphical user interface for R); a popular R package to process methylation array data is Minfi . After parsing the.IDAT files, intensities are typically normalized to correct for unwanted signal variation, and probes that are unreliable or map to polymorphic regions of the genome are removed. The most informative probes are then selected for further analysis (feature selection) based on variability in the dataset (e.g., choosing the top 10,000 most variable probes based on standard deviation). A frequently used and practical approach for delineating tumor types with methylation data is dimensionality reduction. Similar to principal component analysis (PCA), these unsupervised algorithms reduce high-dimensional data (e.g. thousands of brain tumor samples, each with ~20–30k of data points) to a lower dimension (2 or 3) for visualization. The most widely used techniques are t-Distributed Stochastic Neighbor Embedding (t-SNE) and Uniform Manifold Approximation and Projection (UMAP). The underlying mathematics is beyond the scope of this review, suffice it to say both techniques produce similar visual information when clustering bulk CNS tumor samples . However, a few details are worth mentioning. The scalability of t-SNE is limited and, thus, PCA is performed before embedding. It is also thought that UMAP preserves the global structure of the data better than t-SNE. Finally, as a general rule, the distances between the observed clusters are not as meaningful as the distances within the clusters. For example, points that cluster together are more similar than points in other clusters; however, the distances between clusters are not as informative. It is also important to mention other diagnostically useful information that can be derived from methylation array data. It was recognized early on that the sum of methylated and unmethylated signal intensities can be used to infer genome-wide copy number. The segmented copy number data can then be used to detect diagnostically useful alterations such as whole chromosome (+7/−10 signature) and arm-level (1p/19q codeletion) changes, as well as gene-level alterations (e.g. EGFR amplification, CDKN2A/B homozygous deletion) . Furthermore, analysis of copy number breakpoints can also be used to infer possible fusion events resulting from unbalanced translocations (e.g. BRAF , FGFR fusions). The diagnostic impact of integrating methylation and copy number data in neuropathology practice was recently reviewed. Finally, both O-6-methylguanine-DNA methyltransferase (MGMT) and mutL homologue 1 (MLH1) promoter methylation can be quantified with the array-based platform . While discrepancies with pyrosequencing-based methods have been noted, DNA methylation-based MGMT promoter methylation status has been shown to provide clinically relevant prognostic information. , Both copy number and MGMT promoter methylation status are included in the report generated from the “Classifier” (see below). Since its adaptation as an investigative diagnostic tool in 2018, Infinium methylation arrays have led to the identification of new CNS tumor types and subtypes, many of which harbor meaningful associations with clinical course and outcome. One of the most important contributions from this study is the implementation of a machine learning-based classifier to prospectively evaluate new samples (the Heidelberg “Classifier,” https://www.molecularneuropathology.org/mnp ). Benchmarking of the Classifier with a routine diagnostic work-up (e.g. histology, IHC) revealed high concordance (838/1104, 76%) with a correspondingly low number of cases with profiles that were discordant with clinical, pathologic, and molecular features (“misleading” profile). Notably, methylation results led to diagnostic revisions in a significant proportion (129/139 or 12% of the entire cohort) of cases. This finding was confirmed in five external centers (50/401, 12%), with reclassification rates in 6%-25% of cases. A recent study independently applied the Classifier to a large cohort of samples in the work-up of routine and challenging cases. The authors included 502 samples from 480 pediatric and adult primary and recurrent tumors. To assess its diagnostic impact, they adopted similar outcome categories for the Classifier results, including confirmation/refinement of diagnosis, establishment of a new diagnosis, misleading profile, discrepant but non-contributory, and “no match.” The results extended those reported by Capper et al. : 54.4% of cases were concordant with histopathology, with a precise match achieved in 41% of samples and 13.3% receiving a refined diagnosis. Classifier performance was comparable between primary and recurrent tumors (66.1% vs. 65.5% matching a class, respectively). A new diagnosis was established in favor of the methylation class in 9.8% of cases (vs. 12% ). The majority of cases in which a new diagnosis was rendered (49/502) also resulted in a change in WHO grade: 22.5% (11/49, vs. 30% ) were downgraded and 48.9% (24/49, vs. 41% ) were upgraded. A similar proportion of cases had a misleading profile (5/502 vs. 10/1104 ). Interestingly, there was a demonstrated benefit in cases not meeting the cutoff for a class match (calibrated score <0.9). In unmatched cases with a calibrated score between 0.3 and 0.9 (130/502), the histopathologic diagnosis could be confirmed or refined in 50.7%, and a new diagnosis rendered in 7.7% of cases. These results provide independent confirmation of the diagnostic and prognostic impact of DNA methylation-based classification. Finally, differences in clinical outcome based on methylation data were recently demonstrated in tumors previously classified as primitive neuroectodermal tumors, or CNS-PNET, which are now known to resolve into biologically distinct tumor types with specific alterations. These include embryonal tumor with multilayered rosettes (ETMR) with a high frequency of C19MC alterations, CNS neuroblastoma with FOXR2 -activation (CNS-NB-FOXR2), CNS tumors with BCOR internal tandem duplication (CNS-BCOR-ITD), CIC -rearranged sarcomas (EFT-CIC), CNS tumors with MN1 alteration, and a subset of high-grade gliomas (HGG). While detection of these alterations can be done through targeted sequencing or cytogenetic techniques, methylation array profiling can resolve CNS-PNET types independent of the underlying alteration, , some of which have disparate clinical outcomes. A retrospective survival analysis of CNS-PNET types showed prolonged 5-year progression-free (5y-PFS) and overall survival (5y-OS) in CNS-NB-FOXR2 (52%, 96%, respectively) compared with HGG (12%, 12%) and ETMR (12%, 18%). Although MN1 -altered tumors typically contain an astroblastoma-like histology, , a subset may present with PNET-like features. Distinction from other embryonal tumors is thus important as MN1 -altered tumors have a comparatively favorable overall survival. , Collectively, distinctions based on unsupervised analysis of DNA methylation data among CNS-PNET subtypes have important clinical implications. In addition to its utility in recognizing common CNS tumor types, an emerging advantage of methylation profiling is its ability to discover novel types and provide confirmation/refinement of existing types. For example, IDH1/2 mutations define a broad subtype of diffuse gliomas that are clinically and molecularly distinct from isocitrate dehydrogenase (IDH)-wildtype diffuse gliomas. This genetic distinction was confirmed with DNA methylation profiling and has recently been refined to identify a clinically-distinct subtype (e.g. mismatch repair-deficient ). For many of the newly-described tumor types, however, methylation profiling has largely served as a tool for discovery. Below, we outline selected new and anticipated tumor types where methylation profiling has significantly contributed to its discovery or refinement. In the majority of cases, this method represents a highly sensitive assay that captures a wide range of histology and genetic alterations (e.g. mutations, CNVs, fusions) that converge on a common epigenetic signature. High-grade astrocytoma with piloid features (HGAP) is a recently described tumor type that comprises a distinct methylation group and harbors a variable combination of alterations in telomere maintenance ( ATRX , TERT promoter mutations), the cell cycle pathway ( CDKN2A/B deletion/mutation , CDK4 amplification), and the MAPK pathway ; the latter can include NF1 deletion/mutations (somatic or germline), BRAF fusions (e.g. KIAA1549 : BRAF ), FGFR1 fusion/mutations, or, rarely, KRAS or BRAF p.V600E mutations. The histologic findings largely overlap with other gliomas and can include features of pilocytic astrocytoma (PA) and pleomorphic xanthoastrocytoma (PXA), but can also mimic prototypic glioblastoma (GBM) with increased mitotic activity, microvascular proliferation, and necrosis. , HGAP are most commonly seen in young adults (median 40 years) and are found predominantly in the posterior fossa but can be seen in the supratentorial region. Notably, HGAP is thought to be rare in children (age 0–16), with the majority of morphologically-diagnosed “pilocytic astrocytoma with anaplasia” in this age group clustering with other methylation classes, including pilocytic astrocytoma and IDH-wildtype GBM. HGAP forms a distinct group on t-SNE or UMAP embedding of DNA methylation array data . The methylation class “anaplastic pilocytic astrocytoma” (ANA PA, not to be confused with the aforementioned “pilocytic astrocytoma with anaplasia”) is included in the current version of the Classifier (v11b4), and is anticipated to be included in an upcoming expanded iteration as HGAP (v12). This change in designation is intended to underline the aggressiveness and wider morphological spectrum compared to pilocytic astrocytoma. Presently, methylation profiling is the only method to diagnose this tumor type. The genetic alterations found in HGAP (e.g., CDKN2A/B homozygous deletion, ATRX mutation, BRAF fusion) are, in isolation, not specific and can be seen in other CNS tumor types. However, the combination of these genetic alterations, or even the presence of ATRX loss by immunohistochemistry in a pilocytic-like IDH- and H3-wildtype astrocytoma, should raise the suspicion of HGAP. Clinically, distinction from other high-grade gliomas is important as HGAP has a more aggressive outcome than PA and PXA, but significantly better overall survival than IDH-wildtype GBM. The importance of an accurate diagnosis is underscored by the fact that HGAP is not infrequently misdiagnosed as GBM. As highlighted in our review, resolving histologically-similar tumor types into biologically and clinically relevant types is among the significant advantages of DNA methylation-based classification . Infant-type (infantile) hemispheric glioma (IHG) is a genetically heterogeneous group of high-grade gliomas that are primarily seen in children less than one year of age (median age 2.8 months). The histopathology of IHG is not specific and may include gemistocytic, gangliocytic, ependymal, and primitive features, in addition to the typical appearance of a high-grade glioma. , IHG tend to involve the leptomeninges and can disseminate within the neuraxis. A common genetic feature in this group is the presence of receptor tyrosine kinase (RTK) alterations consisting of fusions in the ALK , ROS1 , NTRK1/2/3 , and MET genes present in 61–83% of reported cases. , Despite the range of histopathology and the variety of fusion genes, a common DNA methylation signature helps define IHG. However, identifying the specific fusion can provide valuable options for therapeutic targeting. In the US, two drugs are currently FDA-approved for use in any solid tumor harboring an NTRK fusion or TRK oncoprotein (“tumor-agnostic”): entrectinib and larotrectinib. , Importantly, entrectinib recently showed in vitro efficacy and sustained CNS exposure in an intracranial tumor model. While ALK and ROS1 fusions are relatively specific to IHG among gliomas, NTRK fusions have also been observed in adult-type glioblastoma subtypes, pilocytic astrocytoma, pleomorphic xanthoastrocytoma, and H3K27M-mutant diffuse midline glioma. Methylation profiling thus serves as a sensitive diagnostic assay for IHG and can direct therapeutically-relevant sequencing. There are currently two CNS tumor types harboring an MYB or MYBL alteration: diffuse astrocytoma, MYB or MYBL-altered (DA-MYB/L), and angiocentric glioma (AG). Collectively, these are IDH/H3-wildtype diffuse gliomas with low-grade histologic features that harbor structural alterations in the MYB or MYBL genes. These tumors commonly arise in the cerebral hemispheres but can rarely present in the infratentorial region. , DA-MYB/L are frequently associated with seizures due to their cortical location in many cases and belong to a group of developmental CNS neoplasms called long-term epilepsy-associated tumors (LEAT). There are often overlapping histologic features between DA-MYB/L and AG. The monomorphic appearance of DA-MYB/L, fine fibrillary matrix, and diffuse infiltration led to its initial description as “isomorphic diffuse astrocytoma” in 2004. AG frequently consists of bipolar spindle tumor cells oriented radially in a rosette-like pattern around vessels. In 22 sequenced cases, eight DA-MYB/L contained an MYBL1 fusion and three contained an MYB fusion. In contrast, the majority of AG harbor an MYB - QKI fusion (found in up to 99% of cases). The methylation class in the current version of the Classifier, “LGG, MYB / MYBL -altered,” contains both isomorphic diffuse glioma and angiocentric glioma. However, unsupervised analysis of these tumors with other high- and low-grade gliomas has demonstrated a clear distinction between the two on both dimensionality reduction and hierarchical clustering. While the MYB - QKI fusion is present in the vast majority of AG, non- QKI partners have also been detected. Interestingly, t-SNE embedding also suggests a separation between these tumors and the previously reported “pediatric-type MYB / MYBL diffuse astrocytoma”. , It is likely that diffuse gliomas harboring an MYB or MYBL alteration (fusion or CNV) comprise multiple methylation classes with distinct clinicopathologic features. Diffuse glioneuronal tumor with oligodendroglioma-like features and nuclear clusters (DGONC) is a recently proposed, methylation-defined tumor type discovered through unsupervised clustering of >25,000 CNS tumors. This represents true class discovery as a result of the increasing use of the Classifier. The existence of this class was further supported through iterative resampling and cluster stability analysis, as well as the identification of copy number changes involving chromosome 14 in almost all cases. This was independently confirmed in a separate series that also noted chromosome 1p and 3p loss. The histologic appearance of DGONC may have significant overlap with other tumor types, including oligodendroglioma and CNS-PNET. Genetic alterations typically seen in low-grade glial or glioneuronal tumors, such as FGFR1 and BRAF , have not been identified in sequenced cases. , As genetic drivers have yet to be elucidated in this tumor type, DNA methylation profiling remains the only method for its detection. Rosette-forming glioneuronal tumor (RGNT) is a low-grade (WHO grade 1) glioneuronal tumor with hybrid histologic features including neurocytic rosettes/pseudorosettes and a glial component most frequently resembling pilocytic astrocytoma. A subset of RGNT may lack this biphasic appearance and contain a predominantly oligodendroglioma-like or PA-like histology. , Indeed, 3 of 10 methylation-defined RGNT in a recent report were initially characterized as “low-grade oligodendroglial tumor NOS (not otherwise specified)” and lacked the characteristic neurocytic rosettes. Genetically, RGNT are among the FGFR1 -altrered spectrum of LGG/LGNT and harbor either an FGFR1 p.N546 or p.K656 mutation, identified in 40 of 40 methylation-defined cases to date. , There is also a high frequency of co-occurring PIK3CA / PIK3R1 and NF1 mutations detected in 28/40 and 14/40 cases, respectively. , The combination of hotspot FGFR1 and PIK3CA / PIK3R1 or NF1 mutations is reasonably specific for RGNT with compatible histology. RGNT forms a distinct cluster on both dimensionality reduction (t-SNE, UMAP) and hierarchical clustering. , , When faced with incomplete histologic features, RGNT may closely resemble PA. Furthermore, FGFR1 p.N546K or p.K656 hotspot mutations have been detected in methylation-defined PA. Conversely, tumors epigenetically aligned with PA may contain RGNT-like histology with neurocytic rosettes. Therefore, a subset of tumors clustering with RGNT may contain overlapping genetic and histologic features of PA, and vice versa . Further studies are required to determine the clinical and/or biologic significance of these discrepancies. Nevertheless, methylation profiling serves as a sensitive method for the diagnosis of RGNT without the need for hotspot detection of FGFR1 mutations. Confirmation of FGFR1 mutations as a ubiquitous finding in RGNT may also support methylation profiling as a useful surrogate for this alteration and for directing targeted therapy (e.g. FGFR1 inhibition ). Embryonal tumor with multilayered rosettes (ETMR) is a high-grade (WHO grade 4) primitive CNS tumor with distinct histologic features consisting of foci of pseudostratified primitive neuroepithelial cells arranged circumferentially around a lumen (true “ependymoblastic” rosettes) and containing variable amounts of interspersed neuropil. ETMR is now the term given for three morphologic variants: embryonal tumor with abundant neuropil and true rosettes (ETANTR), ependymoblastoma, and medulloepithelioma. The common molecular alteration is a structural alteration in the microRNA cluster on chromosome 19q13.42 (C19MC); this mostly frequently manifests as a TTYH1 - C19MC fusion with concurrent C19MC amplification, the latter present in 87.9% (167/190) of primary ETMR in a recent large series, but can also include DICER1 mutations (germline or somatic) in 11.4% (8/70), or amplification of the miR-17–92 miRNA cluster on chromosome 13 in 4.2% (3/70). A subset of ETMR do not show miRNA cluster amplification or DICER1 mutations and can harbor alternative C19MC partners including MYO9B . This genomic heterogeneity is anticipated to refine the “embryonal tumor with multilayered rosettes, C19MC-altered” nomenclature in the 5th edition of the WHO blue book to encompass non-amplified and DICER1 -mutant cases. ETMR methylation profiles show clear separation from other CNS tumor types and form a relative homogenous group regardless of C19MC amplification or DICER1 mutation status. The characteristic histology of multilayered rosettes with admixed regions of neuropil is present in the vast majority of cases. In combination with LIN28A immunohistochemistry, the diagnosis can be readily made at the microscope. However, reports of a glioneuronal-like appearance and divergent differentiation (osteoid, myeloid, epithelial) , may complicate its recognition, particularly in relapse specimens. Despite its high sensitivity, LIN28A expression is not specific to ETMR and has been reported in AT/RT, , germ cell tumors, and HGG. DNA methylation profiling thus serves as a sensitive and specific method for the diagnosis of ETMR, regardless of the underlying histopathology or genetic alteration. While outcome differences between amplified and non-amplified ETMR have yet to be established, methylation-based copy number assessment may also serve as a useful feature to identify potential subtypes. CNS tumor with BCOR internal tandem duplication (CNS-BCOR-ITD) is a recently described high-grade tumor of uncertain differentiation (often termed “neuroepithelial”). These rare tumors can occur on either side of the tentorium cerebelli and are primarily seen in children and young adults (median age 3.5 years). High-grade histology is typically encountered (increased proliferation, necrosis) with a conspicuous paucity of microvascular proliferation in one series. A myxoid background is usually seen, and ependymoma-like (perivascular pseudorosettes) or embryonal-like (Homer Wright-like rosettes) features may be encountered. , Currently, the defining genetic alteration for this tumor type is an internal tandem duplication (ITD) in exon 15 of the BCOR gene on Xp11.4. Clinical follow-up is limited, but 4/10 patients in one series had evidence of recurrence (range 4–49 months) and combined survival analysis of 24 patients showed a poor prognosis (median OS: 1.7 years). While specific among CNS tumors, the exon 15 ITD in BCOR that defines this tumor type is also present in clear cell sarcoma of the kidney and undifferentiated round cell sarcoma/primitive myxoid mesenchymal tumor of infancy. , The methylation profile of CNS-BCOR-ITD is distinct and the diagnosis can be reliably made with methylation array profiling. Interestingly, genetic alterations other than BCOR ITD may also occur in tumors related to this methylation class. A recent study reported two high-grade CNS tumors harboring an EP300 : BCOR fusion that clustered with the “CNS high grade neuroepithelial tumor with BCOR alteration” methylation class ; we independently found a sample from the TCGA harboring an EP300 : BCOR fusion also clustered near tumors of this methylation class . While the further characterization is needed, BCOR -altered CNS tumors may extend beyond the exon 15 ITD. CNS neuroblastoma, FOXR2-activated (CNS-NB-FOXR2) is a rare embryonal tumor with typical primitive histologic features (Homer Wright rosettes) accompanied by occasional neurocytic and/or gangliocytic differentiation. The spectrum of genetic alterations in CNS-NB-FOXR2 is limited to a single study. In 6/8 sequenced samples, structural alterations involving the FOXR2 gene on Xp11.21 were identified, including three samples harboring FOXR2 fusion transcripts mediated through gene translocation or tandem duplication. A novel mitochondrial-nuclear fusion was also reported. As discussed previously, the distinction of CNS-NB-FOXR2 from other tumors manifesting as CNS-PNET may have clinical relevance, with a reported better outcome compared to other CNS embryonal tumors. The utility of methylation profiling for CNS-NB-FOXR2 may encompass both diagnosis and surrogate identification of the underlying alteration with copy number changes. In ~28% of cases (13/46), 450k-derived copy number demonstrated FOXR2 alterations including deletion and breakpoints, of which three cases that were sequenced harbored a FOXR2 fusion trancript. There are currently four SMARCB1 -inactivated (INI-1-deficient) primary CNS tumor types that are either established or provisionally recognized: atypical teratoid/rhabdoid tumor (AT/RT), desmoplastic myxoid tumor (DMT), cribriform neuroepithelial tumor (CRINET), and poorly differentiated chordoma (PDC). Comprehensive analyses of AT/RT (WHO grade 4) have revealed distinct methylation subtypes (SHH, TYR, MYC) and their clinicopathologic significance is discussed in detail elsewhere. , As shown in , these tumor types often cluster in close proximity to each other (CRINET data not available). Loss of the INI-1 protein by immunohistochemistry has classically been associated with AT/RT and is due to inactivation of the SMARCB1 gene (22q11.2) through structural alterations or mutations , ; however, a small subset can harbor an inactivating alteration in the SMARCA4 gene with corresponding loss of the BRG1 protein. DMT are, in contrast, low-grade appearing tumors restricted to the pineal region and similarly harbor inactivating alterations in SMARCB1 . CRINET are low-grade neuroepithelial tumors that show INI-1 loss primarily through 22q hemizygous deletion with a “second hit” inactivating mutation/deletion in SMARCB1 . PDC are primarily sacrococcygeal tumors with focal rhabdoid histology and INI-1 loss. SMARCB1 inactivation in PDC has only been demonstrated through gene deletion, with no SMARCB1 mutations reported to date. The distinction between these tumor types is clinically important and can be frequently resolved with histopathology and imaging. However, atypical cases may pose diagnostic difficulty. AT/RT and PDC (spinal) have a mean OS of 14.4 and 51 months, respectively; however, survival estimates vary with AT/RT subtype. Survival data for CRINET and DMT are limited but have been reported with a mean OS of 125 (vs. 37 for AT/RT-TYR) and 36 months, respectively. Specific DNA methylation profiles for AT/RT, PDC, and DMT are evident despite sharing a common genetic alteration . Limited data on CRINET has shown that it clusters with the AT/RT-TYR subtype on methylation profiling. The data so far has shown that DNA methylation profiling can resolve most SMARCB1 -inactivated CNS tumors that may pose diagnostic challenges. CIC-rearranged sarcoma is a rare high-grade mesenchymal neoplasm that occurs in both CNS and extra-CNS sites. The extra-CNS counterpart was recently introduced in the 5th edition of the WHO Classification of Soft Tissue and Bone Tumors in 2020. Histologically, CIC -rearranged sarcomas are characterized by spindled and/or round tumor cell morphology with high-grade features (increased proliferation, necrosis) and variable amounts of myxoid matrix. The characteristic genetic alteration is a translocation involving CIC with various partner genes resulting in a fusion that acts as a dominant oncogene. In the CNS, this most frequently occurs as a CIC - NUTM1 fusion resulting from a t(15;19) translocation. False negatives with CIC break-apart FISH were been reported in 14% of CIC fusion-positive cases in one series. Additionally, a CIC frameshift mutation had previously been identified in a fusion-negative case. A recent report also noted the occurrence of a non- CIC fusion gene involving ATXN1 and DUX4 . Originally termed a “Ewing-like” sarcoma, survival outcomes of CIC -rearranged sarcomas are significantly worse than Ewing sarcoma. Furthermore, treatment with Ewing sarcoma-based therapy showed an inferior pathologic response in 70% of CIC -rearranged sarcomas. Thus, distinction from EWS is clinically and prognostically important. CIC -rearranged sarcomas cluster in a distinct group on methylation profiling independent of the underlying fusion gene/mutation or anatomic site . DICER1-mutant primary intracranial sarcoma is a newly-recognized high-grade sarcoma that is defined by either somatic and/or germline mutation(s) in the DICER1 gene on chromosome 14q32.13. Tumors that occur in the context of a germline DICER1 mutation may represent a hereditary or syndromic association (i.e. DICER1 syndrome). Histologic features have been frequently reported with myogenic differentiation and morphologic and immunophenotypic overlap with embryonal rhabdomyosarcoma. Consequently, a tentative name “spindle cell sarcoma with rhabdomyosarcoma-like features, DICER1 mutant” was given. Similar to DICER1 -mutant pineoblastoma, a subset of these tumors do not harbor a germline mutation and may have multiple somatic DICER1 mutations or copy number loss that results in biallelic inactivation. DICER1 mutations are not specific to intracranial sarcomas in this tumor class. As previously discussed, a subset of methylation-defined ETMR also harbors somatic DICER1 mutations in the absence of C19MC alterations. A recent report also noted a loss of the H3K27me3 epigenetic mark in DICER1 -mutant tumors including intracranial sarcoma, pineoblastoma, and ETMR, thus expanding the spectrum of H3K27me3-negative CNS tumors. In this context, morphologic and immunohistochemical distinction from H3K27me3-negative MPNST may be difficult. Additionally, the histologic findings of DICER1 -mutant intracranial sarcoma are not specific and may overlap those of other sarcomas such as synovial sarcoma, fibrosarcoma, and gliosarcoma. DNA methylation profiling readily distinguishes DICER1 -mutant primary intracranial sarcoma from its histologic mimics. However, further studies are needed to assess whether it can be differentiated from metastatic DICER1 -mutant tumors. Here we present examples of the contribution of DNA methylation profiling to the subtyping of established WHO CNS tumors. The 5 th edition of the WHO classification for CNS tumors is anticipated to include an increased number of molecularly-defined subtypes. Methylation array profiling has been shown to effectively identify and, in some cases, define these subtypes, as outlined below. Diffuse leptomeningeal glioneuronal tumor (DLGNT) was recently introduced in the revised 4 th edition of the WHO Classification of CNS Tumors in 2016. DLGNT are rare, predominantly pediatric tumors characterized by glial and neuronal differentiation, leptomeningeal involvement, and a monomorphous oligodendroglial-like cellular morphology. The histologic features, however, can be variable and may show substantial overlap with other tumor types including pilocytic astrocytoma, anaplastic astrocytoma, primitive neuroectodermal tumor, ganglioglioma, and atypical neurocytoma. While they generally contain low-grade histologic features, a subset of DLGNT demonstrate features of anaplasia (mitoses, microvascular proliferation, necrosis) and have been associated with shorter overall survival in this context. Genetically, there is a high frequency (75%) of the KIAA1549 - BRAF fusion, chromosome 1p deletion (59–100%), , and, less commonly, 1p/19q codeletion (18–30%). , IDH1 or IDH2 mutations were not detected in sequenced cases (n=10 and n=5, respectively). Importantly, alternative mechanisms of MAPK pathway activation may be seen in DLGNT including BRAF p.V600E mutations, , NTRK1/2/3 , and RAF1 fusions, and FGFR1 mutations. MAPK alterations are not specific and can be seen in other glial or glioneuronal tumors. The presence of the KIAA1549 - BRAF fusion, in conjunction with an oligodendroglial-like low-grade histology, may be difficult to distinguish from pilocytic astrocytoma; this is further confounded by the fact that Rosenthal fibers and eosinophilic granular bodies (EGBs) may also be seen in DLGNT. Thus, there appears to be significant molecular and histologic heterogeneity in DLGNT which can be diagnostically challenging at both the microscopic and genetic level. In DLGNT, methylation profiling is useful both as a diagnostic and prognostic assay. A recent study analyzing 30 methylation-confirmed cases of DLGNT identified two subclasses: MC-1 and MC-2. In this series, methylation array-based copy number profiling demonstrated chromosome 1p loss as a frequent finding in DLGNT. A focal gain of 7q34 was observed in 20/30 DLGNT cases, indicating the likely presence of the KIAA1549 - BRAF gene fusion. Age at presentation was also significantly different between the two methylation subclasses, with a median age of 5 for MC-1 and 14 years for MC-2. Most significant, however, was that the DLGNT MC-2 subclass showed shorter OS and PFS in this cohort. While this finding needs to be confirmed, it is likely DNA methylation-based DLGNT subtypes will provide valuable prognostic information going forward. Posterior fossa ependymomas are a clinically and molecularly heterogeneous group of ependymal tumors. They are the prototypic epigenetically-driven CNS tumor and have been shown to (largely) lack recurrent genetic alterations. There are two broad types of posterior fossa ependymomas supported by clinical, gene expression, and DNA methylation data: group A (PFA) and group B (PFB). Morphologic features alone are insufficient to distinguish the two. At the microscope, loss of the H3K27me3 mark is currently diagnostic of PFA in the correct histologic and anatomic context, and results from PRC2 inhibition via EZHIP overexpression. PFA ependymomas have a significantly worse overall survival compared to PFB, highlighting the importance of this distinction. Histologic grading is likely to have little impact on prognosis within molecular subgroups of ependymomas, including PFA and PFB. A subset of subependymomas within the fourth ventricle and posterior fossa (methylation class “subependymoma, posterior fossa”) may contain mixed histologic features of ependymoma and subependymoma. , In our experience, features of classic ependymoma may be encountered in this class even in the total absence of characteristic subependymoma histology. The clinical significance of these histologically-confirmed ependymomas that cluster with posterior fossa subependymoma is unclear. Methylation array profiling is a powerful diagnostic tool for posterior fossa ependymomas and may also provide tumor type-specific prognostic information. In a study of 675 PFA ependymomas, two subgroups and nine subtypes were identified. Within the PFA-1 subgroup, poorly prognostic subtypes were identified by methylation clustering. Currently, distinguishing PFA from PFB is the most clinically relevant use of methylation profiling in posterior fossa ependymomas; however, additional analyses may identify or confirm these clinically aggressive PFA subtypes. PFB ependymomas have similarly been shown to harbor molecular heterogeneity. Five distinct methylation-based subtypes were identified in a cohort of 212 PFB ependymomas. Differences in patient demographics and copy number (13q loss as a prognostic marker) were the predominant findings from subtype analysis. Thus, in addition to the reliable separation of PF ependymoma types by methylation profiling, copy number analysis may provide type-specific prognostic information. Medulloblastomas (MB) are among the most histologically, molecularly, and clinically characterized CNS tumors. The revised 4th edition of the WHO blue book introduced genetically-defined entries for MB subtypes that reflect their significant biologic and clinical heterogeneity. These subtypes include WNT-activated (wingless signal transduction pathway), SHH (sonic hedgehog signaling pathway), and non-WNT/non-SHH (groups 3 and 4). It is clinically important to make this molecular distinction: WNT-activated MB have a significantly better prognosis (>95% OS at 5 years in pediatric patients), while groups 3 and 4 have poor OS and often have metastatic disease at presentation. The clinical and genetic features that comprise these subtypes has been reviewed elsewhere. A recent large-scale combined analysis of 1,501 group 3 and 4 MB revealed eight subtypes using both methylation and gene expression data. In addition to genetic differences among these subtypes, three risk groups were identified with significantly different survival outcomes: group 1 (“high-risk”) had a 5-year OS of 50%, compared to the “standard risk” group with a 5-year OS of 82%. This refinement of group 3/4 MB by methylation profiling was also translated into an ancillary random forest classifier that can be used in conjunction with the original Heidelberg Classifier. The additional refinement of medulloblastoma subtypes by methylation profiling is likely to be translated into the clinic in the future. The integration of DNA methylation profiling into the routine work-up of CNS tumors has demonstrated improved diagnostic precision and clinically meaningful stratification of tumor types. The ability to capture the immense morphologic and genetic heterogeneity in primary CNS tumors remains the most robust feature of DNA methylation profiling. This is particularly valuable in histologically-ambiguous tumor types that may harbor targetable alterations (e.g., IHG). The technique has been demonstrated to be robust in confirming histologic diagnoses and may soon be adapted to providing useful diagnostic information even prior to receiving fixed tissue (H&E slides) for evaluation (e.g. intraoperative long-read sequencing). A notable limitation of the Infinium BeadChip assay is the relatively higher amount of DNA that is required for hybridization (~250 ng). While this is significantly less than WGBS and RRBS requirements, brain tumor biopsies are often small and frequently yield a suboptimal amount of DNA. However, in our practice, we have been able to achieve meaningful classification results down to 20 ng of DNA (unpublished data). An inherent limitation to the use of bulk tissue for analysis is the presence of admixed non-neoplastic cells that may attenuate or obscure the signal of interest. These contaminants are particularly prevalent in brain tumor specimens where infiltration of normal brain tissue is a defining feature of many CNS tumor types. While other variables exist (DNA quality, batch effects), DNA yield and tumor purity are notable challenges in the translation of DNA methylation arrays to routine clinical practice. Not infrequently, classification results may be non-contributory (suboptimal classifier score) or not congruent with the clinical, histopathologic, or molecular features of the case (“misleading profile”). While this can be due to the aforementioned pre-analytic factors, interpretation can often be improved with visual inspection of unsupervised UMAP or t-SNE embedding(s). In our experience, the power of this approach is largely dependent on the sample size and distribution of tumor types. In the setting of a “suggestive” classifier score (e.g. below 0.85), our experience has shown that concordance of orthogonal methods to assess methylation class (t-SNE, UMAP) serves to add confidence in suggesting a specific diagnosis for a tumor. Alternatively, in the instance where these complementary methods are discordant, confidence for a specific tumor class is lower, and often in these cases a purely descriptive diagnosis is rendered. Additionally, appropriate bioinformatics techniques and quality control should be carefully employed. As we move forward with machine learning-based classifier development, real-time unsupervised clustering of tumor samples and employment of complementary methods of classification will undoubtedly form a useful adjunct for difficult-to-classify cases. Finally, the recent implementation of DNA methylation array-based classification in bone and soft tissue pathology may herald a similar paradigm shift for other subspecialties. Of the 33 cancer types profiled through the TCGA, comprehensive molecular profiling has revealed biologically and/or prognostically-meaningful subgroups in all types based on unsupervised clustering of DNA methylation data (see references in the ), further underscoring its potential clinical utility in other cancers. Given the increasing reliance on methylation profiling in CNS tumor diagnostics and its adoption in clinical classification criteria, further study and utilization of this clinically useful platform are warranted in the practice of diagnostic neuropathology, in particular for uncommon and rare CNS neoplasms. noab143_suppl_Supplemental_material Click here for additional data file. |
Therapeutic exercises, manual therapy, and health education program for adolescents with temporomandibular disorders: face-to-face and online multimodal rehabilitation protocol for a randomized controlled clinical trial | a1279b44-2f30-4f79-8bae-d4affc84588e | 11829351 | Patient Education as Topic[mh] | Background and rationale {6a} Temporomandibular disorder (TMD) is a public health problem with a multifactorial etiology that is influenced by biopsychosocial factors . Multimodal rehabilitation has shown good results in adults with TMD; however, there is still doubt about the ideal format (in-person or online) and its effectiveness in adolescents who need healthy growth and development; however, there are few studies involving this population. It is more frequent in adults , although studies have reported that the symptoms begin during childhood or adolescence, which corresponds to the period of maturation of the musculoskeletal system with hormonal, physiological, and behavioral changes. Factors related to TMD include kinesiophobia and parafunction, which significantly influence the prognosis . Therefore, it is important that these factors are incorporated into the treatment of adolescents with TMD to avoid perpetuation or progression, since TMD symptoms and harmful oral habits can manifest early and persist into adulthood, leading to serious consequences such as joint overload and pain . In many cases, TMD predominantly affects the muscles, resulting in the reduced availability of circulating oxygen, accumulation of specific metabolites that interfere with muscle contractile function, and increased metabolic demand for functional activities, which generates pain and fatigue . A cross-sectional study investigating hemodynamic variations in the masseter muscle at rest and during contraction in adolescents revealed that those with TMD exhibited reduced levels of oxyhemoglobin compared with healthy individuals, highlighting the need for early intervention in this population . TMD rehabilitation using physiotherapeutic resources aims to conservatively improve biopsychosocial aspects . This approach seeks to alleviate pain, improve function, stimulate proprioception, promote the production of synovial fluid in the joint, and improve the elasticity of muscle fibers . Multimodal physiotherapeutic intervention in TMD is based on the literature, demonstrating positive effects in the short and medium term, with the reduction of signs and symptoms of TMD or its severity. The highlighted modalities included manual therapy (MT), therapeutic exercises, and self-care guidance. Patient counseling provides autonomy, and responsibility , especially when associated with relaxation therapies,thus, showing effective results in chewing muscle pain in patients with TMD . MT is effective in improving the pain, range of motion, function, and severity of TMD ; hence, it shows better results when associated with home exercises , with training of the craniocervical flexor muscles demonstrating a significant reduction in orofacial pain and headache . Specific exercises for TMD are indicated both in face-to-face physiotherapy , and telerehabilitation (TR) . TR can be another vehicle for use in the evaluation or treatment of TMDs, by physiotherapists or health professionals specialized in the area. The advantages of TR over face-to-face treatment include ease of screening, referral, clarification, time, distance optimization, and cost reduction . Some studies have demonstrated the effectiveness of TR in the evaluation and treatment of TMD in adults and in children and adolescents with other conditions . However, investigations on TR and TMD in adolescents are scarce . Objectives {7} This study aims to describe a multimodal protocol with therapeutic exercises in-person and via telerehabilitation among adolescents with TMD. To evaluate effectiveness, the peripheral muscle oxygenation, pain intensity, range of mandibular movement, kinesiophobia, and parafunction will be assessed. Trial design {8} It will be a longitudinal study with a quantitative and analytical nature and is designed as a controlled and blinded randomized clinical trial (RCT) for statistical analyses. We followed the SPIRIT (Standard Protocol Items: Recommendations for Interventional Trials) in the conduction of this protocol. All participants will be asked to provide written informed consent using standard forms prior to randomization. Participants aged 18 years or parents or guardians of children under 18 years will sign the Free and Informed Consent Form (TCLE) and the consent form for videos, photographs, and recordings. The Free and Informed Assent Form (TALE) will be signed by adolescents under 18 years of age after given a complete explanation of all the procedures. Temporomandibular disorder (TMD) is a public health problem with a multifactorial etiology that is influenced by biopsychosocial factors . Multimodal rehabilitation has shown good results in adults with TMD; however, there is still doubt about the ideal format (in-person or online) and its effectiveness in adolescents who need healthy growth and development; however, there are few studies involving this population. It is more frequent in adults , although studies have reported that the symptoms begin during childhood or adolescence, which corresponds to the period of maturation of the musculoskeletal system with hormonal, physiological, and behavioral changes. Factors related to TMD include kinesiophobia and parafunction, which significantly influence the prognosis . Therefore, it is important that these factors are incorporated into the treatment of adolescents with TMD to avoid perpetuation or progression, since TMD symptoms and harmful oral habits can manifest early and persist into adulthood, leading to serious consequences such as joint overload and pain . In many cases, TMD predominantly affects the muscles, resulting in the reduced availability of circulating oxygen, accumulation of specific metabolites that interfere with muscle contractile function, and increased metabolic demand for functional activities, which generates pain and fatigue . A cross-sectional study investigating hemodynamic variations in the masseter muscle at rest and during contraction in adolescents revealed that those with TMD exhibited reduced levels of oxyhemoglobin compared with healthy individuals, highlighting the need for early intervention in this population . TMD rehabilitation using physiotherapeutic resources aims to conservatively improve biopsychosocial aspects . This approach seeks to alleviate pain, improve function, stimulate proprioception, promote the production of synovial fluid in the joint, and improve the elasticity of muscle fibers . Multimodal physiotherapeutic intervention in TMD is based on the literature, demonstrating positive effects in the short and medium term, with the reduction of signs and symptoms of TMD or its severity. The highlighted modalities included manual therapy (MT), therapeutic exercises, and self-care guidance. Patient counseling provides autonomy, and responsibility , especially when associated with relaxation therapies,thus, showing effective results in chewing muscle pain in patients with TMD . MT is effective in improving the pain, range of motion, function, and severity of TMD ; hence, it shows better results when associated with home exercises , with training of the craniocervical flexor muscles demonstrating a significant reduction in orofacial pain and headache . Specific exercises for TMD are indicated both in face-to-face physiotherapy , and telerehabilitation (TR) . TR can be another vehicle for use in the evaluation or treatment of TMDs, by physiotherapists or health professionals specialized in the area. The advantages of TR over face-to-face treatment include ease of screening, referral, clarification, time, distance optimization, and cost reduction . Some studies have demonstrated the effectiveness of TR in the evaluation and treatment of TMD in adults and in children and adolescents with other conditions . However, investigations on TR and TMD in adolescents are scarce . This study aims to describe a multimodal protocol with therapeutic exercises in-person and via telerehabilitation among adolescents with TMD. To evaluate effectiveness, the peripheral muscle oxygenation, pain intensity, range of mandibular movement, kinesiophobia, and parafunction will be assessed. It will be a longitudinal study with a quantitative and analytical nature and is designed as a controlled and blinded randomized clinical trial (RCT) for statistical analyses. We followed the SPIRIT (Standard Protocol Items: Recommendations for Interventional Trials) in the conduction of this protocol. All participants will be asked to provide written informed consent using standard forms prior to randomization. Participants aged 18 years or parents or guardians of children under 18 years will sign the Free and Informed Consent Form (TCLE) and the consent form for videos, photographs, and recordings. The Free and Informed Assent Form (TALE) will be signed by adolescents under 18 years of age after given a complete explanation of all the procedures. Study setting {9} Rooms are available at two physiotherapy clinics in different locations to facilitate movement and improve participant adherence. Assessments and face-to-face consultations will take place in these clinics according to the participants’ choice, with each room having the same standard of care and an adequate, lit, and air-conditioned environment. For TR services, participants will be questioned in advance about the quality and stability of their internet network and informed about the need for a silent and illuminated environment, and the use of the social network WhatsApp® via smartphone, as a guarantee of service. Eligibility criteria {10} The inclusion criteria were the following: (1) adolescents aged between 10 and 18 years of both sexes; (2) regularly enrolled and attending school activities; and (3) diagnosed with TMD, confirmed by Diagnostic Criteria of Temporomandibular Disorders (DC/TMD), both pain-related TMD or intra-articular TMD subtypes, will be included in this study, since it is a multimodal protocol that covers all aspects of TMDs. The diagnosis will be carried out by a trained examiner. (4) Adolescents with chronic or acute pain, considering that the protocol uses strategies to improve oxygenation and pain reduction, and preventive counseling, that will actuate for both pain modalities. We adopted as a cut-off criterion for pain, classifying pain as acute (less than 3 months) or chronic (more than 3 months) . The following exclusion criteria will be adopted: (1) adolescents who presented with neurological or respiratory diseases, musculoskeletal disorders, or other disabilities that may interfere with understanding or carrying out the protocol; (2) received physiotherapeutic care (for the region to be treated, within 3 months prior to the evaluation) or medication treatment (analgesics and/or muscle relaxant) 24 h before the evaluation; and (3) used an orthodontic appliance (mobile or fixed) or interocclusal splint. Who will take informed consent? {26a} Before commencing the intervention, written documentation of informed consent is acquired subsequent to approval by the Ethics Committee. Upon meeting the inclusion criteria, adolescents aged over 18 years or their parents/legal representatives, in the case of minors under 18, will be furnished with a comprehensive parent information packet (comprising the Free and Informed Consent Form, the Free and Informed Assent Form, and the Consent Form for photographs, videos, and recordings), elucidating details about the trial and intervention. They retain the prerogative to withdraw their child from the trial at any juncture and for any reason, without impeding the child's ongoing treatment. Additional consent provisions for collection and use of participant data and biological specimens {26b} This trial does not involve collecting biological specimens for storage. Rooms are available at two physiotherapy clinics in different locations to facilitate movement and improve participant adherence. Assessments and face-to-face consultations will take place in these clinics according to the participants’ choice, with each room having the same standard of care and an adequate, lit, and air-conditioned environment. For TR services, participants will be questioned in advance about the quality and stability of their internet network and informed about the need for a silent and illuminated environment, and the use of the social network WhatsApp® via smartphone, as a guarantee of service. The inclusion criteria were the following: (1) adolescents aged between 10 and 18 years of both sexes; (2) regularly enrolled and attending school activities; and (3) diagnosed with TMD, confirmed by Diagnostic Criteria of Temporomandibular Disorders (DC/TMD), both pain-related TMD or intra-articular TMD subtypes, will be included in this study, since it is a multimodal protocol that covers all aspects of TMDs. The diagnosis will be carried out by a trained examiner. (4) Adolescents with chronic or acute pain, considering that the protocol uses strategies to improve oxygenation and pain reduction, and preventive counseling, that will actuate for both pain modalities. We adopted as a cut-off criterion for pain, classifying pain as acute (less than 3 months) or chronic (more than 3 months) . The following exclusion criteria will be adopted: (1) adolescents who presented with neurological or respiratory diseases, musculoskeletal disorders, or other disabilities that may interfere with understanding or carrying out the protocol; (2) received physiotherapeutic care (for the region to be treated, within 3 months prior to the evaluation) or medication treatment (analgesics and/or muscle relaxant) 24 h before the evaluation; and (3) used an orthodontic appliance (mobile or fixed) or interocclusal splint. Before commencing the intervention, written documentation of informed consent is acquired subsequent to approval by the Ethics Committee. Upon meeting the inclusion criteria, adolescents aged over 18 years or their parents/legal representatives, in the case of minors under 18, will be furnished with a comprehensive parent information packet (comprising the Free and Informed Consent Form, the Free and Informed Assent Form, and the Consent Form for photographs, videos, and recordings), elucidating details about the trial and intervention. They retain the prerogative to withdraw their child from the trial at any juncture and for any reason, without impeding the child's ongoing treatment. This trial does not involve collecting biological specimens for storage. Explanation for the choice of comparators {6b} The chosen comparison variables were pain, ROM function, and main points of physiotherapeutic treatment, with the variable peripheral muscle oxygenation, a distinguishing feature of this study. Intervention description {11a} After the initial evaluations, randomization will be performed for the distribution of individuals in the following groups: (1) face-to-face group (GF) and (2) telerehabilitation group (GT). Initially, an evaluator will categorize participants into the eligibility criteria by collecting information related to personal data, mass, height, body mass index, medical history, previous illnesses, concomitant illnesses, use of interocclusal devices, medications, previous surgeries, physiotherapy, and other aspects. After screening, acceptance, and signing of ethical terms, anamnesis will begin with participant identification data, demographics, anthropometric assessment, and assessment using the DC/TMD to confirm the diagnosis. This multimodal rehabilitation protocol used guidance resources, manual therapy, and exercises, and are in Table , which were considered low-cost resources based on the literature. We sought to simplify them to allow self-management. The protocol will be applied online and in-person. The choice of 3 sessions was based on the satisfactory results of recent RCTs, which carried out a few (2 to 5) physiotherapy sessions in the treatment of musculoskeletal disorders of the head and neck region . Treatment interventions were standardized to improve the internal validity of the design and allow for ease of replication in future clinical trials or by professionals in their work environments. The GT intervention will take place synchronously remotely via video call via WhatsApp® and as can be seen in Table , the GT protocol will be similar to the GF, except for two basic differences in manual therapy and articular manipulation. Counseling The guidelines in this protocol include care, advice, pain education, and stress management strategies, which will be clarified at the beginning of each session, so that participants can learn and incorporate them into their daily lives. This promotion of autonomy for the patient and their responsibility towards the treatment, promotes a more powerful result with a change in the general perception of pain and a reduction in recurrences . To make it easier for participants and examiners to remember each orientation, the author created the acronym “TEPEDI,” which refers to the initials of thermotherapy, exercises, posture, explanation, lying down, and the importance of quality of life. “T” for thermotherapy, which can be carried out at home with a cold or heat pack, to reduce pain or relax muscles . “E” home exercises, which must be repeated every day, as taught in the sessions and verified for their correct and satisfactory execution . The GF and GT participants will be instructed to repeat the exercises, self-massage, and all instructions at home daily. “P” for correct posture in resting the tongue, jaw, head, and shoulders,the tongue should be on the hard palate and the jaw relaxed , the head aligned with the dorsal column with the shoulders lowered . “E” for explanation, the patient who receives an explanation and understands the use of TMJ and the precipitating and predisposing factors of TMD, may have a better prognosis and adherence to treatment . “D” for decreased parafunction and unclenched teeth from contact, in order to manage parafunctional habits, which can be remembered and made aware through post-it notes or through the “Desencoste” app. Participants will be informed that when the jaw is at rest, their teeth should not touch, except when swallowing or eating . “I” to remember the importance of basic habits in the quality of life, such as sleep hygiene , the practice of physical activity , and relaxation. This relaxation can be done in several ways and in this protocol, diaphragmatic breathing was used, with re-education of the breathing pattern, stimulation of the diaphragm, and relaxation of the accessory respiratory muscles . Manual therapy The manual approaches to this protocol include passive myofascial release (in GF) or self-massage (in GT), and it is demonstrated in Fig. , with guidance that this pressure is gentle and comfortable manner depending on individual tolerance. A previous study showed that myofascial release applied for 10 min to the trapezius muscle was effective in the hemodynamic variable of the total saturation index (TSI), which reinforces the choice of technique and application time chosen in this RCT. The participants in both groups will still be instructed to add self-massage to their daily self-care tasks. To facilitate the checklist for each region to apply MT, the acronym “MAN” was created, which refers to the initials of the regions: masticatory muscles, articulation TMJ, and neck. Masticatory muscles: Manual therapy of the masticatory muscles in the intraoral and extraoral regions . It will be performed in the GF and GT groups. In the IO maneuver, the patient will be instructed to place the contralateral thumb on the mandible that will receive the maneuver, and the other fingers must rest on the external region of this cheek. Attention will be placed on some details upon carrying out the maneuver, including the use of gloves in the GF, hygiene of hands, and whether the patient is free from internal injuries or limitations that cause discomfort. TMJ: Non-specific joint mobilization of the TMJ, with rhythmic, oscillatory movement, in the posteroanterior direction, for 1 min 30 s, 3 to 5 times, with an interval of 30 s between series . This maneuver will only be performed in the GF. Neck muscles: Manual therapy in the posterior and anterior region of the muscles neck (scalene, sternocleidomastoid, trapezius, elevator scapula, suboccipital, supra, and infrahyoid muscles) . This maneuver will be performed in both groups. Therapeutic exercises Patients will be informed of the objective of each exercise, the need to communicate in the event of pain, caution in each performance, and pauses to avoid muscle fatigue . The regions that will receive the exercises will be the same as those of the TM,therefore, the same acronym “MAN” was used for the checklist of the “masticatory” muscles, TMJ “articulation”, and “neck” muscles. The use of a mirror will be recommended in favor of awareness in movement, correcting deviations, and excessive amplitude, minimizing joint noises; thus, which facilitates learning, since the exercises will be advised to be repeated at home every day . A 30-s rest interval will be recommended between each exercise. The exercises aim to achieve muscle relaxation and optimize function. The jaw exercises will be isotonic and isometric and are demonstrated in Fig. : isotonic exercises in a free and maximum range of motion for all mandibular ranges (lateralization, opening, closing, protrusion, with 6 repetitions of each movement) . Strengthening isometrics, with small resistance given by the fingers, and tongue on the palate, for all mandibular ranges (lateralization, opening, closing, protrusion, with 6 repetitions of each movement) . The TMJ exercise chosen was the “N position” with opening and protrusion of the mouth with the tongue on the palate (3 sets of 10 repetitions, 30-s intervals) . The exercises for the neck region are demonstrated in Fig. , and it will be as follows: stretching in all planes of movement (extension, flexion, rotation and inclination right and left, inclination with flexion, six repetitions of each movement), self-growth exercise correcting posture and seeking alignment (for 1 min) and an anterior skull rotation exercise by nodding the head (six times out of six repetitions) . Criteria for discontinuing interventions {11b} The criteria for halting interventions for a trial participant encompass adolescents who request a temporary suspension for personal reasons or whose deteriorating condition prompts them to withdraw from active participation. Strategies to improve adherence to interventions {11c} Two clinics will be available, in different locations, in the city of Florianópolis (Brazil), in order to facilitate movement and improve participant adherence. In-person assessments and consultations take place in these clinics according to the participant's choice. In addition, there will be flexibility in scheduling times, and patients will receive reminders and videos of exercises and care at home. Relevant concomitant care permitted or prohibited during the trial {11d} During this training, adolescents will not be able to undergo other treatments such as the use of an interocclusal splint, medications, or other physiotherapy services, other than those offered in this protocol, nor will they be able to start using orthodontic appliances. Provisions for post-trial care {30} As a form of care post-trial, all participants will receive a report with the results of the instruments used, as well as care guidelines and home exercises. Outcomes {12} Each group will undergo an initial assessment (T0), followed by three treatment sessions: an immediate reassessment (which will be carried out 0–2 days after the third session [T1]), another reassessment 30 days after the end of treatment, and treatment follow-up (T2). Assessments and reassessments will be performed in-person. In addition to the physical aspects of DC/TMD and anthropometric assessment through mass and height measurements, the following instruments will be used: digital pressure algometer, graduated chronic pain scale (GCPS), pain drawing for pain assessment, infrared spectroscopy (NIRS) to assess peripheral muscle oxygenation, Tam scale (TSK/TMD) to assess kinesiophobia, and other psychosocial scales of DC/TMD, such as Generalized Anxiety Disorder 7-item (GAD7), Patient Health Questionnaire-4 (PHQ4), Oral Behaviors Checklist (OBC), and Jaw Functional Limitation Scale-8 (JFLS8). The assessment will last an average of 60 min, and the treatment sessions will last 30 min , with 10 min of guidance, 10 min of manual therapy and relaxation, 10 min of specific exercises, and reassessments lasting 40 min. This 3-session protocol will occur weekly. In reassessments, the same assessment instruments will be applied, with the exception of the GAD7 and PHQ4, as these will only be used to characterize the sample. Finally, all participants will receive a report with the results of the instruments used, as well as care guidelines and home exercises. Diagnostic criteria of temporomandibular disorders—axis 1—physical assessment For the diagnosis of TMD and clinical evaluation of the TMJ, the validated Portuguese version of the DC/TMD will be used, which classifies between myalgia, arthralgia, headache attributed to TMD, disc displacement, degenerative joint disease, or subluxation. There may be more than one diagnosis for each participant. This stage lasts an average of 15 to 20 min and will follow the regulations of the Delphi study Axis 1 of the DC for adolescents . Range of motion (ROM) measurements will be taken with a digital universal caliper. Diagnostic criteria of temporomandibular disorders—axis 2—psychosocial assessment and pain Axis 2 of the DC/TMD allows for psychosocial assessment and determination of the consequences of pain. Some suggestions from the Delphi study for adolescents will be used, such as the GCPS, Pain Drawing, JFLS8, and OBC scales. Even though some questionnaires may be self-administered , it was decided to have one examiner (examiner “A” or “B”) accompany each participant, in order to accommodate doubts, check and reduce biases. Tampa scale for kinesiophobia for temporomandibular disorders The Tampa Kinesiophobia Scale for TMD (TSK/TMD) is a self-administered questionnaire with 18 questions, translated and cross-culturally adapted, and is valid and reliable for assessing kinesiophobia in patients with TMD . Near-infrared spectroscopy To obtain hemodynamic variables with high resolution in real time, such as oxyhemoglobin (HbO2), deoxyhemoglobin (HHb), total hemoglobin (tHb), tissue saturation index (TSI), and the near-infrared spectroscopy (NIRS) (Portamon®, Artinis, Netherlands), with an acquisition frequency of 10 Hz, in a non-invasive evaluation phase that lasts 5 min will be used . The chosen muscle, positioning, and method that will be used were based on an oximetry study in adolescents with TMD , which evaluated the masseter muscle bilaterally, as it is involved in chewing in TMDs. Oxygenation will be measured at two time points: with the muscle at rest (jaw relaxed and teeth disengaged) for 60 s, during muscle contraction (in dental occlusion), followed by the maximum voluntary isometric contraction of the masseter muscle for 20 s, and at moments T0, T1, and T2 of the ECR protocol. The side to start measuring the device was chosen randomly. For the purpose of dental protection, to contract the masseter, a parafilm (Pechinery® Plastic Packaging, USA) folded 15 times to a size of 1.5 cm by 3.5 cm will be used, which will be positioned between the occlusal surfaces of the first and second upper and lower molars. Pressure algometry To evaluate the pressure pain threshold (PPT), a digital pressure algometer from the brand MedEOR® model SP Tech, will be used, which allows for real measurement of pain thresholds and tolerances by mechanical pressure. These are two measurable and useful neurophysiological methods for clinical practice, which assess pain in an objective manner (MedEOR® Medtech LTDA, Brazil, 2018). The position the participant will adopt, as well as the handling of the algometer, was based on a previous study , which evaluated the masseter and temporalis muscles bilaterally. The side to start measuring the device was chosen randomly. Pressure was applied until the volunteer complained of pain, indicated by raising the arm, and the value was recorded on an algometer display (kg/cm 2 ). The PPT was measured three times at each location, with an interval of 5 s between each measurement, and the average was used for statistical analysis. Table summarizes the instruments used, their scores, evaluations, and their respective classifications. The Clinical Relevance of this study is that a multimodal face-to-face and telerehabilitation protocol for adolescents with TMD will be presented, and this protocol could serve as a basis for future research in this area and it can be easily used for clinical practice. Participant timeline {13} The participant’s timeline can be observed through the flowchart in Fig. . Sample size {14} For the sample calculation, the G Power program version 3.1.7 was used based on oxyhemoglobin data in adolescents with TMD , which represents the same population of this trial, and also, the authors used the same NIRS equipment,thus, the physiological variable, peripheral muscle oxygenation, was chosen as the primary outcome to calculate the sample size. The final sample size was 26 adolescents, with a significance level of 5% and power of 80%, the same significance and values used by a previous study . The 26 adolescents will be randomized into two groups with 13 participants in each group. Recruitment {15} Patients will be recruited from communities located in the metropolitan area of Florianopolis City, south of Brazil, by means of announcements, pamphlets, posters, emails, social media, and invitations through schools. The chosen comparison variables were pain, ROM function, and main points of physiotherapeutic treatment, with the variable peripheral muscle oxygenation, a distinguishing feature of this study. After the initial evaluations, randomization will be performed for the distribution of individuals in the following groups: (1) face-to-face group (GF) and (2) telerehabilitation group (GT). Initially, an evaluator will categorize participants into the eligibility criteria by collecting information related to personal data, mass, height, body mass index, medical history, previous illnesses, concomitant illnesses, use of interocclusal devices, medications, previous surgeries, physiotherapy, and other aspects. After screening, acceptance, and signing of ethical terms, anamnesis will begin with participant identification data, demographics, anthropometric assessment, and assessment using the DC/TMD to confirm the diagnosis. This multimodal rehabilitation protocol used guidance resources, manual therapy, and exercises, and are in Table , which were considered low-cost resources based on the literature. We sought to simplify them to allow self-management. The protocol will be applied online and in-person. The choice of 3 sessions was based on the satisfactory results of recent RCTs, which carried out a few (2 to 5) physiotherapy sessions in the treatment of musculoskeletal disorders of the head and neck region . Treatment interventions were standardized to improve the internal validity of the design and allow for ease of replication in future clinical trials or by professionals in their work environments. The GT intervention will take place synchronously remotely via video call via WhatsApp® and as can be seen in Table , the GT protocol will be similar to the GF, except for two basic differences in manual therapy and articular manipulation. Counseling The guidelines in this protocol include care, advice, pain education, and stress management strategies, which will be clarified at the beginning of each session, so that participants can learn and incorporate them into their daily lives. This promotion of autonomy for the patient and their responsibility towards the treatment, promotes a more powerful result with a change in the general perception of pain and a reduction in recurrences . To make it easier for participants and examiners to remember each orientation, the author created the acronym “TEPEDI,” which refers to the initials of thermotherapy, exercises, posture, explanation, lying down, and the importance of quality of life. “T” for thermotherapy, which can be carried out at home with a cold or heat pack, to reduce pain or relax muscles . “E” home exercises, which must be repeated every day, as taught in the sessions and verified for their correct and satisfactory execution . The GF and GT participants will be instructed to repeat the exercises, self-massage, and all instructions at home daily. “P” for correct posture in resting the tongue, jaw, head, and shoulders,the tongue should be on the hard palate and the jaw relaxed , the head aligned with the dorsal column with the shoulders lowered . “E” for explanation, the patient who receives an explanation and understands the use of TMJ and the precipitating and predisposing factors of TMD, may have a better prognosis and adherence to treatment . “D” for decreased parafunction and unclenched teeth from contact, in order to manage parafunctional habits, which can be remembered and made aware through post-it notes or through the “Desencoste” app. Participants will be informed that when the jaw is at rest, their teeth should not touch, except when swallowing or eating . “I” to remember the importance of basic habits in the quality of life, such as sleep hygiene , the practice of physical activity , and relaxation. This relaxation can be done in several ways and in this protocol, diaphragmatic breathing was used, with re-education of the breathing pattern, stimulation of the diaphragm, and relaxation of the accessory respiratory muscles . Manual therapy The manual approaches to this protocol include passive myofascial release (in GF) or self-massage (in GT), and it is demonstrated in Fig. , with guidance that this pressure is gentle and comfortable manner depending on individual tolerance. A previous study showed that myofascial release applied for 10 min to the trapezius muscle was effective in the hemodynamic variable of the total saturation index (TSI), which reinforces the choice of technique and application time chosen in this RCT. The participants in both groups will still be instructed to add self-massage to their daily self-care tasks. To facilitate the checklist for each region to apply MT, the acronym “MAN” was created, which refers to the initials of the regions: masticatory muscles, articulation TMJ, and neck. Masticatory muscles: Manual therapy of the masticatory muscles in the intraoral and extraoral regions . It will be performed in the GF and GT groups. In the IO maneuver, the patient will be instructed to place the contralateral thumb on the mandible that will receive the maneuver, and the other fingers must rest on the external region of this cheek. Attention will be placed on some details upon carrying out the maneuver, including the use of gloves in the GF, hygiene of hands, and whether the patient is free from internal injuries or limitations that cause discomfort. TMJ: Non-specific joint mobilization of the TMJ, with rhythmic, oscillatory movement, in the posteroanterior direction, for 1 min 30 s, 3 to 5 times, with an interval of 30 s between series . This maneuver will only be performed in the GF. Neck muscles: Manual therapy in the posterior and anterior region of the muscles neck (scalene, sternocleidomastoid, trapezius, elevator scapula, suboccipital, supra, and infrahyoid muscles) . This maneuver will be performed in both groups. Therapeutic exercises Patients will be informed of the objective of each exercise, the need to communicate in the event of pain, caution in each performance, and pauses to avoid muscle fatigue . The regions that will receive the exercises will be the same as those of the TM,therefore, the same acronym “MAN” was used for the checklist of the “masticatory” muscles, TMJ “articulation”, and “neck” muscles. The use of a mirror will be recommended in favor of awareness in movement, correcting deviations, and excessive amplitude, minimizing joint noises; thus, which facilitates learning, since the exercises will be advised to be repeated at home every day . A 30-s rest interval will be recommended between each exercise. The exercises aim to achieve muscle relaxation and optimize function. The jaw exercises will be isotonic and isometric and are demonstrated in Fig. : isotonic exercises in a free and maximum range of motion for all mandibular ranges (lateralization, opening, closing, protrusion, with 6 repetitions of each movement) . Strengthening isometrics, with small resistance given by the fingers, and tongue on the palate, for all mandibular ranges (lateralization, opening, closing, protrusion, with 6 repetitions of each movement) . The TMJ exercise chosen was the “N position” with opening and protrusion of the mouth with the tongue on the palate (3 sets of 10 repetitions, 30-s intervals) . The exercises for the neck region are demonstrated in Fig. , and it will be as follows: stretching in all planes of movement (extension, flexion, rotation and inclination right and left, inclination with flexion, six repetitions of each movement), self-growth exercise correcting posture and seeking alignment (for 1 min) and an anterior skull rotation exercise by nodding the head (six times out of six repetitions) . The guidelines in this protocol include care, advice, pain education, and stress management strategies, which will be clarified at the beginning of each session, so that participants can learn and incorporate them into their daily lives. This promotion of autonomy for the patient and their responsibility towards the treatment, promotes a more powerful result with a change in the general perception of pain and a reduction in recurrences . To make it easier for participants and examiners to remember each orientation, the author created the acronym “TEPEDI,” which refers to the initials of thermotherapy, exercises, posture, explanation, lying down, and the importance of quality of life. “T” for thermotherapy, which can be carried out at home with a cold or heat pack, to reduce pain or relax muscles . “E” home exercises, which must be repeated every day, as taught in the sessions and verified for their correct and satisfactory execution . The GF and GT participants will be instructed to repeat the exercises, self-massage, and all instructions at home daily. “P” for correct posture in resting the tongue, jaw, head, and shoulders,the tongue should be on the hard palate and the jaw relaxed , the head aligned with the dorsal column with the shoulders lowered . “E” for explanation, the patient who receives an explanation and understands the use of TMJ and the precipitating and predisposing factors of TMD, may have a better prognosis and adherence to treatment . “D” for decreased parafunction and unclenched teeth from contact, in order to manage parafunctional habits, which can be remembered and made aware through post-it notes or through the “Desencoste” app. Participants will be informed that when the jaw is at rest, their teeth should not touch, except when swallowing or eating . “I” to remember the importance of basic habits in the quality of life, such as sleep hygiene , the practice of physical activity , and relaxation. This relaxation can be done in several ways and in this protocol, diaphragmatic breathing was used, with re-education of the breathing pattern, stimulation of the diaphragm, and relaxation of the accessory respiratory muscles . The manual approaches to this protocol include passive myofascial release (in GF) or self-massage (in GT), and it is demonstrated in Fig. , with guidance that this pressure is gentle and comfortable manner depending on individual tolerance. A previous study showed that myofascial release applied for 10 min to the trapezius muscle was effective in the hemodynamic variable of the total saturation index (TSI), which reinforces the choice of technique and application time chosen in this RCT. The participants in both groups will still be instructed to add self-massage to their daily self-care tasks. To facilitate the checklist for each region to apply MT, the acronym “MAN” was created, which refers to the initials of the regions: masticatory muscles, articulation TMJ, and neck. Masticatory muscles: Manual therapy of the masticatory muscles in the intraoral and extraoral regions . It will be performed in the GF and GT groups. In the IO maneuver, the patient will be instructed to place the contralateral thumb on the mandible that will receive the maneuver, and the other fingers must rest on the external region of this cheek. Attention will be placed on some details upon carrying out the maneuver, including the use of gloves in the GF, hygiene of hands, and whether the patient is free from internal injuries or limitations that cause discomfort. TMJ: Non-specific joint mobilization of the TMJ, with rhythmic, oscillatory movement, in the posteroanterior direction, for 1 min 30 s, 3 to 5 times, with an interval of 30 s between series . This maneuver will only be performed in the GF. Neck muscles: Manual therapy in the posterior and anterior region of the muscles neck (scalene, sternocleidomastoid, trapezius, elevator scapula, suboccipital, supra, and infrahyoid muscles) . This maneuver will be performed in both groups. Patients will be informed of the objective of each exercise, the need to communicate in the event of pain, caution in each performance, and pauses to avoid muscle fatigue . The regions that will receive the exercises will be the same as those of the TM,therefore, the same acronym “MAN” was used for the checklist of the “masticatory” muscles, TMJ “articulation”, and “neck” muscles. The use of a mirror will be recommended in favor of awareness in movement, correcting deviations, and excessive amplitude, minimizing joint noises; thus, which facilitates learning, since the exercises will be advised to be repeated at home every day . A 30-s rest interval will be recommended between each exercise. The exercises aim to achieve muscle relaxation and optimize function. The jaw exercises will be isotonic and isometric and are demonstrated in Fig. : isotonic exercises in a free and maximum range of motion for all mandibular ranges (lateralization, opening, closing, protrusion, with 6 repetitions of each movement) . Strengthening isometrics, with small resistance given by the fingers, and tongue on the palate, for all mandibular ranges (lateralization, opening, closing, protrusion, with 6 repetitions of each movement) . The TMJ exercise chosen was the “N position” with opening and protrusion of the mouth with the tongue on the palate (3 sets of 10 repetitions, 30-s intervals) . The exercises for the neck region are demonstrated in Fig. , and it will be as follows: stretching in all planes of movement (extension, flexion, rotation and inclination right and left, inclination with flexion, six repetitions of each movement), self-growth exercise correcting posture and seeking alignment (for 1 min) and an anterior skull rotation exercise by nodding the head (six times out of six repetitions) . The criteria for halting interventions for a trial participant encompass adolescents who request a temporary suspension for personal reasons or whose deteriorating condition prompts them to withdraw from active participation. Two clinics will be available, in different locations, in the city of Florianópolis (Brazil), in order to facilitate movement and improve participant adherence. In-person assessments and consultations take place in these clinics according to the participant's choice. In addition, there will be flexibility in scheduling times, and patients will receive reminders and videos of exercises and care at home. During this training, adolescents will not be able to undergo other treatments such as the use of an interocclusal splint, medications, or other physiotherapy services, other than those offered in this protocol, nor will they be able to start using orthodontic appliances. As a form of care post-trial, all participants will receive a report with the results of the instruments used, as well as care guidelines and home exercises. Each group will undergo an initial assessment (T0), followed by three treatment sessions: an immediate reassessment (which will be carried out 0–2 days after the third session [T1]), another reassessment 30 days after the end of treatment, and treatment follow-up (T2). Assessments and reassessments will be performed in-person. In addition to the physical aspects of DC/TMD and anthropometric assessment through mass and height measurements, the following instruments will be used: digital pressure algometer, graduated chronic pain scale (GCPS), pain drawing for pain assessment, infrared spectroscopy (NIRS) to assess peripheral muscle oxygenation, Tam scale (TSK/TMD) to assess kinesiophobia, and other psychosocial scales of DC/TMD, such as Generalized Anxiety Disorder 7-item (GAD7), Patient Health Questionnaire-4 (PHQ4), Oral Behaviors Checklist (OBC), and Jaw Functional Limitation Scale-8 (JFLS8). The assessment will last an average of 60 min, and the treatment sessions will last 30 min , with 10 min of guidance, 10 min of manual therapy and relaxation, 10 min of specific exercises, and reassessments lasting 40 min. This 3-session protocol will occur weekly. In reassessments, the same assessment instruments will be applied, with the exception of the GAD7 and PHQ4, as these will only be used to characterize the sample. Finally, all participants will receive a report with the results of the instruments used, as well as care guidelines and home exercises. Diagnostic criteria of temporomandibular disorders—axis 1—physical assessment For the diagnosis of TMD and clinical evaluation of the TMJ, the validated Portuguese version of the DC/TMD will be used, which classifies between myalgia, arthralgia, headache attributed to TMD, disc displacement, degenerative joint disease, or subluxation. There may be more than one diagnosis for each participant. This stage lasts an average of 15 to 20 min and will follow the regulations of the Delphi study Axis 1 of the DC for adolescents . Range of motion (ROM) measurements will be taken with a digital universal caliper. Diagnostic criteria of temporomandibular disorders—axis 2—psychosocial assessment and pain Axis 2 of the DC/TMD allows for psychosocial assessment and determination of the consequences of pain. Some suggestions from the Delphi study for adolescents will be used, such as the GCPS, Pain Drawing, JFLS8, and OBC scales. Even though some questionnaires may be self-administered , it was decided to have one examiner (examiner “A” or “B”) accompany each participant, in order to accommodate doubts, check and reduce biases. Tampa scale for kinesiophobia for temporomandibular disorders The Tampa Kinesiophobia Scale for TMD (TSK/TMD) is a self-administered questionnaire with 18 questions, translated and cross-culturally adapted, and is valid and reliable for assessing kinesiophobia in patients with TMD . Near-infrared spectroscopy To obtain hemodynamic variables with high resolution in real time, such as oxyhemoglobin (HbO2), deoxyhemoglobin (HHb), total hemoglobin (tHb), tissue saturation index (TSI), and the near-infrared spectroscopy (NIRS) (Portamon®, Artinis, Netherlands), with an acquisition frequency of 10 Hz, in a non-invasive evaluation phase that lasts 5 min will be used . The chosen muscle, positioning, and method that will be used were based on an oximetry study in adolescents with TMD , which evaluated the masseter muscle bilaterally, as it is involved in chewing in TMDs. Oxygenation will be measured at two time points: with the muscle at rest (jaw relaxed and teeth disengaged) for 60 s, during muscle contraction (in dental occlusion), followed by the maximum voluntary isometric contraction of the masseter muscle for 20 s, and at moments T0, T1, and T2 of the ECR protocol. The side to start measuring the device was chosen randomly. For the purpose of dental protection, to contract the masseter, a parafilm (Pechinery® Plastic Packaging, USA) folded 15 times to a size of 1.5 cm by 3.5 cm will be used, which will be positioned between the occlusal surfaces of the first and second upper and lower molars. Pressure algometry To evaluate the pressure pain threshold (PPT), a digital pressure algometer from the brand MedEOR® model SP Tech, will be used, which allows for real measurement of pain thresholds and tolerances by mechanical pressure. These are two measurable and useful neurophysiological methods for clinical practice, which assess pain in an objective manner (MedEOR® Medtech LTDA, Brazil, 2018). The position the participant will adopt, as well as the handling of the algometer, was based on a previous study , which evaluated the masseter and temporalis muscles bilaterally. The side to start measuring the device was chosen randomly. Pressure was applied until the volunteer complained of pain, indicated by raising the arm, and the value was recorded on an algometer display (kg/cm 2 ). The PPT was measured three times at each location, with an interval of 5 s between each measurement, and the average was used for statistical analysis. Table summarizes the instruments used, their scores, evaluations, and their respective classifications. The Clinical Relevance of this study is that a multimodal face-to-face and telerehabilitation protocol for adolescents with TMD will be presented, and this protocol could serve as a basis for future research in this area and it can be easily used for clinical practice. For the diagnosis of TMD and clinical evaluation of the TMJ, the validated Portuguese version of the DC/TMD will be used, which classifies between myalgia, arthralgia, headache attributed to TMD, disc displacement, degenerative joint disease, or subluxation. There may be more than one diagnosis for each participant. This stage lasts an average of 15 to 20 min and will follow the regulations of the Delphi study Axis 1 of the DC for adolescents . Range of motion (ROM) measurements will be taken with a digital universal caliper. Axis 2 of the DC/TMD allows for psychosocial assessment and determination of the consequences of pain. Some suggestions from the Delphi study for adolescents will be used, such as the GCPS, Pain Drawing, JFLS8, and OBC scales. Even though some questionnaires may be self-administered , it was decided to have one examiner (examiner “A” or “B”) accompany each participant, in order to accommodate doubts, check and reduce biases. The Tampa Kinesiophobia Scale for TMD (TSK/TMD) is a self-administered questionnaire with 18 questions, translated and cross-culturally adapted, and is valid and reliable for assessing kinesiophobia in patients with TMD . To obtain hemodynamic variables with high resolution in real time, such as oxyhemoglobin (HbO2), deoxyhemoglobin (HHb), total hemoglobin (tHb), tissue saturation index (TSI), and the near-infrared spectroscopy (NIRS) (Portamon®, Artinis, Netherlands), with an acquisition frequency of 10 Hz, in a non-invasive evaluation phase that lasts 5 min will be used . The chosen muscle, positioning, and method that will be used were based on an oximetry study in adolescents with TMD , which evaluated the masseter muscle bilaterally, as it is involved in chewing in TMDs. Oxygenation will be measured at two time points: with the muscle at rest (jaw relaxed and teeth disengaged) for 60 s, during muscle contraction (in dental occlusion), followed by the maximum voluntary isometric contraction of the masseter muscle for 20 s, and at moments T0, T1, and T2 of the ECR protocol. The side to start measuring the device was chosen randomly. For the purpose of dental protection, to contract the masseter, a parafilm (Pechinery® Plastic Packaging, USA) folded 15 times to a size of 1.5 cm by 3.5 cm will be used, which will be positioned between the occlusal surfaces of the first and second upper and lower molars. To evaluate the pressure pain threshold (PPT), a digital pressure algometer from the brand MedEOR® model SP Tech, will be used, which allows for real measurement of pain thresholds and tolerances by mechanical pressure. These are two measurable and useful neurophysiological methods for clinical practice, which assess pain in an objective manner (MedEOR® Medtech LTDA, Brazil, 2018). The position the participant will adopt, as well as the handling of the algometer, was based on a previous study , which evaluated the masseter and temporalis muscles bilaterally. The side to start measuring the device was chosen randomly. Pressure was applied until the volunteer complained of pain, indicated by raising the arm, and the value was recorded on an algometer display (kg/cm 2 ). The PPT was measured three times at each location, with an interval of 5 s between each measurement, and the average was used for statistical analysis. Table summarizes the instruments used, their scores, evaluations, and their respective classifications. The Clinical Relevance of this study is that a multimodal face-to-face and telerehabilitation protocol for adolescents with TMD will be presented, and this protocol could serve as a basis for future research in this area and it can be easily used for clinical practice. The participant’s timeline can be observed through the flowchart in Fig. . For the sample calculation, the G Power program version 3.1.7 was used based on oxyhemoglobin data in adolescents with TMD , which represents the same population of this trial, and also, the authors used the same NIRS equipment,thus, the physiological variable, peripheral muscle oxygenation, was chosen as the primary outcome to calculate the sample size. The final sample size was 26 adolescents, with a significance level of 5% and power of 80%, the same significance and values used by a previous study . The 26 adolescents will be randomized into two groups with 13 participants in each group. Patients will be recruited from communities located in the metropolitan area of Florianopolis City, south of Brazil, by means of announcements, pamphlets, posters, emails, social media, and invitations through schools. Sequence generation {16a} A random sequence of participants distributed proportionally in each study group will be performed using the website, randomization.com. Concealment mechanism {16b} The concealed allocation of individuals will be accomplished using sealed and opaque envelopes. Randomization and concealed allocation will be performed by an independent researcher, who will not be involved in the recruitment, evaluation, or intervention processes. Implementation {16c} Examiner X, which will not be involved in the recruitment, evaluation, or intervention processes, will allocate the participants, and collect the signatures of the ethical terms. The interventions will be made by examiners A and B. All stages of the DC/TMD axis 1, algometry and oximetry, will be carried out by the same examiner (examiner “A”), trained and experienced in the area of TMD for 20 years, with the intention of reducing biases. The other questionnaires will be administered by examiner “A” and examiner “B,” also trained and qualified. All GF sessions will be carried out by the same physiotherapist and an experienced clinician in the area (examiner A). The GT sessions will be conducted by examiners A and B. The groups (GT and GF) will receive similar care; however, in the manual therapy stage (TM), the GT group will be guided to perform self-massage, while the GF group will be massaged by the physiotherapist. A random sequence of participants distributed proportionally in each study group will be performed using the website, randomization.com. The concealed allocation of individuals will be accomplished using sealed and opaque envelopes. Randomization and concealed allocation will be performed by an independent researcher, who will not be involved in the recruitment, evaluation, or intervention processes. Examiner X, which will not be involved in the recruitment, evaluation, or intervention processes, will allocate the participants, and collect the signatures of the ethical terms. The interventions will be made by examiners A and B. All stages of the DC/TMD axis 1, algometry and oximetry, will be carried out by the same examiner (examiner “A”), trained and experienced in the area of TMD for 20 years, with the intention of reducing biases. The other questionnaires will be administered by examiner “A” and examiner “B,” also trained and qualified. All GF sessions will be carried out by the same physiotherapist and an experienced clinician in the area (examiner A). The GT sessions will be conducted by examiners A and B. The groups (GT and GF) will receive similar care; however, in the manual therapy stage (TM), the GT group will be guided to perform self-massage, while the GF group will be massaged by the physiotherapist. Who will be blinded {17a} This protocol will be blinding only for statistical analysis, by examiner “C.” Given the logistical and regulatory challenges in implementing randomization and intervention, examiners X, A, and B will not be feasible. Procedure for unblinding if needed {17b} The design of the study is open-label, with only the outcome assessors being blinded, thus preventing unblinding. This protocol will be blinding only for statistical analysis, by examiner “C.” Given the logistical and regulatory challenges in implementing randomization and intervention, examiners X, A, and B will not be feasible. The design of the study is open-label, with only the outcome assessors being blinded, thus preventing unblinding. Plans for assessment and collection of outcomes {18a} Participant data will be collected by researchers and stored anonymously in Google Drive. The data will be stored in accordance with the country’s Data Protection guidelines. Examiner A received training for face-to-face assessment and treatment and has experience in the area for over 20 years. A database was included in the study folder accessible to all researchers. There will be several moments for data entry: screening, initial assessment, immediate reassessment, and follow-up. Plans to promote participant retention and complete follow-up {18b} Patients will receive confirmation and reminders of the scheduled treatment date, and this will facilitate reaching the calculated sample size, as well as careful reminders of the advice they should take. Data management {19} The Google Drive platform will be used for data management and storage of study data (in a database), including data backup. The data will be validated according to the data validation plan. After carrying out all data validation and the final review, the study database will be considered complete and the data it contains is reliable. As soon as the study is concluded, the study database will be closed and transferred to the blinded Team of biostatisticians for data analysis, with codes to avoid the groups’ identification. At the end of the study, a copy of the site-specific document. Records will be provided to each principal investigator. Data will be validated according to the data validation plan. Confidentiality {27} The study staff will ensure that the participant´s anonymity is maintained. The participants will be identified only by a participant code on the case report forms (eCRF) and any electronic database. All documents will be stored securely and only accessible by study staff and authorized personnel. Applicable regulations for storage, transmittal, and disclosure of patient information will be followed at all times. The study will comply with the Data Protection Legislation in each country. Following formal admission to the study, patient data will be recorded in the Clinic case record in the usual way including the circumstances of their entry to the study. Additionally, data will be held in eCRF. These files will be identified by a study code, date of birth, and participant code only. Representatives from the Sponsor and from the regulatory authorities will be given access to the records that relate to the study. They will have full access to the anonymous eCRFs for the purposes of data validation. The results of the study may be communicated at scientific meetings and will contribute to the scientific literature. At no time, this will be done in such a way that an individual participant may be identified. Plans for collection, laboratory evaluation, and storage of biological specimens for genetic or molecular analysis in this trial/future use {33} See above 26b; there will be no biological specimens collected. Participant data will be collected by researchers and stored anonymously in Google Drive. The data will be stored in accordance with the country’s Data Protection guidelines. Examiner A received training for face-to-face assessment and treatment and has experience in the area for over 20 years. A database was included in the study folder accessible to all researchers. There will be several moments for data entry: screening, initial assessment, immediate reassessment, and follow-up. Patients will receive confirmation and reminders of the scheduled treatment date, and this will facilitate reaching the calculated sample size, as well as careful reminders of the advice they should take. The Google Drive platform will be used for data management and storage of study data (in a database), including data backup. The data will be validated according to the data validation plan. After carrying out all data validation and the final review, the study database will be considered complete and the data it contains is reliable. As soon as the study is concluded, the study database will be closed and transferred to the blinded Team of biostatisticians for data analysis, with codes to avoid the groups’ identification. At the end of the study, a copy of the site-specific document. Records will be provided to each principal investigator. Data will be validated according to the data validation plan. The study staff will ensure that the participant´s anonymity is maintained. The participants will be identified only by a participant code on the case report forms (eCRF) and any electronic database. All documents will be stored securely and only accessible by study staff and authorized personnel. Applicable regulations for storage, transmittal, and disclosure of patient information will be followed at all times. The study will comply with the Data Protection Legislation in each country. Following formal admission to the study, patient data will be recorded in the Clinic case record in the usual way including the circumstances of their entry to the study. Additionally, data will be held in eCRF. These files will be identified by a study code, date of birth, and participant code only. Representatives from the Sponsor and from the regulatory authorities will be given access to the records that relate to the study. They will have full access to the anonymous eCRFs for the purposes of data validation. The results of the study may be communicated at scientific meetings and will contribute to the scientific literature. At no time, this will be done in such a way that an individual participant may be identified. See above 26b; there will be no biological specimens collected. Statistical methods for primary and secondary outcomes {20a} The primary outcomes will be masseter oxygenation, pain, and range of motion. The secondary outcomes will be parafunction and kinesiophobia. Descriptive analysis will be performed with the calculation of mean and standard deviation; minimum, maximum, and median values for quantitative variables; and frequencies and percentages for categorized variables, stratified by group at baseline. Statistical analysis will be conducted based on an intention-to-treat analysis. Thus, individuals will be analyzed in the groups to which they were randomly allocated. Histograms will be used to verify the data distribution. Baseline means comparisons will be made using the Student’s t -test for asymmetric data. Mean questionnaire scores will be compared using a Poisson distribution fit. Associations between the categorized variables and groups will be assessed using the chi-square test. The evaluation of group means and evaluations will be performed using a repeated-measures model to test the interaction between groups and evaluations. For symmetric data, ANOVA will be used, followed by Tukey's multiple comparison test. The scores will be evaluated by fitting a Poisson distribution model, followed by Wald’s multiple comparison test. Pearson’s correlations between the pain and oximetry variables will be obtained. Statistical significance was set at p -value 5%. All analyses will be performed using SAS for Windows, v.9.4. The outcomes will also be compared by analyzing the minimum important clinical differences in relation to pain intensity, range of movement, and pressure pain threshold . Interim analyses {21b} No detrimental problems to the study participants are anticipated. No interim analyses are planned. Methods for additional analyses (e.g., subgroup analyses) {20b} Important prognostic variables are in the complex diagnoses. Further subgroup analysis will be conducted in line with the primary analysis. For continuous endpoints, similar modeling strategies will be used, but instead of logistic regression linear regression models will be used. External statistical assistance was hired, which had no contact with the study data collection. Methods in analysis to handle protocol non-adherence and any statistical methods to handle missing data {20c} Missing data in Clinical Trials, sensitivity analyses will be conducted with diverse strategies for handling missing data, incorporating adjustments for clustering. This will encompass considering missing data scenarios such as completely at random, missing at random, and missing not at random. Plans to give access to the full protocol, participant-level data, and statistical code {31c} The datasets analyzed during the current study and statistical code are available from the corresponding author on reasonable request, as is the full protocol. The primary outcomes will be masseter oxygenation, pain, and range of motion. The secondary outcomes will be parafunction and kinesiophobia. Descriptive analysis will be performed with the calculation of mean and standard deviation; minimum, maximum, and median values for quantitative variables; and frequencies and percentages for categorized variables, stratified by group at baseline. Statistical analysis will be conducted based on an intention-to-treat analysis. Thus, individuals will be analyzed in the groups to which they were randomly allocated. Histograms will be used to verify the data distribution. Baseline means comparisons will be made using the Student’s t -test for asymmetric data. Mean questionnaire scores will be compared using a Poisson distribution fit. Associations between the categorized variables and groups will be assessed using the chi-square test. The evaluation of group means and evaluations will be performed using a repeated-measures model to test the interaction between groups and evaluations. For symmetric data, ANOVA will be used, followed by Tukey's multiple comparison test. The scores will be evaluated by fitting a Poisson distribution model, followed by Wald’s multiple comparison test. Pearson’s correlations between the pain and oximetry variables will be obtained. Statistical significance was set at p -value 5%. All analyses will be performed using SAS for Windows, v.9.4. The outcomes will also be compared by analyzing the minimum important clinical differences in relation to pain intensity, range of movement, and pressure pain threshold . No detrimental problems to the study participants are anticipated. No interim analyses are planned. Important prognostic variables are in the complex diagnoses. Further subgroup analysis will be conducted in line with the primary analysis. For continuous endpoints, similar modeling strategies will be used, but instead of logistic regression linear regression models will be used. External statistical assistance was hired, which had no contact with the study data collection. Missing data in Clinical Trials, sensitivity analyses will be conducted with diverse strategies for handling missing data, incorporating adjustments for clustering. This will encompass considering missing data scenarios such as completely at random, missing at random, and missing not at random. The datasets analyzed during the current study and statistical code are available from the corresponding author on reasonable request, as is the full protocol. Composition of the coordinating center and trial steering committee {5d} In the trial, the School Clinic of Santa Catarina State University (UDESC), served as the coordinating center and trial steering committee. It comprised the chief investigator, technical coordinator, financial officer, and legal officer. Additionally, each participating center had its own chief investigator overseeing the trial’s day-to-day operations. Composition of the data monitoring committee, its role and reporting structure {21a} Due to the nature of the intervention, which does not concern a medical drug and does not propose extra risk to adolescents, the implementation of a Data Safety Monitoring Board is not deemed necessary. Adverse event reporting and harms {22} Given the intervention’s nature and the patient population’s potential for complications inherent to their biological or medical state, individual reporting of serious adverse events will not occur. Instead, they will be incorporated into the data collection process for assessing benefits and risks. Frequency and plans for auditing trial conduct {23} The electronic data entry system furnishes an audit trail, enabling identified and authorized users to remotely deposit data into the eCRFs. This ensures that all data entries and modifications made by sites in the central database are automatically and chronologically documented. Plans for communicating important protocol amendments to relevant parties (e.g., trial participants, ethical committees) {25} Our communication strategies for significant protocol modifications emphasize the prompt and direct distribution of information to all pertinent stakeholders. This encompasses investigators, Research Ethics Committees/Institutional Review Boards (REC/IRBs), and trial registries. Through streamlined communication channels, we will guarantee timely dissemination of updates, enabling stakeholders to remain informed and adapt or decide as needed. This is the third version of the protocol presented to the C Research Ethics Committee of the State University of Santa Catarina, to comply with Consubstantiated Opinion number 5,605,641 issued on October 10, 2022, CAAE 60845922.4.0000.0118. Dissemination plans {31a} The trial funding will undergo consideration for publication and presentation at scientific symposia or congresses. As participant data will be recorded anonymously, utmost care will be taken to ensure participant privacy. The results obtained from the trial will contribute to the enhancement of existing guidelines and enable the publication of new ones. Additionally, we will explore the possibility of conducting stakeholder workshops to engage with relevant groups and gather valuable perspectives. In the trial, the School Clinic of Santa Catarina State University (UDESC), served as the coordinating center and trial steering committee. It comprised the chief investigator, technical coordinator, financial officer, and legal officer. Additionally, each participating center had its own chief investigator overseeing the trial’s day-to-day operations. Due to the nature of the intervention, which does not concern a medical drug and does not propose extra risk to adolescents, the implementation of a Data Safety Monitoring Board is not deemed necessary. Given the intervention’s nature and the patient population’s potential for complications inherent to their biological or medical state, individual reporting of serious adverse events will not occur. Instead, they will be incorporated into the data collection process for assessing benefits and risks. The electronic data entry system furnishes an audit trail, enabling identified and authorized users to remotely deposit data into the eCRFs. This ensures that all data entries and modifications made by sites in the central database are automatically and chronologically documented. Our communication strategies for significant protocol modifications emphasize the prompt and direct distribution of information to all pertinent stakeholders. This encompasses investigators, Research Ethics Committees/Institutional Review Boards (REC/IRBs), and trial registries. Through streamlined communication channels, we will guarantee timely dissemination of updates, enabling stakeholders to remain informed and adapt or decide as needed. This is the third version of the protocol presented to the C Research Ethics Committee of the State University of Santa Catarina, to comply with Consubstantiated Opinion number 5,605,641 issued on October 10, 2022, CAAE 60845922.4.0000.0118. The trial funding will undergo consideration for publication and presentation at scientific symposia or congresses. As participant data will be recorded anonymously, utmost care will be taken to ensure participant privacy. The results obtained from the trial will contribute to the enhancement of existing guidelines and enable the publication of new ones. Additionally, we will explore the possibility of conducting stakeholder workshops to engage with relevant groups and gather valuable perspectives. This trial will verify and compare the effect of a multimodal intervention in the online TR and in the face-to-face format, in among adolescents with TMD, in terms of peripheral oxygenation of the masseter muscle, pain, range of motion, kinesiophobia, and parafunctional habits, highlighting the evaluation of the outcomes of the biopsychosocial sphere affected by patients with TMD. Furthermore, high internal validity is expected because this protocol was designed for randomization, concealed allocation, blinding for statistical analysis, intention-to-treat analysis, and adequate sample size with calculations considering the primary outcome. This study is a pioneer in comparing multimodal treatment among adolescents with TMD in face-to-face and TR formats, but it has some limitations. Owing to the nature of the proposed interventions, the physiotherapist responsible for implementing the treatment will not be blinded, nor will the participants know to which group they belong to. The need for an available Wi-Fi environment and network as well as the importance of engaging with families in teenagers’ participation can also be considered. The aim of this study was to contribute to technical-scientific advancement, supporting the clinical applicability of in-person and online multimodal rehabilitation based on evidence for the age group of adolescents with TMD, through a 3-session protocol, reducing comorbidities and costs of the disease, and generating better development and quality of life for this age group. This protocol could serve as a basis for future research in this area and has the advantage of being accessible to health professionals using tools easily found in clinical practice, such as biopsychosocial assessment, self-management, and pain education, which are part of the current scenario of evidence-based physiotherapy. The data will be published after the completion of the study. This manuscript is based on the trial protocol “Exercises applied online and multimodal intervention face to face: randomized controlled clinical trial in adolescents with Temporomandibular Dysfunction”. The project began in June 2022, and patient enrollment began in October 2022, and the recruitment began in November 2022. The study is in the main study phase. This study was registered in the Brazilian Registry of Clinical Trials (REBEC—RBR −5scd5tm, UTN: U1111-1288–4495), registered in May 2023. This is the third version of the protocol presented to the C Research Ethics Committee of the State University of Santa Catarina, to comply with Consubstantiated Opinion number 5,605,641 issued on October 10, 2022, CAAE 60845922.4.0000.0118. The entire study is scheduled to be completed by the end of December 2025. Supplementary Material 1. |
Metataxonomics study of dental bioaerosols affected by waterline disinfection | 06149275-1bd1-4008-82b6-967ec5e471e0 | 11689511 | Microbiology[mh] | Microorganisms found in aerosol-generating dental procedures are not only derived from a patient’s oral cavity and secretions but are primarily composed of microorganisms from dental unit waterlines (DUWLs) . Therefore, these procedures become potential routes for disease transmission, a consideration that dentists and dental personnel should integrate into their infection control practices. The Centers for Disease Control and Prevention (CDC) has underscored the importance of DUWL disinfection in response to outbreaks of nontuberculous Mycobacteria (NTM) infections among children who underwent pulpotomy in pediatric dental clinics where DUWL bacterial levels exceeded standard limits . Furthermore, the Guidelines for Infection Control in Dental Health-Care Settings have documented numerous cases of infections linked to contaminated DUWLs . Microorganisms in dental aerosols can originate from DUWL contamination, such as water lines connecting high-speed handpieces, ultrasonic scalers, or air/water syringes . Sources of contamination include dental unit water bottles, direct contamination of DUWLs, or backflow from dental handpieces . DUWLs typically feature narrow, lengthy pipes with fluctuating water flow and configurations that promote fluid retention, making them optimal reservoirs for microorganisms . Studies have shown DUWL microorganism levels can reach up to 200,000 CFU/ml within five days of installing a new dental unit . These microorganisms include viruses, bacteria, fungi, protozoa, and oral flora, primarily heterotrophic water bacteria . The American Dental Association (ADA) follows CDC guidelines stipulating that water used in nonsurgical dental procedures must contain heterotrophic bacteria no more than 500 CFU/ml, consistent with the EPA drinking water standard . Several methods can minimize dental bioaerosol contamination, including preoperative antiseptic mouth rinses, rubber dams, high-volume evacuators (HVE), high-efficiency particulate air (HEPA) filters, germicidal UV light, and DUWL disinfection. The well-known DUWL disinfection methods range from tablet systems, continuous-release straws or cartridges, and initial or periodic shock treatments to centralized systems . These methods potentially reduce microorganism levels in generated bioaerosols. Previous studies have shown effective reductions in fungi and viral contamination in dental bioaerosols, but studies focusing specifically on bacteria in clinical settings are lacking . Based on our previous data, a plasma sterilization system (PSS) for DUWL disinfection has demonstrated effectiveness . The PSS generates free radicals inducing oxidative stress in bacterial cells, leading to their demise. Additionally, iodine-releasing cartridges have been proven effective in controlling DUWL contamination . Iodide ions from the cartridge disrupt bacterial cell walls and kill the planktonic bacteria in the waterline. If these methods used to decontaminate DUWL also help improve air quality in dental clinics is worth exploring. In summary, our research aims to identify the source of bacterial contamination in dental bioaerosol after dental procedure and evaluate the efficacy of PSS and iodine-releasing cartridges for DUWL disinfection on bacterial counts and compositions in dental bioaerosols. Experimental designs Two methods of DUWL disinfection were implemented: an iodine-releasing cartridge system and a plasma sterilization system (PSS). Two dental chair units located in 2 different isolated treatment rooms within the faculty’s dental clinic, each with ventilation of 10 air changes per hour (ACH), were used (Fig. ). The first room was equipped with the plasma sterilization system (Dentozone ® Plasma System DPS2, Dentozone Corporation, Seoul, Korea), while the second room was fitted with the iodine-releasing cartridge system (DentaPure™ DP365B Independent Water Bottle Cartridge, HuFriedyGroup, Chicago, USA). The PSS was installed in the dental unit water pipeline system by professional maintenance staff, whereas the iodine-releasing cartridge system was directly submerged in the dental unit water bottle reservoir. Before implementing the disinfection systems, every dental unit water underwent initial treatment with reverse osmosis. The experiment was divided into two phases: the first phase, without waterline disinfectant installed, served as the control period, and the second phase, with waterline disinfectant installed, served as the test period. An adult volunteer served as a control variable in the experiment. Dental bioaerosols were generated using a high-speed handpiece for 5 minutes to mimic anterior teeth preparation, ensuring no physical contact occurred between the dental bur and the volunteer’s tooth. Only saliva ejectors from the dental assistant were utilized during the procedure. All dental procedures adhered to standard infection control protocols. The study protocol was approved by the Ethics Committee (HREC-DCU 2023 − 129) of the Faculty of Dentistry, Chulalongkorn University, Bangkok, Thailand, and biosafety approval was granted by the Institutional Biosafety Committee (DENT CU-IBC 034/2023) of the Faculty of Dentistry, Chulalongkorn University, Bangkok, Thailand. Sample collection and processing Air samples were collected from each treatment room, both before and after the dental procedure, using a two-headed active air sampling machine (VWR ® DUO Surface Air System 360) with trypticase soy agar (TSA) plates for air particle impaction sampling. One of the plates was covered with sterile filter paper (Whatman™ grade 1 qualitative filter paper) for subsequent DNA extraction. The machine was positioned as close to the volunteer’s mouth as possible on the dentist’s cabinet (Fig. ). The sampled TSA plates were then incubated at 37 degrees Celsius for 24 h for colony count and further studies. Additionally, stimulated saliva was collected from the volunteer before the procedure. The volunteer chewed parafilm for five minutes and then spat into a sterile test tube. Furthermore, 500 mL samples of dental unit water were collected directly from the tubing of the high-speed handpiece, with a 2–3-minute flush before collection. These water samples were plated on R2A agar for colony count and subsequent studies. Moreover, the remaining saliva samples, dental unit water samples, and air samples underwent DNA extraction using pre-made kits (DNeasy ® PowerWater ® Kit, Qiagen, Hilden, Germany), followed by whole-genome amplification (REPLI-G ® Mini Kit, Qiagen, Hilden, Germany). Targeted amplification of the V3-V4 variable regions was performed using 341F and 805R primers and SparQ HiFi PCR Master Mix (QuantaBio, USA). The amplified samples were purified using SparQ Puremag Beads (QuantaBio, USA) and subjected to 16S rDNA sequencing (MiSeqTM System, Illumina ® , CA, USA). DNAse- and RNase-free sterilized distilled water (Invitrogen™ UltraPure™ Distilled Water, NY, USA) was used as the blank. All procedures were performed in triplicate. Statistical analyses Finally, statistical analysis was performed using IBM SPSS Statistics (IBM Corp., NY, USA). Descriptive statistics was reported as mean and standard deviation for bacterial plate count, median and interquartile range for Shannon H index, and percentage for bacterial compositions. Alpha diversity was analyzed with the Shannon H index, while beta diversity was assessed with Bray-Curtis Dissimilarity using QIIME 2 , and visualized using OmicsBox software (Biobam Bioinformatics, Valencia, Spain). Analysis of microbiome composition (ANCOM) was conducted using ANCOM statistics. Two methods of DUWL disinfection were implemented: an iodine-releasing cartridge system and a plasma sterilization system (PSS). Two dental chair units located in 2 different isolated treatment rooms within the faculty’s dental clinic, each with ventilation of 10 air changes per hour (ACH), were used (Fig. ). The first room was equipped with the plasma sterilization system (Dentozone ® Plasma System DPS2, Dentozone Corporation, Seoul, Korea), while the second room was fitted with the iodine-releasing cartridge system (DentaPure™ DP365B Independent Water Bottle Cartridge, HuFriedyGroup, Chicago, USA). The PSS was installed in the dental unit water pipeline system by professional maintenance staff, whereas the iodine-releasing cartridge system was directly submerged in the dental unit water bottle reservoir. Before implementing the disinfection systems, every dental unit water underwent initial treatment with reverse osmosis. The experiment was divided into two phases: the first phase, without waterline disinfectant installed, served as the control period, and the second phase, with waterline disinfectant installed, served as the test period. An adult volunteer served as a control variable in the experiment. Dental bioaerosols were generated using a high-speed handpiece for 5 minutes to mimic anterior teeth preparation, ensuring no physical contact occurred between the dental bur and the volunteer’s tooth. Only saliva ejectors from the dental assistant were utilized during the procedure. All dental procedures adhered to standard infection control protocols. The study protocol was approved by the Ethics Committee (HREC-DCU 2023 − 129) of the Faculty of Dentistry, Chulalongkorn University, Bangkok, Thailand, and biosafety approval was granted by the Institutional Biosafety Committee (DENT CU-IBC 034/2023) of the Faculty of Dentistry, Chulalongkorn University, Bangkok, Thailand. Air samples were collected from each treatment room, both before and after the dental procedure, using a two-headed active air sampling machine (VWR ® DUO Surface Air System 360) with trypticase soy agar (TSA) plates for air particle impaction sampling. One of the plates was covered with sterile filter paper (Whatman™ grade 1 qualitative filter paper) for subsequent DNA extraction. The machine was positioned as close to the volunteer’s mouth as possible on the dentist’s cabinet (Fig. ). The sampled TSA plates were then incubated at 37 degrees Celsius for 24 h for colony count and further studies. Additionally, stimulated saliva was collected from the volunteer before the procedure. The volunteer chewed parafilm for five minutes and then spat into a sterile test tube. Furthermore, 500 mL samples of dental unit water were collected directly from the tubing of the high-speed handpiece, with a 2–3-minute flush before collection. These water samples were plated on R2A agar for colony count and subsequent studies. Moreover, the remaining saliva samples, dental unit water samples, and air samples underwent DNA extraction using pre-made kits (DNeasy ® PowerWater ® Kit, Qiagen, Hilden, Germany), followed by whole-genome amplification (REPLI-G ® Mini Kit, Qiagen, Hilden, Germany). Targeted amplification of the V3-V4 variable regions was performed using 341F and 805R primers and SparQ HiFi PCR Master Mix (QuantaBio, USA). The amplified samples were purified using SparQ Puremag Beads (QuantaBio, USA) and subjected to 16S rDNA sequencing (MiSeqTM System, Illumina ® , CA, USA). DNAse- and RNase-free sterilized distilled water (Invitrogen™ UltraPure™ Distilled Water, NY, USA) was used as the blank. All procedures were performed in triplicate. Finally, statistical analysis was performed using IBM SPSS Statistics (IBM Corp., NY, USA). Descriptive statistics was reported as mean and standard deviation for bacterial plate count, median and interquartile range for Shannon H index, and percentage for bacterial compositions. Alpha diversity was analyzed with the Shannon H index, while beta diversity was assessed with Bray-Curtis Dissimilarity using QIIME 2 , and visualized using OmicsBox software (Biobam Bioinformatics, Valencia, Spain). Analysis of microbiome composition (ANCOM) was conducted using ANCOM statistics. Bacterial plate count in DUWL and air samples Bacterial counts in DUWL samples showed a decreasing trend after treatment with iodine or PSS (Fig. ). In contrast, the trend was the opposite for air samples (Fig. ). Bacterial composition in generated dental bioaerosols A total of 48 samples from saliva, DUWL, and air were analyzed using 16S rDNA sequencing, identifying an average of 80,069 reads per sample. Alpha rarefaction curves generated using the Shannon H index determined the appropriate sequencing depth, which stabilized at 1,762 species-level Operational Taxonomic Units (OTUs) for the iodine-treated groups, while the PSS-treated groups stabilized at 673 species-level OTUs. The bacterial composition in generated dental bioaerosols was analyzed using Amplicon Sequence Variants (ASVs) for greater accuracy. Bacterial dispersion in the air following aerosol-generating procedures (AGP) was confirmed to originate from the air before the dental procedure, volunteer’s saliva, and DUWL for both iodine and plasma-treated groups, with bacteria from the air before AGP constituting the largest proportion, accounting for 65.13% and 60.33% in iodine-treated and plasma-treated groups, respectively (Fig. ). Additionally, a shift in bacterial composition in the air after installing DUWL disinfecting systems was observed, although the trend was inconclusive. However, bacteria from DUWL samples exhibited reduced diversity after treatment with water disinfectants, observed in both iodine and PSS groups. Among the bacterial taxa identified in the air and DUWL samples, pathogenic genera and species were found, including Bacillus spp., Campylobacter concisus , Capnocytophaga leadbetteri , Corynebacterium spp., Cutibacterium spp., Escherichia-Shigella spp., Haemophilus spp., Novosphingobium spp., Sphingomonas spp., Staphylococcus spp., Streptococcus salivarius , Streptococcus spp., and Veillonella spp. The phylum Proteobacteria was the most predominant in DUWL and air samples, whereas the phyla Firmicutes and Bacteroidota were predominant in saliva samples (Fig. ). The saliva samples exhibited the highest alpha diversity, which refers to diversity within the samples, while the air samples showed the lowest alpha diversity in both iodine and PSS groups (Fig. ). Additionally, treatment with PSS led to a slight increase in DUWL alpha diversity (Fig. ). The differences in diversity among samples (beta diversity) were analyzed using Bray-Curtis Dissimilarity. The results indicated no significant differences in diversity among samples, except for the saliva samples, which clustered together distinctly and were notably separated from the other DUWL and air samples (Fig. ). The data were further analyzed with the Analysis of Composition of Microbiomes (ANCOM) using ANCOM test statistics, which showed no predominant species among the samples (Fig. ). Bacterial counts in DUWL samples showed a decreasing trend after treatment with iodine or PSS (Fig. ). In contrast, the trend was the opposite for air samples (Fig. ). A total of 48 samples from saliva, DUWL, and air were analyzed using 16S rDNA sequencing, identifying an average of 80,069 reads per sample. Alpha rarefaction curves generated using the Shannon H index determined the appropriate sequencing depth, which stabilized at 1,762 species-level Operational Taxonomic Units (OTUs) for the iodine-treated groups, while the PSS-treated groups stabilized at 673 species-level OTUs. The bacterial composition in generated dental bioaerosols was analyzed using Amplicon Sequence Variants (ASVs) for greater accuracy. Bacterial dispersion in the air following aerosol-generating procedures (AGP) was confirmed to originate from the air before the dental procedure, volunteer’s saliva, and DUWL for both iodine and plasma-treated groups, with bacteria from the air before AGP constituting the largest proportion, accounting for 65.13% and 60.33% in iodine-treated and plasma-treated groups, respectively (Fig. ). Additionally, a shift in bacterial composition in the air after installing DUWL disinfecting systems was observed, although the trend was inconclusive. However, bacteria from DUWL samples exhibited reduced diversity after treatment with water disinfectants, observed in both iodine and PSS groups. Among the bacterial taxa identified in the air and DUWL samples, pathogenic genera and species were found, including Bacillus spp., Campylobacter concisus , Capnocytophaga leadbetteri , Corynebacterium spp., Cutibacterium spp., Escherichia-Shigella spp., Haemophilus spp., Novosphingobium spp., Sphingomonas spp., Staphylococcus spp., Streptococcus salivarius , Streptococcus spp., and Veillonella spp. The phylum Proteobacteria was the most predominant in DUWL and air samples, whereas the phyla Firmicutes and Bacteroidota were predominant in saliva samples (Fig. ). The saliva samples exhibited the highest alpha diversity, which refers to diversity within the samples, while the air samples showed the lowest alpha diversity in both iodine and PSS groups (Fig. ). Additionally, treatment with PSS led to a slight increase in DUWL alpha diversity (Fig. ). The differences in diversity among samples (beta diversity) were analyzed using Bray-Curtis Dissimilarity. The results indicated no significant differences in diversity among samples, except for the saliva samples, which clustered together distinctly and were notably separated from the other DUWL and air samples (Fig. ). The data were further analyzed with the Analysis of Composition of Microbiomes (ANCOM) using ANCOM test statistics, which showed no predominant species among the samples (Fig. ). We analyzed the effectiveness of DUWL disinfectants, iodine-releasing cartridge system, and PSS in terms of bacterial amount and composition in both DUWL and aerosols using cultivation and metataxonomic workflows. Our experiment employed solid plate impactors for active air sampling, which involves directing air onto agar plates to capture potentially inhalable pathogens. This method was chosen because it can detect airborne pathogens that passive air sampling methods cannot and is widely used for this purpose . We opted to use stimulated saliva samples instead of unstimulated saliva due to their greater bacterial diversity, which better represents the volunteer’s oral microbiome . Our findings regarding bacterial composition in aerosols post-AGP align with previous studies, indicating that airborne bacteria originate from patient saliva, DUWL, and pre-AGP air . However, our results diverged in terms of relative proportions, likely due to variability in bacterial taxa counts across samples. Previous studies on the effectiveness of DUWL disinfectants by Allison et al. demonstrated the effective use of DUWL disinfectants (Alpron ® and A-dec ICX ® ) in reducing viable virus count in generated dental bioaerosols. However, their studies were conducted using a dental mannequin . Two other studies by Patil et al., which utilized 0.1% sodium hypochlorite and 0.2% chlorhexidine, and Sethi et al., which used 0.2% chlorhexidine and cinnamon extract 20% w/v, also conducted similar experiments focusing on bacterial counts and demonstrated the effectiveness of DUWL disinfection . However, the null hypothesis regarding the efficacy of the iodine-releasing cartridge system and PSS in reducing bacterial counts and altering composition in our study cannot be rejected due to several factors. Although the iodine-releasing cartridge system showed a decreasing trend in bacterial counts in both DUWL and air samples, no significant difference was observed. Our iodine and plasma-treated DUWL results were within the acceptable range when compared to the ADA standard, which specifies that DUWLs should not exceed 500 CFU/mL of heterotrophic water bacteria . Additionally, there were no significant differences in bacterial composition observed. The similarity in bacterial composition before and after treatment with waterline disinfectants suggests that no specific bacterial taxa were consistently removed from the water or generated aerosols, consistent with our ANCOM results, which showed no predominant bacterial species among the samples. Several factors could have contributed to our insignificant results. Firstly, the design of both waterline disinfectants—the iodine-releasing cartridge system and PSS—was intended for continuous use, necessitating control over low iodide concentrations and short half-life plasma ions. However, these regulatory constraints likely resulted in lower disinfection efficacy against bacteria dispersed in the air after AGP, possibly due to limited substantivity of disinfectants. Secondly, the experiment was conducted only three times in triplicate, which could have increased data variance and contributed to insignificant results compared to previous studies conducted on larger sample sizes . Lastly, the agar impaction method was intended solely for culturable analysis. Additional steps beyond colony counting were necessary for a comprehensive investigation . DNA extraction from filter paper was a crucial step in our metataxonomic study workflow. Opting for DNA extraction from filter paper posed a risk of the filter paper absorbing most of the lysis buffer, potentially reducing cell lysis efficiency. Additionally, a liquid medium was preferred for air sampling when extensive extraction was required . The bacterial composition in the generated dental bioaerosols originates from the patient’s saliva, dental unit water, and air before the dental procedure, dispersing into aerosols after the procedure, with the most predominant phylum being Proteobacteria. Both the iodine-releasing cartridge system and plasma sterilization system were able to control bacterial contamination in dental unit water to be within standard limits but minimally alter bacterial composition in both treated dental unit water and generated dental bioaerosols. However, we still recommend continuing waterline disinfection alongside other infection control measures to minimize disease transmission in dental clinics. |
PEGDA-based HistoBrick for increasing throughput of cryosectioning and immunohistochemistry in organoid and small tissue studies | 53f1402b-bb5b-400a-8b5d-74556df6d768 | 11696907 | Anatomy[mh] | Histology provides insights into tissue structure and cellular morphology and is the gold standard analytical method for clinicians and researchers. Tissue sectioning along with immunohistochemical staining reveals changes in tissue morphology or physiology in response to different treatments and environments. Histological sections are most commonly prepared via paraffin infiltration or cryosectioning. In terms of antigen accessibility for immunohistochemical stainings, cryosections are considered to be superior to paraffin sections. Paraffin infiltration hides some of the antigens, necessitating long non-standardized antigen retrieval protocols for immunostaining. Cryosectioning generally preserves antigens well and enables their visualization via immunohistochemistry , . Microtissues (organoids, spheroids, tumoroids and related complex 3D cellular models) serve as invaluable in vitro models of tissues, offering insights into organ development, disease phenotypes and drug responses , . While some microtissues can be produced in large numbers, their high-throughput analysis remains a challenge – . Preparing histological sections to analyze microtissues or small tissues is a low-throughput, labor-intensive and operator-dependent process , . Typically, the operator manually embeds a few (less than ten) microtissues into a matrix to form a block (Fig. a). Because individual microtissues in the block cannot be easily traced during conventional embedding and subsequent analysis steps, often one or multiple blocks need to be prepared per experimental condition. The time-consuming histological process is especially a bottleneck in the field of drug discovery, where large numbers of compounds are often validated in multiple sample replicates . Strategies have been deployed to increase the throughput of microtissue histology. We and others have described planar arrays to spatially organize samples within blocks – or to centrifuge samples for arrangement on the same sectioning plane . For cryosectioning, tissue microarray-inspired approaches remain limited either in terms of alignment precision, intensive labor, or process compatibility with various microtissue culture approaches , . An ideal and versatile method to increase throughput of microtissues histological analysis should provide: 1. Spatially organizing microtissues in the sectioning plane to trace individual samples throughout the entire study, allowing the combination of different experimental conditions within one block; 2. Highly increasing the number of microtissues within one block; 3. Aligning microtissues in a narrow horizontal plane in the embedded block to reduce the number of sections required for the analysis; 4. Ease-of-use and standardized procedure involving readily accessible tools and materials with integration potential in automated workflows. We aimed to create a solution adapted to cryosectioning that addresses the previous advantages by improving our previously established HistoBrick tool, in which spheroid arrangement is spatially controlled in a paraffin embedded agarose block with arrayed mini-wells (Fig. b). Because the original agarose-based HistoBrick is not suited for cryosectioning, there is a need to develop a HistoBrick made of a cryosectioning-compatible material. The selection of a suitable embedding matrix for HistoBrick preparation is crucial to ensure the integrity of the frozen block (also called cryoblock) throughout the sectioning process. Fragile samples that lack rigidity (neuronal tissue), contain cavities (cochlea biopsy), or present complex surface topography (retinal organoids) can be structurally distorted during cryosectioning, rendering analysis of their immunostaining less informative – . Optimal Cutting Temperature compound (OCT) is a widely used matrix for embedding samples for cryosectioning. However, OCT melts at room temperature, leaving the histological sections which may disrupt fragile sample structures . Gelatine is preferred by some scientists to embed soft tissues such as brain and cochlea. One advantage of gelatine is the mechanically stable sample-matrix interfaces , , . However, gelatine presents the disadvantage of needing to be kept above 37 °C to stay liquid. At room temperature, the viscosity of gelatine increases, making pipetting difficult. Other hydrogels, such as PEGDA, have a stable viscosity at room temperature and their crosslinking can be precisely controlled using UV light. However, PEG-derived hydrogels alone are not suitable for cryosectioning . Here, we describe a new hydrogel mixture composed of PEGDA, gelatine and sucrose suited for HistoBrick preparation and cryosectioning. The new material composition is tested against different parameters including ease of preparation, integrity of cryosections, vertical alignment of microtissues and preservation of their structure. The impact of the embedding matrix on the structural integrity of the sample is studied using retinal organoids. Retinal organoids, like the adult human retina, are organized in layers containing different cell types , . Along the outer organoid surface, they present particularly fragile, hair-like structures made of light-sensitive photoreceptor outer segments. Cells of the human retina, in particular photoreceptors and their outer segments, are affected by several diseases leading to blindness , . Reliable visualization of retinal layers, photoreceptors and their outer segment structures is thus essential. We confirm that the new embedding material for cryosectioning preserves layering of the organoids, photoreceptor appearance and the fragile outer segments throughout the histological process. Taking advantage of PEGDA-gelatine HistoBricks, we analyze a two-year time-course of retinal organoid development and describe features of human retina degeneration (such as the loss of photoreceptor outer segments) and features of human retina aging (such as displaced photoreceptors in the region of outer segments). Embedding up to 16 retinal organoids from different experimental conditions into one HistoBrick increases the throughput of microtissue histological analysis for cryosectioning, while decreasing cost of the analysis by saving reagents and time. Optimizing hydrogel formulation for HistoBrick preparation and cryosectioning We aimed to identify a hydrogel to adapt the HistoBrick method to cryosectioning. The HistoBrick was prepared by first filling liquid embedding material into a silicone mold, and subsequently crosslinking to obtain a gel well plate. The gel well plate contained an array of mini-wells in which microtissues were transferred. After sedimentation of the microtissues to the well bottom, the wells were filled by liquid embedding material and the resulting HistoBrick was snap frozen and cut into thin Sections (10–30 µm) on a cryostat (Fig. ). In this manuscript, the word HistoBrick refers to both the embedding method and the resulting block containing the gel well plate, the microtissues and the embedding matrix. The gel well plate is the empty arrayed hydrogel and the embedding matrix is the hydrogel in direct contact with the microtissues in the wells. The gel well plate and the embedding matrix are always made of the same material. After snap freezing, the frozen HistoBrick is called cryoblock. We first tested the fabrication and cryosectioning of the HistoBrick with either gelatine or PEGDA, not containing microtissues. OCT was not considered for HistoBrick fabrication because an OCT HistoBrick would need to be loaded while frozen which would not allow sedimentation of small samples that would freeze along the cold wells at random heights. We successfully molded gel well plates with the two materials. While crosslinking of the gel well plate prepared with gelatine required one hour at 4 °C, crosslinking with UV for the PEGDA solution only took a few minutes. The unmolding of the gelatine gel well plate was more challenging than the one prepared of PEGDA, as solidified gelatine tended to stick to the silicone mold. We subsequently investigated the integrity of the sections, and the interface between the gel well plate and the embedding matrix. Sectioning the PEGDA HistoBrick (8 v% and 10 v% PEGDA) resulted in frequent wrinkling and breakage of the sections, highlighting poor mechanical stability. Additionally, filling the PEGDA gel well plate with PEGDA solution resulted in a non-adherent interface, which led to the disassociation of the embedded matrix from the gel well plate (Fig. a left). Consequently, some embedding matrix regions were lost or folded during sectioning. During an analysis experiment containing organoids this would result in the loss of microtissues located inside. The sections of the gelatine HistoBricks were stable without major cracks or folds, and with good cohesion at the interface of the well plate and embedding matrix (Fig. a middle). Due to its poor sectionability, the PEGDA hydrogel was not used for further testing with organoids. To promote a coherent interface between organoids and the embedding matrix, organoids were incubated in the embedding solutions for 15 min prior to the transfer into the gel well plate. The pipetting of organoids, incubated in gelatine solution, into the wells trapped air bubbles at the bottom. The presence of air bubbles at the bottom of the wells prevented the organoids from sedimenting and aligning on a unique plane while reducing the integrity of the sections during cutting. An additional barrier to a planar arrangement of organoids in gelatine HistoBricks was their tendency to attach to the sides of the wells during transfer. Consequently, we had to gently push them down to the bottom of the wells using a pipette tip. To optimize organoid planar embedding and HistoBrick sectioning, we aimed to build on the advantages of both tested materials. We developed a new embedding matrix by mixing 8 v% PEGDA with 2.5 wt% of gelatine (thereafter called PEGDA-gelatine hydrogel). The PEGDA-gelatine HistoBrick combined the fast and easy unmolding of the PEGDA HistoBrick and the structural stability and coherence of embedding matrix-gel well plate interface during cryosectioning of the gelatine HistoBrick (Fig. a right). We did not observe air bubble trapping when pipetting organoids incubated in PEGDA-gelatine solution and organoids efficiently sedimented to the bottom of the wells. We then investigated the interface between the organoids and the embedding matrix in more detail with a hematoxilin and eosin staining, as a coherent interface is important to support the tissue structure. Eosin stained the proteins contained in gelatine and thus enabled embedding matrix visualization. Both gelatine and PEGDA-gelatine matrix presented a stable interface with the organoids, meaning that the embedding matrices did not detach from the organoid surface throughout the whole histological process (Fig. b). The gelatine sections have a structure less homogeneous than the PEGDA-gelatine sections but this was not found to impact the interface with the organoids. Conventional embedding in OCT was also performed, as it is known to present a weak interface . Indeed, during the staining procedure the OCT was washed away, not maintaining an interface with the tissue (Fig. b). Hydrogel rheological analysis We next aimed to understand how viscosity of the hydrogel solutions affected their compatibility with organoid transfer procedures. Using rheology, we measured changes in viscosity of the PEGDA, gelatine, and PEGDA-gelatine solutions over time at temperatures representative of the HistoBrick loading step. To recreate similar conditions, the measurements were started with a temperature ramp going from 37 to 20 °C (5 °C/min), followed by an isotherm at 20 °C (Fig. c). As the solutions underwent temperature-driven crosslinking during the tests, we measured the complex viscosity that considered the elastic and viscous component of crosslinked hydrogels. The complex viscosity of the PEGDA solution stayed stable throughout the whole experiment. The isotherm at 20 °C was shortened as the PEGDA solution showed a stable behavior. The viscosity of the gelatine solution started to increase during the temperature ramp at 22 °C and stabilized at the end of the ramp when the temperature reached 20 °C. The increase of viscosity was caused by the crosslinking of the gelatine polymers inside the solution upon cooling. On the other hand, the viscosity of the PEGDA-gelatine solution started to increase at 20 °C and stabilized 6 min later. The PEGDA polymers do not influence the viscosity as in these experiments no UV crosslinking was performed. Therefore, the difference between the two solutions is interpreted to come from the difference in gelatine concentration. These results support experimental observations that PEGDA-gelatine solution was easier to handle when pipetting into the gel well plate due to its slower increase of viscosity during the process. PEGDA-gelatine preserves fragile organoid substructures To investigate if the embedding process of the HistoBrick has an impact on the structure of fragile organoid substructures, we used human retinal organoids. Like the human retina, retinal organoids are structured into layers with each layer containing specific cell types. Photoreceptor cell bodies are arranged in a layer called outer nuclear layer along the outside of retinal organoids. The inner nuclear layer and ganglion cell layers of retinal organoids, like the human adult retina, contain the other major cell types of the neural retina as described in Cowan et al . (Fig. a). Along the outer surface, retinal organoids display fragile outer segments. Outer segments are the light-sensitive “antennae” of photoreceptor cells. We embedded PFA-fixed, outer-segment-containing retinal organoids from the same organoid differentiation and compared conventionally embedded to HistoBrick embedded samples using both gelatine and PEGDA-gelatine as embedding materials. We evaluated both, overall organoid integrity, and the maintenance of outer segments on cryosections (Fig. , Supplementary Fig. ). As a negative control, we embedded retinal organoids in OCT, as we and others have described that this treatment does not preserve outer segments , . Due to the liquid state of OCT at room temperature and its very high viscosity, we used it only for conventional embedding. To analyze retinal organoid sections obtained using the different embedding materials we performed immunostainings (Fig. b–f). For visualization of cone photoreceptors, we stained them with an antibody against arrestin 3. Rod photoreceptor morphology was visualized using an antibody against rhodopsin labelling rod cell bodies in the outer nuclear layer and rod outer segments as thin appendices and dots surrounding the organoid. For both conventional and HistoBrick embedding using gelatine and PEGDA-gelatine, and for OCT embedding, the outer nuclear layer containing cone (ARR3 positive) and rod (RHO positive) photoreceptor cell bodies was well preserved (Fig. b–f). We stained for various retinal cell types, like cones expressing M- and L-opsin, horizontal cells (ONECUT2 positive), ON bipolar cells (TRPM1 positive), Müller cells (RLBP1 positive), and ribbon synapses (Basson) and observed that organoid sections were well maintained and the cell types, as well as ribbon synapses, could be detected across all conditions (Supplementary Fig. a–e). To observe cone outer segments, we stained organoid cryosections with the lectin Peanut agglutinin (PNA). PNA in retinal organoids labels both inner segments and outer segments of cones (Supplementary Fig. f). Cone inner segments appear bud-like adjacent to the outer nuclear layer, and the outer segments extend significantly farther than the inner segments (Supplementary Fig. f). In both gelatine and PEGDA-gelatine cryoblock preparations (conventional and HistoBrick embedding), RHO and PNA staining confirm the presence of cone and rod outer segments along the outside of retinal organoids (Fig. b–f). Upon OCT embedding, the RHO and PNA positive outersegments were completely lost, and only PNA-positive and RHO positive inner segment buds remained (Fig. d). Our stainings revealed that gelatine in some cases did not preserve outer segments well, leading to stretched outer segments, holes in the outer segment area that overall appeared rather porous, or outer segments ripped off the organoid (Supplementary Fig. g–i). While we observed those artefacts for both conventional and HistoBrick embedding using gelatine, it happened more frequently using the HistoBrick. Interestingly, we observed less damage to outer segments in the form of stretching or holes using PEGDA-gelatine than using gelatine embedding, irrespective of conventional or HistoBrick embedding. To quantify the preservation of outer segments, we estimated the average thickness ϑ of the PNA-positive outer and inner segments (“OS + IS thickness”) under different embedding conditions. We first defined a region-of-interest (ROI) of the retinal organoid by the PNA signal (PNA ROI), which contained outer segments, inner segments and the inner part of the retinal organoid. We then defined a ROI of the retinal organoid by the Hoechst signal (Hoechst ROI), which contained nuclei and the inner part of the retinal organoid but excluded outer and inner segments. The area of the outer and inner segments (OS + IS area) was estimated by subtracting the area of the Hoechst ROI (Hoechst area) from the area of the PNA ROI (PNA area; Fig. g and Supplementary Fig. b–e). The OS + IS thickness ϑ was estimated from the PNA area and the Hoechst area by a circular approximation (Methods) and was independent of the Hoechst area (Supplementary Figs. f, g and i, j), which varied due to variance in organoid size (Supplementary Fig. h,i) or cross-section area. A significant reduction in ϑ was observed comparing samples conventionally embedded in OCT to the gelatine control. There were no significant differences in ϑ for organoids treated with conventional gelatine embedding compared to gelatine HistoBrick, conventional PEGDA-gelatine, or PEGDA-gelatine HistoBrick embedding, respectively. Overall, these results show that PEGDA-gelatine is suitable for HistoBrick preparation and well preserves organoid layering and fragile photoreceptor outer segments. The HistoBrick improves organoid traceability and analysis throughput We optimized the HistoBrick design to contain up to 16 retinal organoids while keeping external dimensions compatible with cryosectioning (Fig. a). Organoid diameters are on average 1.17 mm (+ /− 0.25 mm SD; n = 16,306 organoid images quantified from but can measure up to 2.66 mm (Supplementary Fig. i). We designed the HistoBrick wells to be 2.4 mm in diameter. Up to three sections of the HistoBrick fit on one glass slide. Here, we show an example with two sections enabling the simultaneous analysis of up to 32 organoids (Fig. b). Sometimes, the interface between the embedding matrix and the gel well plate is not continuous. Residual PBS, from the pre-wetting of the wells, is hypothesized to not be fully mixed with the embedding matrix, thereby disrupting the interface. Sedimentation of organoids at the bottom of the wells results in serial cryosections with maximized information content per section. More than 80% of organoids were present in the cryoblock over a thickness of 390 ± 212 μm, which corresponds to 19 ± 10 consecutive sections of 20 μm. This large standard deviation is hypothesized to come from the heterogeneity of organoids size (Supplementary Fig. h,i) and morphology. In addition, the HistoBrick is often cut at a slightly tilted angle on the cryotome. A tilting angle between the HistoBrick and the blade was indicated by the fact that some wells only appeared in later sections (Supplementary Fig. a,b). Due to this misalignment, aligned organoids in the blocks would appear on different sections. The tilting angle inherent to manual sectioning was analyzed in more detail in our previous work and was hypothesized to reduce the number of visible organoids on the Sections . The individualization of organoids in an array facilitates their traceability throughout the analysis process from embedding into the HistoBrick to the sectioned samples on slides. When serial sections are prepared on consecutive glass slides, each organoids position can be tracked within the array. The different slides can be stained with various sets of dyes and antibodies, thus providing multiple datasets for a single organoid. First, the high information content obtained by immunohistochemistry on consecutive sections facilitates the correlation between the organoid structure and immunolocalization (Fig. b, c). Second, consecutive organoid sections can provide some insight into organoid 3D structure (Fig. b, d; Supplementary Fig. b). Finally, increasing the number of organoids embedded within one block makes their analysis cheaper. While in conventional blocks our lab previously embedded on average 4 organoids, the HistoBrick fits 16 organoids. Thus, depending on the scale of a performed experiment, using the HistoBrick can save 91% of sectioning time and 88% of reagents necessary for antibody stainings in an experiment in which 2 organoid replicates per treatment are analyzed (Supplementary Fig. d). Structural changes of human retinal organoids over time Retinal organoids require over six months of development to reach a stage where they exhibit transcriptomic and morphological characteristics closely resembling those of the adult human retina . We have previously shown that at week 46, the transcriptome of rod photoreceptors changes and some rod marker genes are down-regulated . It was not known whether from that age onward, organoids would lose photoreceptors, and until when photoreceptors can be maintained in retinal organoids. We cultured retinal organoids for 98 weeks, fixed organoids at week 30, 38, 46, 52, 70, 81 and 98 (n = 3–8 retinal organoids per time-point), and embedded them in PEGDA-gelatine HistoBricks. Processing the organoids using HistoBrick embedding greatly reduced the number of blocks that needed to be sectioned and stained (3 HistoBricks instead of 11 conventional blocks). We then assessed several quality criteria that are potentially compromised in aging organoids: 1. the presence of layered retina structures, 2. the presence of rod- and cone photoreceptors and 3. the presence of photoreceptor outer segments. We observed organoids that maintained both their outer nuclear layer and inner nuclear layer up to week 98 (Fig. a). Staining for the rod marker rhodopsin (RHO) and cone marker arrestin 3 (ARR3) revealed the presence and maintenance of both photoreceptor types up to week 98 in culture. However, while organoids are surrounded by dense outer segments between weeks 30 and 52, we observed a gradual loss of photoreceptor outer segments past week 52 (Fig. b). To quantify whether outer segments decrease over time, we estimated the OS + IS thickness [12pt]{minimal} $$$$ . For each individual organoid, we manually selected the section with the most abundant outer segments for quantification. Our measurements revealed that OS + IS thickness [12pt]{minimal} $$$$ negatively correlates with time (Fig. c, Spearman rank correlation r = − 0.38, P = 0.01). Up to week 52, OS + IS thickness [12pt]{minimal} $$$$ was significantly larger than the negative control (Fig. d). At week 98, we measured a significant decrease of OS + IS thickness [12pt]{minimal} $$$$ as compared to the positive control (Fig. d). Although retina structures and photoreceptors could still be detected, some degeneration events very likely occurred during the long culture period. While analyzing the stained images, in older organoids we unexpectedly observed displaced nuclei outside the outer nuclear layer, in the area where outer segments are located (Fig. b, Supplementary Fig. ). To determine the cell type identity of these displaced cells, we performed further stainings. The most common cell types for retinal organoids were examined using the markers SOX9 for Müller cell nuclei, MiTF for retinal pigment epithelium cells, and MAP2 for neuronal cells. The displaced cells were negative for Sox9 and MiTF (Supplementary Fig. a). The nuclei of the displaced cells in the Hoechst staining overlapped with the stained cell bodies in the MAP2 stained images (Supplementary Fig. b–h, left), confirming a neuronal cell identity. Further evaluation of the ARR3 stained images indicated that these cells were cones, (Supplementary Fig. b–h, right). Gartner et al . and Lai et al . described displaced nuclei in the region of outer segments in human retinas upon aging , . We found examples of displaced cells in both young and old organoids. However, while the displaced cells appeared individually in organoids at week 30 and 38, we saw examples of displaced nuclei arranged chain-like near each other in older organoids (Supplementary Fig. b–h). We aimed to identify a hydrogel to adapt the HistoBrick method to cryosectioning. The HistoBrick was prepared by first filling liquid embedding material into a silicone mold, and subsequently crosslinking to obtain a gel well plate. The gel well plate contained an array of mini-wells in which microtissues were transferred. After sedimentation of the microtissues to the well bottom, the wells were filled by liquid embedding material and the resulting HistoBrick was snap frozen and cut into thin Sections (10–30 µm) on a cryostat (Fig. ). In this manuscript, the word HistoBrick refers to both the embedding method and the resulting block containing the gel well plate, the microtissues and the embedding matrix. The gel well plate is the empty arrayed hydrogel and the embedding matrix is the hydrogel in direct contact with the microtissues in the wells. The gel well plate and the embedding matrix are always made of the same material. After snap freezing, the frozen HistoBrick is called cryoblock. We first tested the fabrication and cryosectioning of the HistoBrick with either gelatine or PEGDA, not containing microtissues. OCT was not considered for HistoBrick fabrication because an OCT HistoBrick would need to be loaded while frozen which would not allow sedimentation of small samples that would freeze along the cold wells at random heights. We successfully molded gel well plates with the two materials. While crosslinking of the gel well plate prepared with gelatine required one hour at 4 °C, crosslinking with UV for the PEGDA solution only took a few minutes. The unmolding of the gelatine gel well plate was more challenging than the one prepared of PEGDA, as solidified gelatine tended to stick to the silicone mold. We subsequently investigated the integrity of the sections, and the interface between the gel well plate and the embedding matrix. Sectioning the PEGDA HistoBrick (8 v% and 10 v% PEGDA) resulted in frequent wrinkling and breakage of the sections, highlighting poor mechanical stability. Additionally, filling the PEGDA gel well plate with PEGDA solution resulted in a non-adherent interface, which led to the disassociation of the embedded matrix from the gel well plate (Fig. a left). Consequently, some embedding matrix regions were lost or folded during sectioning. During an analysis experiment containing organoids this would result in the loss of microtissues located inside. The sections of the gelatine HistoBricks were stable without major cracks or folds, and with good cohesion at the interface of the well plate and embedding matrix (Fig. a middle). Due to its poor sectionability, the PEGDA hydrogel was not used for further testing with organoids. To promote a coherent interface between organoids and the embedding matrix, organoids were incubated in the embedding solutions for 15 min prior to the transfer into the gel well plate. The pipetting of organoids, incubated in gelatine solution, into the wells trapped air bubbles at the bottom. The presence of air bubbles at the bottom of the wells prevented the organoids from sedimenting and aligning on a unique plane while reducing the integrity of the sections during cutting. An additional barrier to a planar arrangement of organoids in gelatine HistoBricks was their tendency to attach to the sides of the wells during transfer. Consequently, we had to gently push them down to the bottom of the wells using a pipette tip. To optimize organoid planar embedding and HistoBrick sectioning, we aimed to build on the advantages of both tested materials. We developed a new embedding matrix by mixing 8 v% PEGDA with 2.5 wt% of gelatine (thereafter called PEGDA-gelatine hydrogel). The PEGDA-gelatine HistoBrick combined the fast and easy unmolding of the PEGDA HistoBrick and the structural stability and coherence of embedding matrix-gel well plate interface during cryosectioning of the gelatine HistoBrick (Fig. a right). We did not observe air bubble trapping when pipetting organoids incubated in PEGDA-gelatine solution and organoids efficiently sedimented to the bottom of the wells. We then investigated the interface between the organoids and the embedding matrix in more detail with a hematoxilin and eosin staining, as a coherent interface is important to support the tissue structure. Eosin stained the proteins contained in gelatine and thus enabled embedding matrix visualization. Both gelatine and PEGDA-gelatine matrix presented a stable interface with the organoids, meaning that the embedding matrices did not detach from the organoid surface throughout the whole histological process (Fig. b). The gelatine sections have a structure less homogeneous than the PEGDA-gelatine sections but this was not found to impact the interface with the organoids. Conventional embedding in OCT was also performed, as it is known to present a weak interface . Indeed, during the staining procedure the OCT was washed away, not maintaining an interface with the tissue (Fig. b). We next aimed to understand how viscosity of the hydrogel solutions affected their compatibility with organoid transfer procedures. Using rheology, we measured changes in viscosity of the PEGDA, gelatine, and PEGDA-gelatine solutions over time at temperatures representative of the HistoBrick loading step. To recreate similar conditions, the measurements were started with a temperature ramp going from 37 to 20 °C (5 °C/min), followed by an isotherm at 20 °C (Fig. c). As the solutions underwent temperature-driven crosslinking during the tests, we measured the complex viscosity that considered the elastic and viscous component of crosslinked hydrogels. The complex viscosity of the PEGDA solution stayed stable throughout the whole experiment. The isotherm at 20 °C was shortened as the PEGDA solution showed a stable behavior. The viscosity of the gelatine solution started to increase during the temperature ramp at 22 °C and stabilized at the end of the ramp when the temperature reached 20 °C. The increase of viscosity was caused by the crosslinking of the gelatine polymers inside the solution upon cooling. On the other hand, the viscosity of the PEGDA-gelatine solution started to increase at 20 °C and stabilized 6 min later. The PEGDA polymers do not influence the viscosity as in these experiments no UV crosslinking was performed. Therefore, the difference between the two solutions is interpreted to come from the difference in gelatine concentration. These results support experimental observations that PEGDA-gelatine solution was easier to handle when pipetting into the gel well plate due to its slower increase of viscosity during the process. To investigate if the embedding process of the HistoBrick has an impact on the structure of fragile organoid substructures, we used human retinal organoids. Like the human retina, retinal organoids are structured into layers with each layer containing specific cell types. Photoreceptor cell bodies are arranged in a layer called outer nuclear layer along the outside of retinal organoids. The inner nuclear layer and ganglion cell layers of retinal organoids, like the human adult retina, contain the other major cell types of the neural retina as described in Cowan et al . (Fig. a). Along the outer surface, retinal organoids display fragile outer segments. Outer segments are the light-sensitive “antennae” of photoreceptor cells. We embedded PFA-fixed, outer-segment-containing retinal organoids from the same organoid differentiation and compared conventionally embedded to HistoBrick embedded samples using both gelatine and PEGDA-gelatine as embedding materials. We evaluated both, overall organoid integrity, and the maintenance of outer segments on cryosections (Fig. , Supplementary Fig. ). As a negative control, we embedded retinal organoids in OCT, as we and others have described that this treatment does not preserve outer segments , . Due to the liquid state of OCT at room temperature and its very high viscosity, we used it only for conventional embedding. To analyze retinal organoid sections obtained using the different embedding materials we performed immunostainings (Fig. b–f). For visualization of cone photoreceptors, we stained them with an antibody against arrestin 3. Rod photoreceptor morphology was visualized using an antibody against rhodopsin labelling rod cell bodies in the outer nuclear layer and rod outer segments as thin appendices and dots surrounding the organoid. For both conventional and HistoBrick embedding using gelatine and PEGDA-gelatine, and for OCT embedding, the outer nuclear layer containing cone (ARR3 positive) and rod (RHO positive) photoreceptor cell bodies was well preserved (Fig. b–f). We stained for various retinal cell types, like cones expressing M- and L-opsin, horizontal cells (ONECUT2 positive), ON bipolar cells (TRPM1 positive), Müller cells (RLBP1 positive), and ribbon synapses (Basson) and observed that organoid sections were well maintained and the cell types, as well as ribbon synapses, could be detected across all conditions (Supplementary Fig. a–e). To observe cone outer segments, we stained organoid cryosections with the lectin Peanut agglutinin (PNA). PNA in retinal organoids labels both inner segments and outer segments of cones (Supplementary Fig. f). Cone inner segments appear bud-like adjacent to the outer nuclear layer, and the outer segments extend significantly farther than the inner segments (Supplementary Fig. f). In both gelatine and PEGDA-gelatine cryoblock preparations (conventional and HistoBrick embedding), RHO and PNA staining confirm the presence of cone and rod outer segments along the outside of retinal organoids (Fig. b–f). Upon OCT embedding, the RHO and PNA positive outersegments were completely lost, and only PNA-positive and RHO positive inner segment buds remained (Fig. d). Our stainings revealed that gelatine in some cases did not preserve outer segments well, leading to stretched outer segments, holes in the outer segment area that overall appeared rather porous, or outer segments ripped off the organoid (Supplementary Fig. g–i). While we observed those artefacts for both conventional and HistoBrick embedding using gelatine, it happened more frequently using the HistoBrick. Interestingly, we observed less damage to outer segments in the form of stretching or holes using PEGDA-gelatine than using gelatine embedding, irrespective of conventional or HistoBrick embedding. To quantify the preservation of outer segments, we estimated the average thickness ϑ of the PNA-positive outer and inner segments (“OS + IS thickness”) under different embedding conditions. We first defined a region-of-interest (ROI) of the retinal organoid by the PNA signal (PNA ROI), which contained outer segments, inner segments and the inner part of the retinal organoid. We then defined a ROI of the retinal organoid by the Hoechst signal (Hoechst ROI), which contained nuclei and the inner part of the retinal organoid but excluded outer and inner segments. The area of the outer and inner segments (OS + IS area) was estimated by subtracting the area of the Hoechst ROI (Hoechst area) from the area of the PNA ROI (PNA area; Fig. g and Supplementary Fig. b–e). The OS + IS thickness ϑ was estimated from the PNA area and the Hoechst area by a circular approximation (Methods) and was independent of the Hoechst area (Supplementary Figs. f, g and i, j), which varied due to variance in organoid size (Supplementary Fig. h,i) or cross-section area. A significant reduction in ϑ was observed comparing samples conventionally embedded in OCT to the gelatine control. There were no significant differences in ϑ for organoids treated with conventional gelatine embedding compared to gelatine HistoBrick, conventional PEGDA-gelatine, or PEGDA-gelatine HistoBrick embedding, respectively. Overall, these results show that PEGDA-gelatine is suitable for HistoBrick preparation and well preserves organoid layering and fragile photoreceptor outer segments. We optimized the HistoBrick design to contain up to 16 retinal organoids while keeping external dimensions compatible with cryosectioning (Fig. a). Organoid diameters are on average 1.17 mm (+ /− 0.25 mm SD; n = 16,306 organoid images quantified from but can measure up to 2.66 mm (Supplementary Fig. i). We designed the HistoBrick wells to be 2.4 mm in diameter. Up to three sections of the HistoBrick fit on one glass slide. Here, we show an example with two sections enabling the simultaneous analysis of up to 32 organoids (Fig. b). Sometimes, the interface between the embedding matrix and the gel well plate is not continuous. Residual PBS, from the pre-wetting of the wells, is hypothesized to not be fully mixed with the embedding matrix, thereby disrupting the interface. Sedimentation of organoids at the bottom of the wells results in serial cryosections with maximized information content per section. More than 80% of organoids were present in the cryoblock over a thickness of 390 ± 212 μm, which corresponds to 19 ± 10 consecutive sections of 20 μm. This large standard deviation is hypothesized to come from the heterogeneity of organoids size (Supplementary Fig. h,i) and morphology. In addition, the HistoBrick is often cut at a slightly tilted angle on the cryotome. A tilting angle between the HistoBrick and the blade was indicated by the fact that some wells only appeared in later sections (Supplementary Fig. a,b). Due to this misalignment, aligned organoids in the blocks would appear on different sections. The tilting angle inherent to manual sectioning was analyzed in more detail in our previous work and was hypothesized to reduce the number of visible organoids on the Sections . The individualization of organoids in an array facilitates their traceability throughout the analysis process from embedding into the HistoBrick to the sectioned samples on slides. When serial sections are prepared on consecutive glass slides, each organoids position can be tracked within the array. The different slides can be stained with various sets of dyes and antibodies, thus providing multiple datasets for a single organoid. First, the high information content obtained by immunohistochemistry on consecutive sections facilitates the correlation between the organoid structure and immunolocalization (Fig. b, c). Second, consecutive organoid sections can provide some insight into organoid 3D structure (Fig. b, d; Supplementary Fig. b). Finally, increasing the number of organoids embedded within one block makes their analysis cheaper. While in conventional blocks our lab previously embedded on average 4 organoids, the HistoBrick fits 16 organoids. Thus, depending on the scale of a performed experiment, using the HistoBrick can save 91% of sectioning time and 88% of reagents necessary for antibody stainings in an experiment in which 2 organoid replicates per treatment are analyzed (Supplementary Fig. d). Retinal organoids require over six months of development to reach a stage where they exhibit transcriptomic and morphological characteristics closely resembling those of the adult human retina . We have previously shown that at week 46, the transcriptome of rod photoreceptors changes and some rod marker genes are down-regulated . It was not known whether from that age onward, organoids would lose photoreceptors, and until when photoreceptors can be maintained in retinal organoids. We cultured retinal organoids for 98 weeks, fixed organoids at week 30, 38, 46, 52, 70, 81 and 98 (n = 3–8 retinal organoids per time-point), and embedded them in PEGDA-gelatine HistoBricks. Processing the organoids using HistoBrick embedding greatly reduced the number of blocks that needed to be sectioned and stained (3 HistoBricks instead of 11 conventional blocks). We then assessed several quality criteria that are potentially compromised in aging organoids: 1. the presence of layered retina structures, 2. the presence of rod- and cone photoreceptors and 3. the presence of photoreceptor outer segments. We observed organoids that maintained both their outer nuclear layer and inner nuclear layer up to week 98 (Fig. a). Staining for the rod marker rhodopsin (RHO) and cone marker arrestin 3 (ARR3) revealed the presence and maintenance of both photoreceptor types up to week 98 in culture. However, while organoids are surrounded by dense outer segments between weeks 30 and 52, we observed a gradual loss of photoreceptor outer segments past week 52 (Fig. b). To quantify whether outer segments decrease over time, we estimated the OS + IS thickness [12pt]{minimal} $$$$ . For each individual organoid, we manually selected the section with the most abundant outer segments for quantification. Our measurements revealed that OS + IS thickness [12pt]{minimal} $$$$ negatively correlates with time (Fig. c, Spearman rank correlation r = − 0.38, P = 0.01). Up to week 52, OS + IS thickness [12pt]{minimal} $$$$ was significantly larger than the negative control (Fig. d). At week 98, we measured a significant decrease of OS + IS thickness [12pt]{minimal} $$$$ as compared to the positive control (Fig. d). Although retina structures and photoreceptors could still be detected, some degeneration events very likely occurred during the long culture period. While analyzing the stained images, in older organoids we unexpectedly observed displaced nuclei outside the outer nuclear layer, in the area where outer segments are located (Fig. b, Supplementary Fig. ). To determine the cell type identity of these displaced cells, we performed further stainings. The most common cell types for retinal organoids were examined using the markers SOX9 for Müller cell nuclei, MiTF for retinal pigment epithelium cells, and MAP2 for neuronal cells. The displaced cells were negative for Sox9 and MiTF (Supplementary Fig. a). The nuclei of the displaced cells in the Hoechst staining overlapped with the stained cell bodies in the MAP2 stained images (Supplementary Fig. b–h, left), confirming a neuronal cell identity. Further evaluation of the ARR3 stained images indicated that these cells were cones, (Supplementary Fig. b–h, right). Gartner et al . and Lai et al . described displaced nuclei in the region of outer segments in human retinas upon aging , . We found examples of displaced cells in both young and old organoids. However, while the displaced cells appeared individually in organoids at week 30 and 38, we saw examples of displaced nuclei arranged chain-like near each other in older organoids (Supplementary Fig. b–h). In this work, we present a novel embedding matrix PEGDA-gelatine hydrogel to prepare an adapted HistoBrick for organoid cryosectioning. Identifying a novel embedding mixture was key to adapt the HistoBrick for cryosectioning of complex 3D microtissues. The embedding material should meet the following criteria. First, it should provide structural stability of the sections during cryosectioning. Importantly, a coherent interface between the gel well plate and the embedding matrix prevents the loss of valuable samples. Second, the embedding matrix should support and preserve the sample’s structural integrity throughout the whole process. A stable interface between the embedding material and the sample is crucial. Third, handling of the HistoBrick should be easy, including gel well plate unmolding from the silicone mold. Finally, the embedding material solution should allow easy transfer of organoids onto the bottom of the gel well plate. Ferguson et al. presented an embryoid body array providing traceability for six experimental conditions using OCT as an embedding matrix . The major limitation of this technique resides in the use of OCT which is a viscous material difficult to pipette, solidifying only below – 10 °C and not maintaining fragile structures such as outer segments of retinal organoids . Gelatine, which provides good stability to soft samples , was used to create an array of six half rodent brains (cm-scale ). The study highlights gelatine as a promising candidate for organoid or tissue arrays in the mm-scale like the HistoBrick. However, gelatine presents the disadvantage of needing to be kept above 37 °C to stay liquid. At room temperature, the viscosity of gelatine increases, making pipetting difficult. Consequently, UV-crosslinkable hydrogels with stable viscosity at room temperature are an interesting embedding matrix for the HistoBrick, facilitating gel pipetting and promoting organoid sedimentation inside the HistoBrick wells. PEGDA is a promising UV-curable candidate because it is widely commercially available, well studied in literature and exhibits biocompatibility properties . A previous study investigated the cryosectioning of a tissue-engineered construct composed of polyethylene glycol hydrogel (PEG) mixed with human mesenchymal stem cells. The untreated PEG-derived hydrogels were incompatible with the cryosectioning process . This behavior is due to the water-rich and polymeric nature of these compounds, which makes them very sensitive to routine tissue histology procedures. The highly ordered tetrahedral structure of water molecules in large ice crystals results in sample brittleness . The authors of the study demonstrated that overnight incubation in OCT or 1% Polyvinyl alcohol (PVA) of the tissue-engineered PEG-derived constructs increases the flexibility of the frozen sample facilitating sectioning . In another study, spheroids cultured in alginate and after fixation embedded in OCT were successfully cryosectioned . Alginate was not considered for the HistoBrick fabrication because its crosslinking relies on the diffusion of ions which is difficult to implement for large volumes. Based on these considerations, we compared gelatine and PEGDA as materials for the HistoBrick. As expected, PEGDA HistoBrick cryoblocks were very brittle and shattered during sectioning. On the other hand, gelatine acts as a plasticizer promoting material flexibility and reducing the sections’ brittleness. A small amount of gelatine (2.5 wt%) added to the PEGDA hydrogel enables the cutting of entire crack-free and fold-free sections while maintaining the advantage of easy handling low-viscous solution at room temperature and UV-crosslinking. Using a mixture of 8 v% PEGDA and 2.5 wt% gelatine, the user can prepare PEGDA-based cryoblocks without the need for prolonged overnight incubation in OCT or PVA solution, as required in a related published protocol . Furthermore, the addition of gelatine promotes the adhesion between the gel well plate and the embedding matrix, which is crucial to avoid the loss of precious samples. During freezing and sectioning, a separation of the tissue-matrix interface may create damage to the tissue. Proper infiltration of the embedding matrix inside the tissue promotes a coherent interface. Analysis of the substructure of human retinal organoids revealed that PEGDA-gelatine mixture preserves the fragile structure of outer segments, as well as the organization of the organoid nuclear layers and cell types. It correlates with the stable interface obtained between the embedding matrix and the organoids. Additionally, the PEGDA-gelatine mixture has a 3.4 times lower viscosity at 37 °C than gelatine 7.5 wt%, which promotes the diffusion of the matrix during incubation supporting the structure of fragile tissue (Supplementary Fig. b,c). Embedding with the gelatine HistoBrick protocol frequently resulted in damaged outer segments. The rapid solidification of gelatine during organoid transfer into the well plate required that organoids were pushed into well bottoms with pipette tips (to prevent their attachment to well walls or displacement by air bubbles), which may have damaged the organoids. PEGDA-gelatine circumvented this problem, as it is less viscous at room temperature and thereby more suited for organoid transfer into wells and retained the substructure of retinal organoid outer segments more efficiently. PEGDA-gelatine is thus the preferred material for fabrication and use of a HistoBrick for cryosectioning. The PEDGA-gelatine HistoBricks easily allowed the analysis of a two-year retinal organoid culture time-course. Instead of preparing and sectioning 11 conventional blocks with different time points and biological replicates, only 3 PEGDA-gelatine HistoBricks were required to analyze all samples. We have previously analyzed the development of retinal organoids in detail by single-cell RNA sequencing and compared them to the adult human retina . At week 46, we have found the cell type diversity of organoids to be decreased and that some cell type marker genes such as rhodopsin were downregulated in rods . One could interpret reduced expression of cell type marker genes as a decrease in cell quality or health. The decrease of outer segments that we visually observed from week 52 supports the indication that photoreceptors degenerate to some degree. In some retinal diseases, outer segments of photoreceptors degenerate, while their cell bodies remain viable . It has been unclear for how long photoreceptors can be maintained in cultured retinal organoids. We were surprised to still find organoids with clearly separated outer nuclear and inner nuclear layers between week 70 and up to week 98. We observed both cone and rod photoreceptors up to week 98. Unexpectedly, in older organoids we found displaced nuclei that we identified as cones, on top of the outer nuclear layer, in the region where outer segments are. It has been described in sections of human retinas that photoreceptor cell bodies can be displaced out of the outer nuclear layer into the region of outer segments during aging , . Further studies are needed to validate whether the gradual loss of photoreceptor outer segments and displacement of cones is a form of degradation in retinal organoids and whether this is similar to degeneration processes happening in human disease or aging. The HistoBrick method faces general challenges associated with the analysis of complex 3D microtissues on two-dimensional sections. First, each section contains only part of an organoid at a random sectioning angle which may or may not contain the tissue region or cell type of interest (Supplementary Fig. c), introducing variability in tissue composition. Second, organoids largely vary in size and shape , and on each given section only a fraction of the entire organoid is visible, introducing variability in size and geometry. Third, retinal organoids contain not only retina, but also regions of non-retinal identity, introducing variability in the ratio of retinal to non-retinal tissues. Fourth, depending on the cutting angle, appearance of retinal organoid layering, outer segment thickness, or distance between layers could be influenced by the sectioning angle. These sources of variability could be reduced by analyzing a larger number of sections per organoid, or optimally their full 3D signals. In addition, the analysis would benefit from automated organoid segmentation based on antibody staining. A future version of the HistoBrick, potentially using yet another material, could be adapted to organoid embedding for wholemount staining, clearing and imaging in 3D. Some aspects of the HistoBrick method can be further optimized to improve its user-friendliness. First, color can be added to the embedding material to help the visualization on the slide after sectioning and staining. Secondly, a marking sign visible on transparent sections could be incorporated into the HistoBrick design to ensure optimal sample traceability and re-orientation. Finally, automated transfer of organoids into the gel well plate and automated staining would further increase the throughput of the HistoBrick approach. We successfully adapted the HistoBrick methodology for an efficient preparation and analysis of organoid frozen sections. We expect the PEGDA-gelatine HistoBrick to provide optimal support to various fragile samples like brain organoids and cochlea biopsies. The PEGDA-gelatine HistoBrick increases the throughput of immunohistochemistry by simultaneously embedding up to 16 human retinal organoids from different experimental conditions in an organized manner. Processing a smaller number of cryoblocks for sectioning and stainings saves time and consumables such as antibodies required for staining the resulting sections (Supplemenatary Fig. d). Depending on the microtissue to analyze, the HistoBrick can easily be adapted to different organoid and tissue sizes, and organoid numbers per block can still be increased when smaller or more wells are used. Simultaneous immunohistochemical staining and imaging of multiple organoid samples on the same section reduces sample-to-sample variation and enables more reliable analysis. The array organization facilitates traceability of individual organoids through different sections and offers high potential for automated organoid transfer and image analysis. Furthermore, the HistoBrick is a versatile labware that can be easily implemented in standard laboratories thanks to its simple methodology and low-cost materials. The ease of HistoBrick re-design enables adoption of the technology across the field of microtissue research. Finally, with this tool in hand, the immunohistochemical validation of organoid high-throughput experiments such as compound-, toxicology-, or gene-therapy screening will be simplified. Gel well plate fabrication The silicone mold was designed similarly to our previous work . The diameter of the wells was increased to 2.4 mm to accommodate large retinal organoids. The outer dimensions of the HistoBrick were set to 20.7× 34.7 mm to contain 16 wells (Supplementary Fig. a). The mold was 3D printed in silicone “TrueSil Shore 50A” by the company Spectoplast. The schematic of the HistoBrick fabrication is depicted in Fig. . The HistoBrick 3D design file is available upon request to the corresponding authors. HistoBrick well-sizes and dimensions can be optimized for different organoid types. Gelatine gel well plate In a first approach, the silicone mold was placed with the pillars facing downward on a petri dish. A gelatine solution (7.5 wt% gelatine (Sigma Aldrich, #G1890), 10 wt% sucrose (Sigma Aldrich, #84,100) in 1 × PBS) was warmed at 37 °C and pipetted inside the silicone mold. The gelatine solution was crosslinked at 4 °C for 1 h. Afterwards, the mold was flipped and the gel well plate was released by gently lifting it from below. The obtained gel well plate was stored in a closed box with damp paper at 4 °C until use. PEGDA-based gel well plate In a second approach the silicone mold was placed with the pillars facing downward on a petri dish. The PEGDA-gelatine solution was prepared by mixing 8 v% PEGDA (Mw = 700, Sigma Aldrich, #455008), 2.5 wt% gelatine (Sigma Aldrich, #G1890), 10 wt% sucrose (Sigma Aldrich, #84100) and 0.05 wt% LAP (Sigma Aldrich, #900889) in 1 × PBS. The solution was pipetted inside the silicone mold and crosslinked by a 30 s exposure to 365 nm wavelength light with 35 mW/cm power (total illumination dose = 1.05 J/s*cm). The UV light was shined from above the mold. Afterward, the silicone mold was vertically lifted to release the gel well plate. The obtained gel well plate was stored in a closed box with humid paper at 4 °C until use. For the material selection, additional HistoBricks were prepared using a solution of 8 v% PEGDA Mw = 700, 10 wt% sucrose, 0.05 wt% LAP in 1 × PBS and 10 v% PEGDA Mw = 700, 10 wt% sucrose, 0.05 wt% LAP in 1 × PBS with the same crosslinking procedure. Organoid transfer into the gel well plate and HistoBrick freezing When the gelatine embedding material was used, the gel well plate was left to equilibrate at room temperature for 1 h. We refer to the liquid hydrogel solution filling the wells of the gel well plate as liquid embedding matrix. Before organoid transfer into the gel well plate, PFA-fixed, sucrose-cryopreserved retinal organoids were incubated in liquid embedding matrix of the same embedding material as the gel well plate for 15 min at 37 °C on a warming plate to enhance organoid-embedding matrix interaction. Incubation can also be performed in an incubator at 37 °C. The organoids were manually transferred with 20–50 µL of liquid embedding matrix to the gel well plate with a 200 µL tip pre-coated with 2% BSA (Sigma Aldrich, #A2153) in deionized water, to prevent organoids from sticking. The end of the tip was cut with a razor blade to create a larger aperture to avoid damaging the organoids. The gel well plate can be pre-wet by adding 20 µL of PBS in each well to minimize air bubble trapping. After the transfer, the wells were filled up with the same liquid embedding matrix used during the incubation of the organoids. Organoids were aligned onto a unique plane by centrifuging the filled gel well plate for 20 s at 200×g. The filled gel well plate underwent a second crosslinking (as previously described) to fix the position of the organoids at the bottom of the wells. The HistoBrick blocks were trimmed on all four sides so that three sections would later fit on the glass slides during cryosectioning. Finally, the HistoBrick was fixed on a labelled cardboard with a drop of OCT and snap frozen for 2 min by submersion in isopentane (Sigma Aldrich, #277258) cooled down to − 40 °C using dry ice pellets. The frozen HistoBricks were wrapped in aluminum foil to avoid drying and stored at − 80 °C until cryosectioning. Conventional cryoblock preparation Unless specified otherwise, the steps were performed using the same solutions and timings as for the gel well plate fabrication and organoid transfer. Gelatine and PEGDA-gelatine cryoblock The respective solution was pipetted inside a plastic weighing boat or an embedding mold and crosslinked to form a base layer of 2–4 mm thickness. During this time, retinal organoids were incubated in their respective embedding solutions for 15 min at 37 °C. Then, they were transferred one by one and placed on the base layer, followed by a second crosslinking. Finally, liquid embedding material was pipetted on top of the organoids to fill up the weighing boat and crosslinked. The blocks were trimmed with a scalpel to leave 1–2 mm space between the organoids and the block edge. The samples were then fixed on a labelled cardboard with a drop of OCT, snap frozen in isopentane at − 40 °C and stored at − 80 °C. OCT-based cryoblock OCT (Sakura, #4583) was poured inside a plastic weigh boat and crosslinked by freezing on dry ice pellets to form a base layer. The organoids were then transferred from their storage solution (without incubation) and placed on the base layer. Additional OCT was poured on the organoids, the blocks were frozen on dry ice and stored wrapped in aluminum foil at − 80 °C. HistoBrick cryoblock and conventional cryoblock sectioning The HistoBrick cryoblocks and the conventionally embedded cryoblocks were stored at − 20 °C overnight to acclimate to sectioning temperature. They were then mounted on a cryostat holder with OCT. If necessary, HistoBrick cryoblocks were again trimmed to fit 3 sections per slide, and conventional cryoblocks were trimmed to fit 8–12 sections per slide. Then they were sectioned in thin slices of 20 µm or 25 µm and placed on SuperFrost ® /Plus glass slides (Biosystems, #85-0911-00 or Epredia, #K5800AMNZ72). For results shown in Figs. , , Supplementary Figs. , and consecutive Histobrick sections were placed one section per glass slide on 5 slides. Afterwards, a second section was placed on each of the slides, followed by a third section, again starting from the first slide. After filling the first 5 slides, a second slide series of 5 slides was prepared using the same procedure. In this manner each slide contains sections from different organoid cutting depths, increasing the likelihood that there will be at least one usable section per organoid. The slides were stored at − 20 °C or − 80 °C until staining. Hematoxylin and Eosin staining To localize the organoids, slides were stained with hematoxylin and eosin. The slides were equilibrated and dried at room temperature for 1 h, then rehydrated in deionized water for 10 min. The slides were then incubated in Harris Hematoxylin (Harris Hematoxylin: Biosystems, #3873.2500) for 5 min and washed with running tap water. They were then differentiated for a few seconds in 1% acid-alcohol (Absolute alcohol: 7:10, VWR #20820.362, Hydrochloric acid 37%: 0.1:10, Sigma Aldrich, #30721 and H2O MiliQ 2.9:10) and washed under running tap water for 10 min. The slides were incubated in Eosine-Phloxine solution for 1 min (Eosin: 1:100, Sigma Aldrich, #E4382; Phloxine: 1:100, Sigma Aldrich, #P2759) and washed a last time. Finally, the slides were mounted with Eukitt (Sigma Aldrich, #03989) and dried overnight at room temperature. Immunostainings The slides were equilibrated and dried at room temperature for 1 h, then rehydrated in 1 × PBS for 10 min. The sections were blocked with 180 µL of 10% NDS blocking solution (10% Normal Donkey Serum (Sigma Aldrich, S30-M), 1% BSA (Sigma Aldrich, #A2153), 0.5% TritonX (Sigma Aldrich, #T9284), 0.1% sodium azide (Biosciences, #786-299), 1 × PBS) for 1 h. The blocking solution was removed, and the primary antibodies were added (ARR3: 1:400, Sigma Aldrich, #HPA063129; PNA: 1:1200, Sigma Aldrich #L6135; Rhodopsin: 1:400, Sigma Aldrich, #R5403; SOX9: 1:200, R&D systems, #AF3075; MiTF: 1:500, Exalpha, #X2398M; MAP2: 1:400, Sigma Aldrich, #AB5622; OneCut2: 1:100, R&D systems, # AF6294; Cralbp: 1:200, Abcam, #ab15051; Trpm1: 1:200, Sigma Aldrich, #HPA014779; Bassoon: 1:500, Enzo, # SAP7F407, L-M Opsin: 1:100, Merck, # AB5405). The primary antibodies were diluted in a 3% NDS solution (3% Normal Donkey Serum, 1% BSA, 0.5% TritonX, 0.1% sodium azide, 1 × PBS), 180 µL was added to each slide. The sections were covered with parafilm and incubated with the primary antibodies overnight at 4 °C in a humidified chamber. Afterwards, the slides were washed 3 times with PBS-T (1 × PBS, 0.05% TritonX) for 10 min each. The secondary antibodies (Hoechst 33342: 1:2000, Invitrogen, #H3570; anti-Mouse 488: 1:500, Invitrogen, #A32766; Streptavidin 555: 1:400, Invitrogen, #S32355; anti-Rabbit 640: 1:500, Invitrogen, #A32795; anti-Goat 488: 1:500, Invitrogen, #A32814; anti-Mouse 568: 1:500, Invitrogen, #A10037; anti-Sheep 488: 1:500, Invitrogen, # A11015) were diluted in 3% NDS and the sections were incubated for 2 h with 180 µL antibody solution at 4 °C. The slides were washed twice with PBS-T for 10 min each and once with 1 × PBS for 10 min. As a last step, the slides were mounted with Prolong Gold (Invitrogen, #P36934), coverslipped, and dried overnight at room temperature, sealed with nail polish (Electron Microscopy Sciences, #72180) and imaged. Image acquisition Images of the H&E-stained sections in Figs. and were acquired with an Olympus VS200 slide scanner. An overview image of the entire glass slide was acquired at 10x (UplaXapo, NA = 0.40) magnification in brightfield. Immunostained sections were acquired with a confocal spinning disc microscope (Olympus IXplore SpinSR10). Different magnifications were used depending on the desired image (4x: UplanSApo, NA = 0.16; 10x: UplanSApo, NA = 0.40; 20x: UplanSApo, NA = 0.75; 40x: UPlanSApo, NA = 1.05 Sil). For higher throughput of image acquisition and in Supplementary Fig. , a slide scanner was used (Zeiss Axio Scan.Z1) with 10 × magnification (Plan-Apochromate, NA = 0.45). OS + IS area measurements were performed on images obtained with the Zeiss Axio Scan.Z1 slide scanner. For each individual organoid, the section containing the highest quality outer segments (criteria: abundant OS with as little disruption as possible, intact organoid section) was manually selected for analysis. Image analysis Analysis of the vertical alignment of the human retinal organoids was performed on consecutive H&E-stained sections on a slide series by manually counting the number of organoids present on each section. For each HistoBrick, we determined the number of sections in which at least 80% of the organoids were visible. The analysis was conducted on two HistoBricks containing 15, respectively 16, organoids. Although the HistoBrick contains a marking at one edge to localize the top right corner, one well can be left empty for experiments in which the embedding matrix cannot easily be visualized during the microscopy step. This improves the certainty of organoid re-identification, because during sectioning and microscopy the sections and the images thereof can be rotated. FIJI was employed in this study for image modification and the quantification of the OS + IS area within images. For images displayed in the figures, brightness and contrast were adjusted for optimal visualization of the signal. Each channel underwent independent analysis. OS + IS area measurements To obtain single, fully filled regions of interest (ROIs), we applied the following procedure separately for each channel. For area masking, first, the “Fire” lookup table (LUT) for color mapping was applied to enhance visualization (Supplementary Fig. c, 2nd panel). A color threshold was manually set for each channel such that the organoid including the OS was above, and the background below threshold. Thresholds were channel-specific and kept constant within one embedding material in Fig. and across all conditions in Fig. . A binary mask was defined as the pixels above threshold (Supplementary Fig. c, 3rd panel). Image parts of the retinal organoids below threshold were included into the mask (Supplementary Fig. d), while parts outside the retinal organoid with values above threshold, including non-retinal, unstructured or disrupted organoid regions were manually separated from the mask (Supplementary Fig. e) using the “Overlay brush” tool. Subsequently, the “Fill Holes” command was executed to include all pixels inside the retinal organoid into the mask. The “Analyze Particles” function was applied to define a region of interest (ROI) by the binary mask with the largest area, corresponding to the retinal organoid (Supplementary Fig. c, right panel, Supplementary Fig. d,e, right panels). Area measurements of the ROIs were performed using the “Analyze Particles” command. To measure the OS + IS area, first, a single, fully filled ROI for the Peanut agglutinin (PNA) signal was determined (PNA ROI, Supplementary Fig. b,c). Second, a single, fully filled ROI for the Hoechst signal was determined (Hoechst ROI, Supplementary Fig. b,c). Finally, to calculate the OS + IS area, the area of the Hoechst ROI was subtracted from the area of the PNA ROI. The used FIJI macro is available in the Supplementary Information. Quantification of OS + IS thickness To quantify outer segments independent of organoid size and cross-section area, we estimated the average thickness [12pt]{minimal} $$$$ of the outer and inner segments labeled by PNA (“OS + IS thickness”) by [12pt]{minimal} $$ = }}{ } - }}{ }$$ where PNA area denotes the area of the PNA ROI and Hoechst area denotes the area of the Hoechst ROI. [12pt]{minimal} $$$$ provides the exact solution for circular organoids with uniform OS + IS segments and an average thickness estimate for organoids with less regular shape and OS + IS segments. [12pt]{minimal} $$$$ can also be expressed as a ratio between the OS + IS area and the Hoechst area, providing an intuitive interpretation of their relationship: [12pt]{minimal} $$ 2} = 2}= ^{2}) }{r 2}= + ^{2}}{2r} ,$$ with radius [12pt]{minimal} $$r$$ of the Hoechst area, radius [12pt]{minimal} $$r+$$ of the PNA area, and [12pt]{minimal} $$ r$$ . Retinal organoid culture and fixation for embedding Retinal organoids were generated from induced pluripotent stem cells as previously described in Cowan et al . and Spirig et al . , . Retinal organoids were generated from the induced pluripotent stem cell line 01F49i-N-B7 (IOBi001-A (RRID:CVCL_C1TR) described in Cowan et al . . Experiments in this study were performed using organoids aged between 30 and 98 weeks. Organoids were fixed using 4% paraformaldehyde (PFA) (Merck, #1.00496) for 4 h at 4 °C. The samples were washed three times for 10 min each in 1 × PBS. Then the organoids were incubated in 30% sucrose (Millipore, #84100) in 1 × PBS until sunken to the bottom of the tube for cryopreservation and stored at − 80 °C until embedding. Statistical analysis Graphpad Prism (Prism10 for MacOS, Version 10.3.1 (464), August 21, 2024) was used for statistical analysis. To test whether the OS + IS thickness [12pt]{minimal} $$$$ labeled by PNA differed significantly for the results shown in Figs. h and d, Welch’s ANOVA was used to compare the different groups followed by posthoc Dunnett’s test. For the posthoc test in Fig. h, we compared all conditions to the conventional gelatine embedding. For the graph in Fig. d, we included data from Fig. as controls. We used the PEGDA-gelatine HistoBrick data as positive controls and the OCT data as negative controls. For the posthoc test, we compared all against negative control and all against positive control. To determine, whether OS + IS thickness [12pt]{minimal} $$$$ decreases over time, we plotted [12pt]{minimal} $$$$ values of the organoids at different developmental ages against their age in weeks. We then calculated the nonparametric Spearman rank correlation. The P value and r are represented in the figure. Multiplicity adjusted P values are indicated as asterisks in the figures as follows: P > 0.05 (ns), P ≤ 0.05 , P ≤ 0.001 , P ≤ 0.001 . Only significant P values were indicated on the graphs, the other comparisons were not significant. Each dot on the graphs represents the results of one organoid. Rheological measurements The gelatine solution, PEGDA-gelatine solution and a 8 v% PEGDA + 10 wt% sucrose solution were measured by pipetting 1 ml of the solution into a rheometer (Anton Paar, MCR 702e). For oscillation testing, the geometry used was a PP-40 with a gap of 0.5 mm. The normal force was set at 0 N so that the gap would adjust to maintain this force during the phase transition. A ramp from 37 to 20 °C (5 °C/min) followed by an isotherm at 20 °C for 60 min was performed for the gelatine solution and the PEGDA-gelatine solution. For the PEGDA solution, a ramp from 37 to 20 °C (5 °C/min) followed by an isotherm at 20 °C for 10 min was performed. The isotherm at 20 °C was shortened for the PEGDA solution because it showed stable behavior. During these 2 intervals, the storage Modulus (G’) and the loss Modulus (G’’) as well as the complex viscosity were recorded at a strain of 1% and 1 Hz. Prior to this experiment, an amplitude sweep test was performed (data not shown) to evaluate the Linear viscoelastic range (LVE range). 1% strain was selected within the LVE range. For rotational testing, the geometry used was a CP-50-1 (diameter of 50 mm with a 1° angle and cone truncation of 0.045 mm). Dynamic viscosity was recorded at 37 °C from a shear rate of 0.1–1000 s −1 . The exploitable data range is from 10 to 1000 s −1 . The rotational tests were performed three times for each sample. The dynamic viscosities of the different solutions were calculated as the average of the three performed tests over the whole shear rate range. The silicone mold was designed similarly to our previous work . The diameter of the wells was increased to 2.4 mm to accommodate large retinal organoids. The outer dimensions of the HistoBrick were set to 20.7× 34.7 mm to contain 16 wells (Supplementary Fig. a). The mold was 3D printed in silicone “TrueSil Shore 50A” by the company Spectoplast. The schematic of the HistoBrick fabrication is depicted in Fig. . The HistoBrick 3D design file is available upon request to the corresponding authors. HistoBrick well-sizes and dimensions can be optimized for different organoid types. Gelatine gel well plate In a first approach, the silicone mold was placed with the pillars facing downward on a petri dish. A gelatine solution (7.5 wt% gelatine (Sigma Aldrich, #G1890), 10 wt% sucrose (Sigma Aldrich, #84,100) in 1 × PBS) was warmed at 37 °C and pipetted inside the silicone mold. The gelatine solution was crosslinked at 4 °C for 1 h. Afterwards, the mold was flipped and the gel well plate was released by gently lifting it from below. The obtained gel well plate was stored in a closed box with damp paper at 4 °C until use. PEGDA-based gel well plate In a second approach the silicone mold was placed with the pillars facing downward on a petri dish. The PEGDA-gelatine solution was prepared by mixing 8 v% PEGDA (Mw = 700, Sigma Aldrich, #455008), 2.5 wt% gelatine (Sigma Aldrich, #G1890), 10 wt% sucrose (Sigma Aldrich, #84100) and 0.05 wt% LAP (Sigma Aldrich, #900889) in 1 × PBS. The solution was pipetted inside the silicone mold and crosslinked by a 30 s exposure to 365 nm wavelength light with 35 mW/cm power (total illumination dose = 1.05 J/s*cm). The UV light was shined from above the mold. Afterward, the silicone mold was vertically lifted to release the gel well plate. The obtained gel well plate was stored in a closed box with humid paper at 4 °C until use. For the material selection, additional HistoBricks were prepared using a solution of 8 v% PEGDA Mw = 700, 10 wt% sucrose, 0.05 wt% LAP in 1 × PBS and 10 v% PEGDA Mw = 700, 10 wt% sucrose, 0.05 wt% LAP in 1 × PBS with the same crosslinking procedure. In a first approach, the silicone mold was placed with the pillars facing downward on a petri dish. A gelatine solution (7.5 wt% gelatine (Sigma Aldrich, #G1890), 10 wt% sucrose (Sigma Aldrich, #84,100) in 1 × PBS) was warmed at 37 °C and pipetted inside the silicone mold. The gelatine solution was crosslinked at 4 °C for 1 h. Afterwards, the mold was flipped and the gel well plate was released by gently lifting it from below. The obtained gel well plate was stored in a closed box with damp paper at 4 °C until use. In a second approach the silicone mold was placed with the pillars facing downward on a petri dish. The PEGDA-gelatine solution was prepared by mixing 8 v% PEGDA (Mw = 700, Sigma Aldrich, #455008), 2.5 wt% gelatine (Sigma Aldrich, #G1890), 10 wt% sucrose (Sigma Aldrich, #84100) and 0.05 wt% LAP (Sigma Aldrich, #900889) in 1 × PBS. The solution was pipetted inside the silicone mold and crosslinked by a 30 s exposure to 365 nm wavelength light with 35 mW/cm power (total illumination dose = 1.05 J/s*cm). The UV light was shined from above the mold. Afterward, the silicone mold was vertically lifted to release the gel well plate. The obtained gel well plate was stored in a closed box with humid paper at 4 °C until use. For the material selection, additional HistoBricks were prepared using a solution of 8 v% PEGDA Mw = 700, 10 wt% sucrose, 0.05 wt% LAP in 1 × PBS and 10 v% PEGDA Mw = 700, 10 wt% sucrose, 0.05 wt% LAP in 1 × PBS with the same crosslinking procedure. When the gelatine embedding material was used, the gel well plate was left to equilibrate at room temperature for 1 h. We refer to the liquid hydrogel solution filling the wells of the gel well plate as liquid embedding matrix. Before organoid transfer into the gel well plate, PFA-fixed, sucrose-cryopreserved retinal organoids were incubated in liquid embedding matrix of the same embedding material as the gel well plate for 15 min at 37 °C on a warming plate to enhance organoid-embedding matrix interaction. Incubation can also be performed in an incubator at 37 °C. The organoids were manually transferred with 20–50 µL of liquid embedding matrix to the gel well plate with a 200 µL tip pre-coated with 2% BSA (Sigma Aldrich, #A2153) in deionized water, to prevent organoids from sticking. The end of the tip was cut with a razor blade to create a larger aperture to avoid damaging the organoids. The gel well plate can be pre-wet by adding 20 µL of PBS in each well to minimize air bubble trapping. After the transfer, the wells were filled up with the same liquid embedding matrix used during the incubation of the organoids. Organoids were aligned onto a unique plane by centrifuging the filled gel well plate for 20 s at 200×g. The filled gel well plate underwent a second crosslinking (as previously described) to fix the position of the organoids at the bottom of the wells. The HistoBrick blocks were trimmed on all four sides so that three sections would later fit on the glass slides during cryosectioning. Finally, the HistoBrick was fixed on a labelled cardboard with a drop of OCT and snap frozen for 2 min by submersion in isopentane (Sigma Aldrich, #277258) cooled down to − 40 °C using dry ice pellets. The frozen HistoBricks were wrapped in aluminum foil to avoid drying and stored at − 80 °C until cryosectioning. Unless specified otherwise, the steps were performed using the same solutions and timings as for the gel well plate fabrication and organoid transfer. Gelatine and PEGDA-gelatine cryoblock The respective solution was pipetted inside a plastic weighing boat or an embedding mold and crosslinked to form a base layer of 2–4 mm thickness. During this time, retinal organoids were incubated in their respective embedding solutions for 15 min at 37 °C. Then, they were transferred one by one and placed on the base layer, followed by a second crosslinking. Finally, liquid embedding material was pipetted on top of the organoids to fill up the weighing boat and crosslinked. The blocks were trimmed with a scalpel to leave 1–2 mm space between the organoids and the block edge. The samples were then fixed on a labelled cardboard with a drop of OCT, snap frozen in isopentane at − 40 °C and stored at − 80 °C. OCT-based cryoblock OCT (Sakura, #4583) was poured inside a plastic weigh boat and crosslinked by freezing on dry ice pellets to form a base layer. The organoids were then transferred from their storage solution (without incubation) and placed on the base layer. Additional OCT was poured on the organoids, the blocks were frozen on dry ice and stored wrapped in aluminum foil at − 80 °C. The respective solution was pipetted inside a plastic weighing boat or an embedding mold and crosslinked to form a base layer of 2–4 mm thickness. During this time, retinal organoids were incubated in their respective embedding solutions for 15 min at 37 °C. Then, they were transferred one by one and placed on the base layer, followed by a second crosslinking. Finally, liquid embedding material was pipetted on top of the organoids to fill up the weighing boat and crosslinked. The blocks were trimmed with a scalpel to leave 1–2 mm space between the organoids and the block edge. The samples were then fixed on a labelled cardboard with a drop of OCT, snap frozen in isopentane at − 40 °C and stored at − 80 °C. OCT (Sakura, #4583) was poured inside a plastic weigh boat and crosslinked by freezing on dry ice pellets to form a base layer. The organoids were then transferred from their storage solution (without incubation) and placed on the base layer. Additional OCT was poured on the organoids, the blocks were frozen on dry ice and stored wrapped in aluminum foil at − 80 °C. The HistoBrick cryoblocks and the conventionally embedded cryoblocks were stored at − 20 °C overnight to acclimate to sectioning temperature. They were then mounted on a cryostat holder with OCT. If necessary, HistoBrick cryoblocks were again trimmed to fit 3 sections per slide, and conventional cryoblocks were trimmed to fit 8–12 sections per slide. Then they were sectioned in thin slices of 20 µm or 25 µm and placed on SuperFrost ® /Plus glass slides (Biosystems, #85-0911-00 or Epredia, #K5800AMNZ72). For results shown in Figs. , , Supplementary Figs. , and consecutive Histobrick sections were placed one section per glass slide on 5 slides. Afterwards, a second section was placed on each of the slides, followed by a third section, again starting from the first slide. After filling the first 5 slides, a second slide series of 5 slides was prepared using the same procedure. In this manner each slide contains sections from different organoid cutting depths, increasing the likelihood that there will be at least one usable section per organoid. The slides were stored at − 20 °C or − 80 °C until staining. To localize the organoids, slides were stained with hematoxylin and eosin. The slides were equilibrated and dried at room temperature for 1 h, then rehydrated in deionized water for 10 min. The slides were then incubated in Harris Hematoxylin (Harris Hematoxylin: Biosystems, #3873.2500) for 5 min and washed with running tap water. They were then differentiated for a few seconds in 1% acid-alcohol (Absolute alcohol: 7:10, VWR #20820.362, Hydrochloric acid 37%: 0.1:10, Sigma Aldrich, #30721 and H2O MiliQ 2.9:10) and washed under running tap water for 10 min. The slides were incubated in Eosine-Phloxine solution for 1 min (Eosin: 1:100, Sigma Aldrich, #E4382; Phloxine: 1:100, Sigma Aldrich, #P2759) and washed a last time. Finally, the slides were mounted with Eukitt (Sigma Aldrich, #03989) and dried overnight at room temperature. The slides were equilibrated and dried at room temperature for 1 h, then rehydrated in 1 × PBS for 10 min. The sections were blocked with 180 µL of 10% NDS blocking solution (10% Normal Donkey Serum (Sigma Aldrich, S30-M), 1% BSA (Sigma Aldrich, #A2153), 0.5% TritonX (Sigma Aldrich, #T9284), 0.1% sodium azide (Biosciences, #786-299), 1 × PBS) for 1 h. The blocking solution was removed, and the primary antibodies were added (ARR3: 1:400, Sigma Aldrich, #HPA063129; PNA: 1:1200, Sigma Aldrich #L6135; Rhodopsin: 1:400, Sigma Aldrich, #R5403; SOX9: 1:200, R&D systems, #AF3075; MiTF: 1:500, Exalpha, #X2398M; MAP2: 1:400, Sigma Aldrich, #AB5622; OneCut2: 1:100, R&D systems, # AF6294; Cralbp: 1:200, Abcam, #ab15051; Trpm1: 1:200, Sigma Aldrich, #HPA014779; Bassoon: 1:500, Enzo, # SAP7F407, L-M Opsin: 1:100, Merck, # AB5405). The primary antibodies were diluted in a 3% NDS solution (3% Normal Donkey Serum, 1% BSA, 0.5% TritonX, 0.1% sodium azide, 1 × PBS), 180 µL was added to each slide. The sections were covered with parafilm and incubated with the primary antibodies overnight at 4 °C in a humidified chamber. Afterwards, the slides were washed 3 times with PBS-T (1 × PBS, 0.05% TritonX) for 10 min each. The secondary antibodies (Hoechst 33342: 1:2000, Invitrogen, #H3570; anti-Mouse 488: 1:500, Invitrogen, #A32766; Streptavidin 555: 1:400, Invitrogen, #S32355; anti-Rabbit 640: 1:500, Invitrogen, #A32795; anti-Goat 488: 1:500, Invitrogen, #A32814; anti-Mouse 568: 1:500, Invitrogen, #A10037; anti-Sheep 488: 1:500, Invitrogen, # A11015) were diluted in 3% NDS and the sections were incubated for 2 h with 180 µL antibody solution at 4 °C. The slides were washed twice with PBS-T for 10 min each and once with 1 × PBS for 10 min. As a last step, the slides were mounted with Prolong Gold (Invitrogen, #P36934), coverslipped, and dried overnight at room temperature, sealed with nail polish (Electron Microscopy Sciences, #72180) and imaged. Images of the H&E-stained sections in Figs. and were acquired with an Olympus VS200 slide scanner. An overview image of the entire glass slide was acquired at 10x (UplaXapo, NA = 0.40) magnification in brightfield. Immunostained sections were acquired with a confocal spinning disc microscope (Olympus IXplore SpinSR10). Different magnifications were used depending on the desired image (4x: UplanSApo, NA = 0.16; 10x: UplanSApo, NA = 0.40; 20x: UplanSApo, NA = 0.75; 40x: UPlanSApo, NA = 1.05 Sil). For higher throughput of image acquisition and in Supplementary Fig. , a slide scanner was used (Zeiss Axio Scan.Z1) with 10 × magnification (Plan-Apochromate, NA = 0.45). OS + IS area measurements were performed on images obtained with the Zeiss Axio Scan.Z1 slide scanner. For each individual organoid, the section containing the highest quality outer segments (criteria: abundant OS with as little disruption as possible, intact organoid section) was manually selected for analysis. Analysis of the vertical alignment of the human retinal organoids was performed on consecutive H&E-stained sections on a slide series by manually counting the number of organoids present on each section. For each HistoBrick, we determined the number of sections in which at least 80% of the organoids were visible. The analysis was conducted on two HistoBricks containing 15, respectively 16, organoids. Although the HistoBrick contains a marking at one edge to localize the top right corner, one well can be left empty for experiments in which the embedding matrix cannot easily be visualized during the microscopy step. This improves the certainty of organoid re-identification, because during sectioning and microscopy the sections and the images thereof can be rotated. FIJI was employed in this study for image modification and the quantification of the OS + IS area within images. For images displayed in the figures, brightness and contrast were adjusted for optimal visualization of the signal. Each channel underwent independent analysis. To obtain single, fully filled regions of interest (ROIs), we applied the following procedure separately for each channel. For area masking, first, the “Fire” lookup table (LUT) for color mapping was applied to enhance visualization (Supplementary Fig. c, 2nd panel). A color threshold was manually set for each channel such that the organoid including the OS was above, and the background below threshold. Thresholds were channel-specific and kept constant within one embedding material in Fig. and across all conditions in Fig. . A binary mask was defined as the pixels above threshold (Supplementary Fig. c, 3rd panel). Image parts of the retinal organoids below threshold were included into the mask (Supplementary Fig. d), while parts outside the retinal organoid with values above threshold, including non-retinal, unstructured or disrupted organoid regions were manually separated from the mask (Supplementary Fig. e) using the “Overlay brush” tool. Subsequently, the “Fill Holes” command was executed to include all pixels inside the retinal organoid into the mask. The “Analyze Particles” function was applied to define a region of interest (ROI) by the binary mask with the largest area, corresponding to the retinal organoid (Supplementary Fig. c, right panel, Supplementary Fig. d,e, right panels). Area measurements of the ROIs were performed using the “Analyze Particles” command. To measure the OS + IS area, first, a single, fully filled ROI for the Peanut agglutinin (PNA) signal was determined (PNA ROI, Supplementary Fig. b,c). Second, a single, fully filled ROI for the Hoechst signal was determined (Hoechst ROI, Supplementary Fig. b,c). Finally, to calculate the OS + IS area, the area of the Hoechst ROI was subtracted from the area of the PNA ROI. The used FIJI macro is available in the Supplementary Information. To quantify outer segments independent of organoid size and cross-section area, we estimated the average thickness [12pt]{minimal} $$$$ of the outer and inner segments labeled by PNA (“OS + IS thickness”) by [12pt]{minimal} $$ = }}{ } - }}{ }$$ where PNA area denotes the area of the PNA ROI and Hoechst area denotes the area of the Hoechst ROI. [12pt]{minimal} $$$$ provides the exact solution for circular organoids with uniform OS + IS segments and an average thickness estimate for organoids with less regular shape and OS + IS segments. [12pt]{minimal} $$$$ can also be expressed as a ratio between the OS + IS area and the Hoechst area, providing an intuitive interpretation of their relationship: [12pt]{minimal} $$ 2} = 2}= ^{2}) }{r 2}= + ^{2}}{2r} ,$$ with radius [12pt]{minimal} $$r$$ of the Hoechst area, radius [12pt]{minimal} $$r+$$ of the PNA area, and [12pt]{minimal} $$ r$$ . Retinal organoids were generated from induced pluripotent stem cells as previously described in Cowan et al . and Spirig et al . , . Retinal organoids were generated from the induced pluripotent stem cell line 01F49i-N-B7 (IOBi001-A (RRID:CVCL_C1TR) described in Cowan et al . . Experiments in this study were performed using organoids aged between 30 and 98 weeks. Organoids were fixed using 4% paraformaldehyde (PFA) (Merck, #1.00496) for 4 h at 4 °C. The samples were washed three times for 10 min each in 1 × PBS. Then the organoids were incubated in 30% sucrose (Millipore, #84100) in 1 × PBS until sunken to the bottom of the tube for cryopreservation and stored at − 80 °C until embedding. Graphpad Prism (Prism10 for MacOS, Version 10.3.1 (464), August 21, 2024) was used for statistical analysis. To test whether the OS + IS thickness [12pt]{minimal} $$$$ labeled by PNA differed significantly for the results shown in Figs. h and d, Welch’s ANOVA was used to compare the different groups followed by posthoc Dunnett’s test. For the posthoc test in Fig. h, we compared all conditions to the conventional gelatine embedding. For the graph in Fig. d, we included data from Fig. as controls. We used the PEGDA-gelatine HistoBrick data as positive controls and the OCT data as negative controls. For the posthoc test, we compared all against negative control and all against positive control. To determine, whether OS + IS thickness [12pt]{minimal} $$$$ decreases over time, we plotted [12pt]{minimal} $$$$ values of the organoids at different developmental ages against their age in weeks. We then calculated the nonparametric Spearman rank correlation. The P value and r are represented in the figure. Multiplicity adjusted P values are indicated as asterisks in the figures as follows: P > 0.05 (ns), P ≤ 0.05 , P ≤ 0.001 , P ≤ 0.001 . Only significant P values were indicated on the graphs, the other comparisons were not significant. Each dot on the graphs represents the results of one organoid. The gelatine solution, PEGDA-gelatine solution and a 8 v% PEGDA + 10 wt% sucrose solution were measured by pipetting 1 ml of the solution into a rheometer (Anton Paar, MCR 702e). For oscillation testing, the geometry used was a PP-40 with a gap of 0.5 mm. The normal force was set at 0 N so that the gap would adjust to maintain this force during the phase transition. A ramp from 37 to 20 °C (5 °C/min) followed by an isotherm at 20 °C for 60 min was performed for the gelatine solution and the PEGDA-gelatine solution. For the PEGDA solution, a ramp from 37 to 20 °C (5 °C/min) followed by an isotherm at 20 °C for 10 min was performed. The isotherm at 20 °C was shortened for the PEGDA solution because it showed stable behavior. During these 2 intervals, the storage Modulus (G’) and the loss Modulus (G’’) as well as the complex viscosity were recorded at a strain of 1% and 1 Hz. Prior to this experiment, an amplitude sweep test was performed (data not shown) to evaluate the Linear viscoelastic range (LVE range). 1% strain was selected within the LVE range. For rotational testing, the geometry used was a CP-50-1 (diameter of 50 mm with a 1° angle and cone truncation of 0.045 mm). Dynamic viscosity was recorded at 37 °C from a shear rate of 0.1–1000 s −1 . The exploitable data range is from 10 to 1000 s −1 . The rotational tests were performed three times for each sample. The dynamic viscosities of the different solutions were calculated as the average of the three performed tests over the whole shear rate range. Supplementary Information. |
Effect of behavior modification combined with health belief model education on adherence to skin moisturizing care in patients with psoriasis vulgaris | 45255d78-2f03-4587-b3ac-527c1e6e62ac | 11685565 | Patient Education as Topic[mh] | Psoriasis vulgaris is an immune-mediated condition characterized by an interplay of genetic, environmental, and immunological factors. Aberrant activation of dendritic cells and T-cells leads to overproduction of cytokines such as IL-17 and TNF-alpha, driving keratinocyte proliferation and inflammation. This disrupts skin barrier function, exacerbating symptoms recurring, chronic inflammatory skin condition characterized like scaling, erythema, and itching . Psoriasis vulgaris affects approximately 2–3% of the global population, with pathogenesis involving complex immune mechanisms. Current treatments primarily target cytokine pathways to control inflammation, but patient adherence to topical therapies remains a significant barrier to successful outcomes. Research has demonstrated that compromised skin barrier function is a pivotal factor in the development of psoriasis vulgaris . Addressing skin barrier repair can effectively alleviate clinical symptoms and prevent disease recurrence . Consequently, alongside conventional treatments, clinical emphasis should be placed on repairing the skin barrier . Skin moisturizing care plays a crucial role in restoring skin barrier function , , encompassing skin cleansing and the application of moisturizing emollients . Appropriate skin cleansing helps prevent exacerbation of skin barrier damage , improves patient comfort by removing scales, and eliminates harmful bacteria on the skin surface, thereby preventing infections and averting psoriasis recurrence. The application of moisturizing emollients post-cleansing aids in repairing the skin barrier, softening skin keratin, increasing skin moisture content, and reducing skin itching. Additionally, it enhances the effectiveness of other topical medications by improving skin permeability , . Research indicates that cognitive and behavioral levels of skin moisturizing care in patients with psoriasis vulgaris are relatively low . While domestic dermatologists acknowledge the importance of daily skin moisturizing care, enhancing patients’ adherence to it remains a significant challenge. Many patients exhibit insufficient compliance due to the time-consuming nature of daily moisturizer application, the inconvenience involved, and the need for assistance in applying to certain body parts. Cost concerns also contribute to reluctance, as the prolonged use of moisturizers for skin barrier repair may seem costly with unclear benefits. Weather-related factors, such as discomfort in winter and summer, further contribute to patients deviating from recommended cognitive behavior. It has been noted that a more positive attitude in patients correlates with a stronger willingness to act, leading to increased adherence to daily skin moisturizing care . Behavior modification has been shown to significantly improve adherence to care routines in chronic conditions, fostering sustained behavioral changes . In response to this cognitive behavior, efforts should be directed towards improving patients’ disease awareness, correcting poor behaviors, enhancing compliance, and alleviating symptoms. Objectives This study primarily explores the intervention effect of behavior modification combined with health belief model education on moisturizing care adherence in patients with psoriasis vulgaris, as detailed below.
This study primarily explores the intervention effect of behavior modification combined with health belief model education on moisturizing care adherence in patients with psoriasis vulgaris, as detailed below.
Participants We enrolled 108 patients diagnosed with psoriasis vulgaris admitted between November 1, 2022, and October 30, 2023, as our study participants. Participants were recruited as inpatient from the Dermatology Department of Hangzhou Third People’s Hospital during regular clinic hours. During the two-month intervention, educational sessions and behavior modification activities were conducted biweekly. These interventions were delivered by physician primarily face-to-face during inpatient, supplemented by WeChat online contact during outpatient. Ethics approval has been obtained from the Ethics Committee of The Third People’s Hospital of Hangzhou (No. 2022KAO49). Informed consent was obtained from all participants, who were over 18 years old and cognitively unimpaired, ensuring no involvement of legal guardians. All methods were performed in accordance with the relevant guidelines and regulations and was subject to the supervision of the Ethics Committee of The Third People’s Hospital of Hangzhou. Inclusion criteria comprised: (1) meeting the diagnostic criteria for psoriasis vulgaris ; (2) age over 18 years; (3) maintaining good mental health with no cognitive impairment; (4) providing informed and signed consent for participation. Exclusion criteria encompassed: (a) presence of other primary diseases, or allergies induced by food or drugs resulting in itching; (b) concurrent acute infectious diseases; (c) mental disorders or cognitive impairment hindering cooperation with the intervention; (d) involvement in other ongoing studies. Utilizing a randomized numerical table method, we allocated 54 cases to both the experimental and control groups. Study design and setting Both groups received standard medications for both internal and external application, specifically loratadine tablets at a dosage of 10 mg per administration, once daily, for a course of 10 days. The treatment’s effectiveness was evaluated after completing one course. Additionally, throughout the treatment period, external applications included pimecrolimus cream, compound flumethasone cream, vitamin E allantoin cream (I), and black light therapy. (i) Control group. The control group underwent standard nursing interventions, which encompassed the distribution of informational brochures and the implementation of health education. Patients were guided to adhere to the prescribed medication regimen, adopt a balanced diet, rectify detrimental habits, engage in suitable physical activities, sustain a positive psychological state, and refrain from scratching. This comprehensive intervention spanned a duration of two months. (ii) Experimental group. The intervention involved a combination of behavior modification, health belief model education, and nursing care. Initial data collection aimed to gather comprehensive information on patients, including their general details, disease history, treatment, skin condition, educational background, and family situation. The study’s purpose was then explained to establish rapport and trust with patients. Health belief model education and nursing: Assessment of Health Beliefs: Evaluate patients’ understanding and coping attitudes towards psoriasis, identifying adverse behaviors, psychological responses, willpower, comprehension, and treatment adherence. Health Education: Deliver personalized education using diverse methods like visuals, videos, and case studies to enhance patients’ knowledge of psoriasis. Assess their informational needs and encourage strict adherence to medical advice, providing guidance on diet and exercise. Psychological Intervention: Understand patients’ psychological states, address negative emotions promptly, and foster positive health beliefs. 2. Behavior modification: Awareness Training: Communicate the significance of moisturizing care, utilizing visuals of aggravated symptoms due to poor compliance. Issue self-supervision cards, instructing patients to record moisturizing details to reinforce the importance. Behavioral Assessment: Analyze self-monitoring card data to infer causes of weak compliance, patient psychology during care, and factors influencing symptom severity. Guide patients to identify and monitor symptoms for timely control. Behavioral Training: Instruct patients to enact counteractive changes when signs of psychological and behavioral abnormalities emerge, using visuals to counter misconceptions. Control of Change: Implement supervision and reminders to correct behavior when changes fall short. Results Evaluation and Cognitive Consolidation: Assess patients’ cognitive mastery and encourage continued training for those with successful implementation. For those with insufficient implementation, identify influencing factors and provide targeted guidance to strengthen health beliefs, enhance awareness, and improve implementation. The entire intervention spanned two months. Observational indicators Itch assessment: Before and after the intervention, the Pruritus Scale was employed to evaluate patients’ itching severity, frequency, and affected area, using a total score of 10. Higher scores indicated more pronounced itching. Self-efficacy evaluation: The General Self-Efficacy Scale (GSES) was administered pre- and post-intervention to gauge patients’ self-efficacy. This Chinese version, comprising 10 items, used a scoring system of 1–4 points for “never,” “occasionally,” “often,” and “always.” Elevated scores denoted enhanced self-efficacy. Medication adherence assessment: The MORISKY Medication Adherence Questionnaire (MMAS8) was utilized before and after the intervention. The scale, featuring 8 items, included true/false questions with scores of 0 for correct and 1 for incorrect responses. The eighth question offered a range of responses, from “all the time” (0 points) to “never” (1 point). Scores were inversely correlated with patient adherence, with a maximum score of 8. Quality of life measurement: The Dermatologic Quality of Life Index (DLQI) was employed pre- and post-intervention. This index encompassed six dimensions—symptom perception, daily life, leisure and recreation, work and study, interpersonal relationships, and treatment—summing up to a total of 30 points. Higher scores indicated a diminished quality of life. Statistical methods The data analysis employed SPSS 23.0 statistical software. Descriptive statistics were used for measurement data, expressed as mean ± standard deviation (mean ± SD), with the t-test applied. Count data were presented as the number of cases and percentage, analyzed using the χ2 test. For ordinal data, the rank and test method was utilized. Ordinal data were analyzed using the Mann-Whitney U test for non-parametric comparisons. Statistical significance was determined at P < 0.05.
We enrolled 108 patients diagnosed with psoriasis vulgaris admitted between November 1, 2022, and October 30, 2023, as our study participants. Participants were recruited as inpatient from the Dermatology Department of Hangzhou Third People’s Hospital during regular clinic hours. During the two-month intervention, educational sessions and behavior modification activities were conducted biweekly. These interventions were delivered by physician primarily face-to-face during inpatient, supplemented by WeChat online contact during outpatient. Ethics approval has been obtained from the Ethics Committee of The Third People’s Hospital of Hangzhou (No. 2022KAO49). Informed consent was obtained from all participants, who were over 18 years old and cognitively unimpaired, ensuring no involvement of legal guardians. All methods were performed in accordance with the relevant guidelines and regulations and was subject to the supervision of the Ethics Committee of The Third People’s Hospital of Hangzhou. Inclusion criteria comprised: (1) meeting the diagnostic criteria for psoriasis vulgaris ; (2) age over 18 years; (3) maintaining good mental health with no cognitive impairment; (4) providing informed and signed consent for participation. Exclusion criteria encompassed: (a) presence of other primary diseases, or allergies induced by food or drugs resulting in itching; (b) concurrent acute infectious diseases; (c) mental disorders or cognitive impairment hindering cooperation with the intervention; (d) involvement in other ongoing studies. Utilizing a randomized numerical table method, we allocated 54 cases to both the experimental and control groups.
Both groups received standard medications for both internal and external application, specifically loratadine tablets at a dosage of 10 mg per administration, once daily, for a course of 10 days. The treatment’s effectiveness was evaluated after completing one course. Additionally, throughout the treatment period, external applications included pimecrolimus cream, compound flumethasone cream, vitamin E allantoin cream (I), and black light therapy. (i) Control group. The control group underwent standard nursing interventions, which encompassed the distribution of informational brochures and the implementation of health education. Patients were guided to adhere to the prescribed medication regimen, adopt a balanced diet, rectify detrimental habits, engage in suitable physical activities, sustain a positive psychological state, and refrain from scratching. This comprehensive intervention spanned a duration of two months. (ii) Experimental group. The intervention involved a combination of behavior modification, health belief model education, and nursing care. Initial data collection aimed to gather comprehensive information on patients, including their general details, disease history, treatment, skin condition, educational background, and family situation. The study’s purpose was then explained to establish rapport and trust with patients. Health belief model education and nursing: Assessment of Health Beliefs: Evaluate patients’ understanding and coping attitudes towards psoriasis, identifying adverse behaviors, psychological responses, willpower, comprehension, and treatment adherence. Health Education: Deliver personalized education using diverse methods like visuals, videos, and case studies to enhance patients’ knowledge of psoriasis. Assess their informational needs and encourage strict adherence to medical advice, providing guidance on diet and exercise. Psychological Intervention: Understand patients’ psychological states, address negative emotions promptly, and foster positive health beliefs. 2. Behavior modification: Awareness Training: Communicate the significance of moisturizing care, utilizing visuals of aggravated symptoms due to poor compliance. Issue self-supervision cards, instructing patients to record moisturizing details to reinforce the importance. Behavioral Assessment: Analyze self-monitoring card data to infer causes of weak compliance, patient psychology during care, and factors influencing symptom severity. Guide patients to identify and monitor symptoms for timely control. Behavioral Training: Instruct patients to enact counteractive changes when signs of psychological and behavioral abnormalities emerge, using visuals to counter misconceptions. Control of Change: Implement supervision and reminders to correct behavior when changes fall short. Results Evaluation and Cognitive Consolidation: Assess patients’ cognitive mastery and encourage continued training for those with successful implementation. For those with insufficient implementation, identify influencing factors and provide targeted guidance to strengthen health beliefs, enhance awareness, and improve implementation. The entire intervention spanned two months.
Itch assessment: Before and after the intervention, the Pruritus Scale was employed to evaluate patients’ itching severity, frequency, and affected area, using a total score of 10. Higher scores indicated more pronounced itching. Self-efficacy evaluation: The General Self-Efficacy Scale (GSES) was administered pre- and post-intervention to gauge patients’ self-efficacy. This Chinese version, comprising 10 items, used a scoring system of 1–4 points for “never,” “occasionally,” “often,” and “always.” Elevated scores denoted enhanced self-efficacy. Medication adherence assessment: The MORISKY Medication Adherence Questionnaire (MMAS8) was utilized before and after the intervention. The scale, featuring 8 items, included true/false questions with scores of 0 for correct and 1 for incorrect responses. The eighth question offered a range of responses, from “all the time” (0 points) to “never” (1 point). Scores were inversely correlated with patient adherence, with a maximum score of 8. Quality of life measurement: The Dermatologic Quality of Life Index (DLQI) was employed pre- and post-intervention. This index encompassed six dimensions—symptom perception, daily life, leisure and recreation, work and study, interpersonal relationships, and treatment—summing up to a total of 30 points. Higher scores indicated a diminished quality of life.
The data analysis employed SPSS 23.0 statistical software. Descriptive statistics were used for measurement data, expressed as mean ± standard deviation (mean ± SD), with the t-test applied. Count data were presented as the number of cases and percentage, analyzed using the χ2 test. For ordinal data, the rank and test method was utilized. Ordinal data were analyzed using the Mann-Whitney U test for non-parametric comparisons. Statistical significance was determined at P < 0.05.
Population characteristics of the experimental and control groups In the experimental group, gender distribution included 34 male cases and 20 female cases, with an average age of (50.96 ± 3.14) years. The duration of the disease was (4.33 ± 1.21) months, and disease severity was classified as follows: 3 cases as mild, 16 cases as moderate, 31 cases as severe, and 4 cases as very severe. Disease severity was assessed using the Pruritus Scale, a validated scoring system measuring patients’ itching severity, frequency, and affected area, using a total score of 10. Educational background comprised 13 cases with primary school education, 16 cases with junior high school education, 8 cases ith high school or junior college education, and 16 cases with college or bachelor’s degrees. Additionally, there were 1 case pursuing postgraduate studies. Similarly, in the control group, there were 34 male cases and 20 female cases, with an average age of (51.58 ± 3.25) years. The duration of the disease was (4.25 ± 1.19) months, and disease severity included 3 cases classified as mild, 17 cases as moderate, and 30 cases as severe, and 4 cases as very severe. Educational backgrounds consisted of 13 cases with primary education, 15 cases with junior high school education, 9 cases with high school or junior college education, 16 cases with college or bachelor’s degrees, and 1 case (1.85%) pursuing postgraduate studies. No statistically significant differences were observed in the general clinical data between the two groups ( P > 0.05). The demographic and baseline characteristics of the experimental and control groups summarized in Table . Comparison of skin itching before and after the intervention in the two groups is shown in Table . Comparison of GSES scores between the two groups before and after the intervention See Table . Comparison of adherence between the two groups See Table . Comparison of DLQI scores between the two groups before and after the intervention see Table . Effect size calculations To calculate the effect sizes for the statistically significant differences in the study, using metrics Cohen’s d for continuous data (e.g., comparing means between two groups) for categorical outcomes. Here’s how effect sizes can be interpreted: [12pt]{minimal}
$${}:{}\,=\,0.{}$$ [12pt]{minimal}
$${}:{}\,=\,0.{}$$ [12pt]{minimal}
$${}:{}\,=\,0.{}$$ Pooled standard deviation (pooled SD) : The pooled SD accounts for the variability within both the pre- and post-intervention measurements and is calculated as: [12pt]{minimal}
$${}\;{}\;{}\; { - 1) SD_{1}^{2}+({n_2} - 1) SD_{2}^{2}}}{{{n_1}+{n_2} - 2}}}$$ Effect size is calculated as: d = Mean difference/Pooled SD. Effect size calculations for Pruritus scores (degree of itchiness) For “degree of itchiness” (pre- and post-intervention), Cohen’s d for the experimental and control groups calculation as below: Experimental group: Pooled standard deviation: 0.38. Cohen’s d: 5.16 (very large effect). Control group: Pooled standard deviation: 0.41. Cohen’s d: 2.67 (very large effect). Therefore, these values indicate that the intervention had a strong impact on reducing itchiness in both groups, with a more pronounced effect in the experimental group. 2. Effect size calculations for GSES scores Experimental group: Pooled SD: 3.73. Cohen’s d: 3.76 (very large effect). Control group: Pooled SD: 3.58. Cohen’s d: 3.06 (very large effect). Therefore, both the experimental and control groups exhibited significant improvements in GSES scores following the intervention, with very large effect sizes. However, the experimental group showed a slightly stronger effect. This suggests that the combined intervention of behavior modification and Health Belief Model education had a significant impact on self-efficacy in psoriasis patients. 3. DLQI dimensions Symptom perception Experimental group: Pooled SD: 0.66. Cohen’s d: 2.60 (very large effect). Control group: Pooled SD: 0.76. Cohen’s d: 1.33 (large effect). Therefore, these values indicate that the intervention had larger effect in the experimental group on Symptom Perception compared with control group. (b) Everyday lives Experimental group: Pooled SD: 0.67. Cohen’s d: 1.04 (large effect). Control group: Pooled SD: 0.54. Cohen’s d: 0.71 (medium to large effect). Therefore, these values indicate that the intervention had larger effect in the experimental group on everyday lives compared with control group. (c) Entertainment Experimental group: Pooled SD: 0.48. Cohen’s d: 1.75 (large effect). Control group: Pooled SD: 0.45. Cohen’s d: 1.28 (large effect). Therefore, these values indicate that the intervention had large effect in both groups on entertainment, with a more pronounced effect in the experimental group. (d) Work-based learning Experimental group: Pooled SD: 0.50. Cohen’s d: 0.91 (large effect). Control group: Pooled SD: 0.43. Cohen’s d: 0.99 (large effect). Therefore, these values indicate that the intervention had large effect in both groups, with a slightly more pronounced effect in the control group. (e) Interpersonal relationships Experimental group: Pooled SD: 0.52. Cohen’s d: 1.71 (large effect). Control group: Pooled SD: 0.47. Cohen’s d: 1.11 (large effect). Therefore, these values indicate that the intervention had large effect in both groups on interpersonal relationships, with a more pronounced effect in the experimental group. (f) Treatment Experimental group: Pooled SD: 0.42. Cohen’s d: 2.50 (very large effect). Control group: Pooled SD: 0.41. Cohen’s d: 1.96 (large effect). Therefore, these values indicate that the intervention had very large effect in experimental groups on treatment, with a more pronounced effect compared to the control group. Overall, the experimental group consistently showed larger effects than the control group, with the most pronounced differences observed in symptom perception, treatment, and interpersonal relationships. The control group also showed large effects, particularly in work-based learning, but the experimental group generally benefitted more from the intervention across the various categories measured. The intervention had a substantial and positive impact on both groups, but the experimental group generally experienced stronger effects, particularly in the areas of symptom perception, treatment, and interpersonal relationships.
In the experimental group, gender distribution included 34 male cases and 20 female cases, with an average age of (50.96 ± 3.14) years. The duration of the disease was (4.33 ± 1.21) months, and disease severity was classified as follows: 3 cases as mild, 16 cases as moderate, 31 cases as severe, and 4 cases as very severe. Disease severity was assessed using the Pruritus Scale, a validated scoring system measuring patients’ itching severity, frequency, and affected area, using a total score of 10. Educational background comprised 13 cases with primary school education, 16 cases with junior high school education, 8 cases ith high school or junior college education, and 16 cases with college or bachelor’s degrees. Additionally, there were 1 case pursuing postgraduate studies. Similarly, in the control group, there were 34 male cases and 20 female cases, with an average age of (51.58 ± 3.25) years. The duration of the disease was (4.25 ± 1.19) months, and disease severity included 3 cases classified as mild, 17 cases as moderate, and 30 cases as severe, and 4 cases as very severe. Educational backgrounds consisted of 13 cases with primary education, 15 cases with junior high school education, 9 cases with high school or junior college education, 16 cases with college or bachelor’s degrees, and 1 case (1.85%) pursuing postgraduate studies. No statistically significant differences were observed in the general clinical data between the two groups ( P > 0.05). The demographic and baseline characteristics of the experimental and control groups summarized in Table . Comparison of skin itching before and after the intervention in the two groups is shown in Table . Comparison of GSES scores between the two groups before and after the intervention See Table . Comparison of adherence between the two groups See Table . Comparison of DLQI scores between the two groups before and after the intervention see Table .
To calculate the effect sizes for the statistically significant differences in the study, using metrics Cohen’s d for continuous data (e.g., comparing means between two groups) for categorical outcomes. Here’s how effect sizes can be interpreted: [12pt]{minimal}
$${}:{}\,=\,0.{}$$ [12pt]{minimal}
$${}:{}\,=\,0.{}$$ [12pt]{minimal}
$${}:{}\,=\,0.{}$$ Pooled standard deviation (pooled SD) : The pooled SD accounts for the variability within both the pre- and post-intervention measurements and is calculated as: [12pt]{minimal}
$${}\;{}\;{}\; { - 1) SD_{1}^{2}+({n_2} - 1) SD_{2}^{2}}}{{{n_1}+{n_2} - 2}}}$$ Effect size is calculated as: d = Mean difference/Pooled SD. Effect size calculations for Pruritus scores (degree of itchiness) For “degree of itchiness” (pre- and post-intervention), Cohen’s d for the experimental and control groups calculation as below: Experimental group: Pooled standard deviation: 0.38. Cohen’s d: 5.16 (very large effect). Control group: Pooled standard deviation: 0.41. Cohen’s d: 2.67 (very large effect). Therefore, these values indicate that the intervention had a strong impact on reducing itchiness in both groups, with a more pronounced effect in the experimental group. 2. Effect size calculations for GSES scores Experimental group: Pooled SD: 3.73. Cohen’s d: 3.76 (very large effect). Control group: Pooled SD: 3.58. Cohen’s d: 3.06 (very large effect). Therefore, both the experimental and control groups exhibited significant improvements in GSES scores following the intervention, with very large effect sizes. However, the experimental group showed a slightly stronger effect. This suggests that the combined intervention of behavior modification and Health Belief Model education had a significant impact on self-efficacy in psoriasis patients. 3. DLQI dimensions Symptom perception Experimental group: Pooled SD: 0.66. Cohen’s d: 2.60 (very large effect). Control group: Pooled SD: 0.76. Cohen’s d: 1.33 (large effect). Therefore, these values indicate that the intervention had larger effect in the experimental group on Symptom Perception compared with control group. (b) Everyday lives Experimental group: Pooled SD: 0.67. Cohen’s d: 1.04 (large effect). Control group: Pooled SD: 0.54. Cohen’s d: 0.71 (medium to large effect). Therefore, these values indicate that the intervention had larger effect in the experimental group on everyday lives compared with control group. (c) Entertainment Experimental group: Pooled SD: 0.48. Cohen’s d: 1.75 (large effect). Control group: Pooled SD: 0.45. Cohen’s d: 1.28 (large effect). Therefore, these values indicate that the intervention had large effect in both groups on entertainment, with a more pronounced effect in the experimental group. (d) Work-based learning Experimental group: Pooled SD: 0.50. Cohen’s d: 0.91 (large effect). Control group: Pooled SD: 0.43. Cohen’s d: 0.99 (large effect). Therefore, these values indicate that the intervention had large effect in both groups, with a slightly more pronounced effect in the control group. (e) Interpersonal relationships Experimental group: Pooled SD: 0.52. Cohen’s d: 1.71 (large effect). Control group: Pooled SD: 0.47. Cohen’s d: 1.11 (large effect). Therefore, these values indicate that the intervention had large effect in both groups on interpersonal relationships, with a more pronounced effect in the experimental group. (f) Treatment Experimental group: Pooled SD: 0.42. Cohen’s d: 2.50 (very large effect). Control group: Pooled SD: 0.41. Cohen’s d: 1.96 (large effect). Therefore, these values indicate that the intervention had very large effect in experimental groups on treatment, with a more pronounced effect compared to the control group. Overall, the experimental group consistently showed larger effects than the control group, with the most pronounced differences observed in symptom perception, treatment, and interpersonal relationships. The control group also showed large effects, particularly in work-based learning, but the experimental group generally benefitted more from the intervention across the various categories measured. The intervention had a substantial and positive impact on both groups, but the experimental group generally experienced stronger effects, particularly in the areas of symptom perception, treatment, and interpersonal relationships.
Zhang Li et al.‘s survey results suggested that patients with psoriasis vulgaris generally held positive attitudes toward skin moisturizing care . They recognized its significance and expressed a willingness to engage in it for an extended period. While psoriasis patients demonstrated relatively high content knowledge about skin moisturizing care, it reflected a lower-level knowledge change. Procedural knowledge, specifically understanding how to establish and implement skin moisturizing care behaviors, scored relatively low. This aspect necessitates attention in health education and promotion, as procedural knowledge is pivotal for behavioral change , . Many patients perceive emollients as non-medicinal with no therapeutic effect, resulting in haphazard usage. To address this, healthcare professionals should provide detailed guidance on the frequency, dosage, and integration with other topical medications when advising patients on skin moisturizing care to enhance their cognitive understanding . In this study, health education played a crucial role in helping patients comprehend moisturizing care knowledge, enhancing their self-care abilities, and promoting self-health management. Studies have highlighted the significant association between self-health management ability and adherence, considering adherence as the foundation of self-health management . Key health education messages included elucidating the role of moisturizing care, emphasizing frequency, timing, and dosage, contributing positively to enhancing patients’ self-health management ability. Post-intervention, the experimental group exhibited higher compliance than the control group, and their General Self-Efficacy Scale (GSES) score was also higher. This indicates that health belief model education guided by behavior modification has the potential to improve adherence to skin moisturizing care in patients with psoriasis vulgaris. Possible contributing factors include: (1) Health belief model education enhancing patients’ awareness of disease severity and cooperation with nursing care; (2) behavior modification targeting the correction of poor behavioral habits and changing procedural knowledge. Self-efficacy, as a personal belief in one’s behavioral ability, plays a crucial role in determining whether individuals can achieve their set goals. Adjusting self-efficacy levels can effectively promote behavioral change . Therefore, behavior modification, when combined with health belief cognitive preaching, serves as a holistic approach to assisting patients in achieving positive changes in their health behaviors. Based on the data analysis, the following statistically significant differences were identified. After the intervention, the Pruritus Scale significantly decreased in the experimental group compared to the control group. The reduction in Pruritus Scale in the experimental group was notably greater than in the control group ( P < 0.05), indicating that the combined behavioral intervention and health belief education effectively alleviated itching. Both groups showed improvements in GSES scores post-intervention, but the experimental group experienced a significantly larger increase than the control group ( P < 0.05). This demonstrates that the intervention significantly enhanced patients’ confidence in managing their care. The experimental group had significantly lower MORISKY Medication Adherence scores after the intervention compared to the control group, reflecting improved adherence in medication timing, frequency, and a reduction in self-discontinuation ( P < 0.05). These results suggest that the intervention was effective in enhancing treatment compliance. Post-intervention, the DLQI scores were significantly lower across all dimensions in the experimental group compared to the control group, including symptom perception, daily life, leisure activities, work and study, interpersonal relationships, and treatment ( P < 0.05). This indicates a substantial improvement in the overall quality of life for the experimental group. The significant reduction in Dermatology Life Quality Index (DLQI) scores among the experimental group underscores the potential for combining behavioral education with clinical treatments. However, further multicenter studies are warranted to establish generalizability across diverse populations. These statistically significant differences highlight that the combined behavioral modification and health belief education intervention was more effective in improving key outcomes—itching severity, self-efficacy, medication adherence, and quality of life—compared to standard care. In China, there is a lack of widespread awareness about psoriasis, resulting in insufficient knowledge among patients. This knowledge gap hinders the scientific and effective prevention and control of psoriasis, leading to a notable increase in recurrence rates and adversely impacting patients’ physical and mental well-being . Existing studies highlight that common clinical symptoms of psoriasis, such as itching, plaques, and papules, often prompt patients to scratch their skin, causing skin breakouts and negatively affecting their external appearance . Evaluating quality of life becomes crucial in clinical research, encompassing both physiological and psychological aspects of patient interventions. A study conducted by Long et al. involved behavioral habit reversal corrective intervention for psoriasis patients, revealing significant relief in patients’ itchiness and a notable improvement in overall quality of life. This underscores the vital role of behavioral interventions in enhancing the well-being of psoriasis patients. The health belief model serves as a primary theoretical framework for elucidating health-related behaviors from a social psychological perspective. This model has demonstrated satisfactory results in various medical conditions , . Its application in understanding and addressing health behaviors is particularly relevant, providing valuable insights for improving patient outcomes in psoriasis management. The results of this study demonstrate statistically significant improvements in self-efficacy and quality of life among patients with psoriasis vulgaris who underwent interventions combining behavioral modification and health belief model education. These findings carry several important implications for clinical practice and the management of chronic dermatological conditions. The integration of behavior modification and health education could be adapted for other chronic skin disorders, such as eczema or atopic dermatitis, where adherence to skin care regimens is similarly critical. The focus on enhancing self-efficacy through tailored education and behavior tracking can address common barriers to compliance, such as misunderstanding of treatments or behavioral resistance. By addressing psychological and behavioral aspects, this intervention moves beyond symptom-focused treatments. The approach is aligned with patient-centered care, emphasizing the importance of empowering individuals to actively participate in their disease management. Improved adherence and symptom management may reduce the long-term costs associated with more severe disease progression, hospital readmissions, or advanced treatments. This is particularly relevant for healthcare systems looking to optimize resource use. The results of effect size demonstrate that the intervention had a larger and more pronounced effect on the experimental group across most domains, particularly in areas related to symptom perception, treatment, and interpersonal relationships. The very large effect sizes observed in the experimental group suggest that the intervention was particularly effective in these areas, offering significant improvements in how participants perceived and dealt with their symptoms, as well as their overall treatment outcomes. However, the control group also showed large effects in several domains, such as work-based learning and entertainment, indicating that even without the specific intervention, participants in the control group experienced positive changes. This suggests that factors outside of the intervention may have contributed to some of the observed effects, and the control group’s improvements should not be overlooked. For instance, natural changes over time or other external influences could have played a role.While the results are promising, particularly in the experimental group, further research is needed to explore the specific mechanisms behind these improvements. In conclusion, the intervention appears to be effective across a variety of domains, with particularly large effects on symptom perception, treatment, and interpersonal relationships in the experimental group. The control group also showed positive results, but the experimental group consistently experienced stronger outcomes, suggesting that the intervention had a meaningful and impactful role in the study. While the combination of behavior modification and health belief education demonstrates significant benefits, it is important to note that these methods are supplementary and do not replace internationally recommended pharmacological therapies. Future studies should integrate these educational approaches with standard international protocols to validate their broader applicability. Limitations of study While the study offers valuable insights into the effectiveness of an educational approach incorporating behavioral modification and the Health Belief Model in individuals with psoriasis vulgaris, several limitations warrant consideration. The study’s single-center design and a relatively modest sample size of 108 patients may limit the generalizability of findings. The short two-month intervention period may not capture sustained long-term effects on adherence to skin moisturizing care, itch severity, self-efficacy, and quality of life. The reliance on self-reported measures, such as the Pruritus Scale, General Self-Efficacy Scale, Medication Adherence Scale (MORISKY), and Dermatologic Quality of Life Index (DLQI), introduces the potential for response bias. Additionally, the absence of a robust placebo or sham intervention in the control group raises concerns about the possibility of a placebo effect influencing outcomes. Future research with a larger, more diverse sample, an extended intervention period, and rigorous control conditions could provide a more comprehensive understanding of the long-term impact and generalizability of the educational interventions studied.
While the study offers valuable insights into the effectiveness of an educational approach incorporating behavioral modification and the Health Belief Model in individuals with psoriasis vulgaris, several limitations warrant consideration. The study’s single-center design and a relatively modest sample size of 108 patients may limit the generalizability of findings. The short two-month intervention period may not capture sustained long-term effects on adherence to skin moisturizing care, itch severity, self-efficacy, and quality of life. The reliance on self-reported measures, such as the Pruritus Scale, General Self-Efficacy Scale, Medication Adherence Scale (MORISKY), and Dermatologic Quality of Life Index (DLQI), introduces the potential for response bias. Additionally, the absence of a robust placebo or sham intervention in the control group raises concerns about the possibility of a placebo effect influencing outcomes. Future research with a larger, more diverse sample, an extended intervention period, and rigorous control conditions could provide a more comprehensive understanding of the long-term impact and generalizability of the educational interventions studied.
The study’s findings indicate that, following the intervention, the Dermatologic Quality of Life Index (DLQI) score in the experimental group surpassed that of the control group. Additionally, the experimental group exhibited an improvement in skin itching compared to the control group. These results suggest that the combination of behavior modification and health belief model education can effectively enhance the quality of life and alleviate itching symptoms in psoriasis patients. This underscores the positive significance of behavior modification in ameliorating clinical symptoms among patients. Several factors may contribute to these outcomes: Health belief model education plays a role in enhancing patients’ understanding of emollient use and fostering greater cooperation in nursing care. Behavior modification targets the correction of patients’ unfavorable behavioral habits, resulting in improved compliance with skin moisturizing care. The integration of behavior modification and health belief model education effectively addresses non-compliance and promotes positive changes in moisturizing care habits, thus reducing the occurrence and progression of undesirable symptoms. These findings underscore the critical role of behavior modification and health belief model education in improving not only adherence to care but also enhancing the quality of life for psoriasis patients by reducing pruritus and emotional distress.
|
Filgotinib: A Clinical Pharmacology Review | 860dce36-5735-439f-adf0-1f065bf60df4 | 9249714 | Pharmacology[mh] |
The Janus kinase (JAK)-signal transducer and activator of transcription (JAK-STAT) pathway is a key driver of rheumatoid arthritis (RA), Crohn’s disease (CD), and ulcerative colitis (UC) by way of mediating the response to multiple proinflammatory cytokines and cellular growth factors . Inhibition of the JAK-STAT pathway has demonstrated efficacy in immune-mediated diseases and has been identified as a novel therapeutic target for the treatment of RA , CD, and UC . Preferential inhibition of JAK1 modulates a subset of proinflammatory cytokines within the JAK-STAT pathway that differ from those inhibited by JAK2 or JAK3 and that could improve the benefit–risk profile in comparison with that of pan-JAK inhibitors . Filgotinib is a second-generation preferential JAK1 inhibitor. In biochemical assays, filgotinib preferentially inhibited the activity of JAK1 and showed > 5-fold higher potency for JAK1 over JAK2, JAK3, and tyrosine kinase 2 . Filgotinib has an active primary metabolite—GS-829845 (previously G254445)—which has a similar JAK1 selectivity profile but 10-fold lower potency in human whole blood assay than the parent compound . In humans, exposure to this primary metabolite is 16- to 20-fold higher than to the parent filgotinib . Filgotinib exposure following 200 mg once-daily (QD) dosing covers the half-maximal inhibitory concentration (IC 50 ) for JAK1 in human whole blood . Filgotinib is approved for the treatment of RA in Europe, the UK, and Japan but has not received approval for any indication in the USA. Clinical trials have shown that filgotinib has an onset of action as early as week 2, sustained efficacy, and a proven safety profile in patients with moderate-to-severe RA who have had an inadequate response to methotrexate, are biologic naïve, or are biologic experienced . Filgotinib has also demonstrated efficacy and safety in combination with methotrexate in a phase III study in patients with active RA who had limited or no prior exposure to methotrexate . Filgotinib is approved for the treatment of UC in Europe, the UK, and Japan, having demonstrated early symptomatic relief and sustained efficacy in a phase III trial in patients with UC who were biologic naïve or biologic experienced . Filgotinib also displayed a promising clinical profile in a phase II study in patients with CD, inducing clinical remission in significantly more patients compared with placebo .
Physicochemical Features Filgotinib [ N -(5-(4-((1,1-dioxidothiomorpholin-4-yl)methyl)phenyl)[1,2,4]triazole[1,5-α]pyridin-2-yl)cyclopropanecarboxamide(2Z)but-2-enedioate] is an orally bioavailable JAK inhibitor (Fig. ). It is categorized as a class II drug according to the Biopharmaceutics Classification System, based on its high permeability and low solubility . A study in human epithelial colorectal adenocarcinoma (Caco-2) cell monolayers showed an apparent permeability of 3.5 × 10 −6 cm/s, with an efflux ratio of 16.0. This was lower than that of the high-permeability reference substance, propranolol (permeability 22.7 × 10 −6 cm/s) . However, following oral administration of 100 mg [ 14 C]-filgotinib in a mass balance study, 87% of radioactivity was excreted in the urine, indicating that filgotinib permeability is high, with near complete absorption . In healthy subject studies, filgotinib was rapidly absorbed after single and repeated oral administration . Analytical Methods Plasma concentrations of filgotinib and its primary metabolite were determined simultaneously using a validated liquid chromatography–tandem mass spectrometry assay . JAK1 inhibition was investigated using fluorescence-activated cell-sorting analysis on blood samples, by measuring STAT1 phosphorylation (pSTAT1) in interleukin-6-stimulated blood . Similarly, JAK2 inhibition was investigated by measuring STAT5 phosphorylation (pSTAT5) to confirm selectivity in vivo.
Filgotinib [ N -(5-(4-((1,1-dioxidothiomorpholin-4-yl)methyl)phenyl)[1,2,4]triazole[1,5-α]pyridin-2-yl)cyclopropanecarboxamide(2Z)but-2-enedioate] is an orally bioavailable JAK inhibitor (Fig. ). It is categorized as a class II drug according to the Biopharmaceutics Classification System, based on its high permeability and low solubility . A study in human epithelial colorectal adenocarcinoma (Caco-2) cell monolayers showed an apparent permeability of 3.5 × 10 −6 cm/s, with an efflux ratio of 16.0. This was lower than that of the high-permeability reference substance, propranolol (permeability 22.7 × 10 −6 cm/s) . However, following oral administration of 100 mg [ 14 C]-filgotinib in a mass balance study, 87% of radioactivity was excreted in the urine, indicating that filgotinib permeability is high, with near complete absorption . In healthy subject studies, filgotinib was rapidly absorbed after single and repeated oral administration .
Plasma concentrations of filgotinib and its primary metabolite were determined simultaneously using a validated liquid chromatography–tandem mass spectrometry assay . JAK1 inhibition was investigated using fluorescence-activated cell-sorting analysis on blood samples, by measuring STAT1 phosphorylation (pSTAT1) in interleukin-6-stimulated blood . Similarly, JAK2 inhibition was investigated by measuring STAT5 phosphorylation (pSTAT5) to confirm selectivity in vivo.
Absorption and Bioavailability Filgotinib was rapidly absorbed following oral administration in healthy subjects, with maximum (peak) plasma concentrations ( C max ) reached within 1 to 3 h postdose (Table ). Filgotinib exposure (both C max and area under the plasma concentration–time curve [AUC]) increased dose proportionally over the dose range of 50 to 200 mg. Overall, inter-subject variability of AUC and C max at steady state was low to moderate (inter-subject coefficient of variation [CV%] range 16–44%) . With regards to the primary metabolite, plasma concentrations were detected within 30 min after single doses and reached a maximum 3 to 8 h postdose (Table ). As with the parent compound, C max and AUC increased dose proportionally within the 50 to 200 mg dose range. Inter-subject variability of AUC and C max of the metabolite at steady state was low (inter-subject CV% <26%). Primary metabolite exposures were on average 16- to 20-fold higher than filgotinib exposures . Various formulations have been used throughout the clinical development of filgotinib, all of which have demonstrated similar exposures, including the drug product used in the phase IIb studies (a hydrochloride trihydrate salt) and the maleate salt tablet formulation selected for use in the phase III studies and subsequently commercialized (Fig. ) . Effect of Food Filgotinib can be administered without regard to food intake, as demonstrated in an evaluation of the commercialized maleate salt tablet formulation. Effects of both a high-fat (approximately 800 calories with 50% from fat) and a low-fat (approximately 400 calories with 20% from fat) meal on the pharmacokinetics of a single 200 mg dose of filgotinib were compared with those of fasting conditions in healthy subjects . Slight decreases were noted for filgotinib C max , with both the high-fat (20%) and the low-fat (11%) meal versus fasting (Fig. a). Food intake delayed absorption of filgotinib, with a median time to reach C max following drug administration ( t max ) of 1 h while fasting, which increased to 2 h with a low-fat meal and to 3 h with a high-fat meal. These effects were marginal and not considered clinically relevant. Neither the high-fat nor the low-fat meal affected the pharmacokinetics of the primary metabolite (Fig. b) . This lack of effect was consistent with the high solubility of the maleate salt formulation in simulated intestinal fluids (data on file). Based on results from this study, the filgotinib tablet formulation was administered without regard to food in phase III studies. Distribution and Protein Binding Binding of filgotinib or its primary metabolite to plasma proteins was determined in vitro by equilibrium dialysis using [ 14 C]-filgotinib or cold compound. Protein binding of filgotinib was low in all evaluated species, with a value of 55–59% in humans and 32–70% in animal species. Binding of the primary metabolite to plasma proteins was similar to that of filgotinib, with values of 39–44% in humans and 29–55% in animal species (data on file). After oral administration of [ 14 C]-filgotinib to healthy male subjects, geometric mean whole blood-to-plasma total radioactivity ratios fluctuated between 0.9 and 1.1, suggesting some affinity of the radioactivity for red blood cell elements . Protein binding of filgotinib and its primary metabolite were unchanged in patients with moderate hepatic impairment compared with healthy controls (bound fraction 56–59% and 39–45%, respectively) . Data from two studies in healthy subjects and a proof-of-concept study in patients with RA were used to develop estimates for several filgotinib pharmacokinetic parameters. The apparent volume of distribution of filgotinib was 3.1 L (95% confidence interval [CI] 2.2–3.6) in the central compartment and 4.7 L (95% CI 4.4–5.0) in the peripheral compartment. These findings were comparable to the apparent volume of distribution of the metabolite compartment (4.4 L [95% CI 4.3–4.4]) . Metabolism Carboxylesterases (CESs) are the enzymes responsible for the formation of the primary metabolite . CESs belong to the serine hydrolase superfamily of enzymes, of which there are six CES isoforms. The two most important forms of CESs for drug metabolism in humans are CES1 and CES2 . Human CES2 is found mainly in the intestine and, to a lesser extent, in the liver, whereas CES1 is found mainly in the liver . Following administration of 100 mg [ 14 C]-filgotinib to healthy subjects, two metabolites were identified and quantified in plasma: the primary metabolite and its N -glucuronide derivative (Fig. ). The primary metabolite accounted for approximately 92% of the total radioactivity in plasma, and both metabolites accounted for 68.6% of the total radioactivity in urine. Cleavage of filgotinib into its primary metabolite released cyclopropanecarboxylic acid (CPCA). CPCA was recorded at very low concentrations (close to the lower limit of quantification of 5.00 ng/mL) in plasma and urine but formed conjugates with endogenous amino acids, such as carnitine, taurine, and glycine . Other minor filgotinib metabolites included M1 and M3 (accounting for 1.55% and 0.294% of the total radioactivity in urine, respectively) (Fig. ). Elimination Following administration of [ 14 C]-filgotinib 100 mg to healthy male subjects, 86.9% of the dose (filgotinib and its metabolites) was recovered from urine, which suggested that the majority of filgotinib and its metabolites were cleared by the kidneys . Urinary excretion was rapid, with approximately half of recovery achieved within 24 h of dosing. Eight metabolites were identified in the urine and seven in the feces. Similar to observations involving plasma, the primary metabolite and its N -glucuronide derivative were the most prevalent, representing 54.0% and 14.6% of the total radioactivity in urine, and 8.9% and 1.9% of the total radioactivity in feces, respectively. Each of the remaining metabolites identified in the urine and feces represented, on average, less than 2.2% and 0.3% of the total radioactivity, respectively. The identical metabolite pattern in urine and feces suggested some biliary excretion of filgotinib and its metabolites (data on file). Accumulation, Half-Life, and Steady-State Pharmacokinetics For filgotinib, no accumulation to steady state was observed after QD dosing, consistent with its relatively short half-life of 4.9 to 10.7 h (Tables and ). Steady state was reached on day 2 of dosing. After repeated dosing with filgotinib, plasma elimination of the primary metabolite displayed a monophasic pattern, with mean apparent terminal elimination half-life ( t ½ ) ranging between 19.6 and 27.3 h (Tables and ). Consistent with the terminal half-life, steady-state levels of the metabolite were achieved within 4 days, with an average 2-fold accumulation of the metabolite after QD dosing .
Filgotinib was rapidly absorbed following oral administration in healthy subjects, with maximum (peak) plasma concentrations ( C max ) reached within 1 to 3 h postdose (Table ). Filgotinib exposure (both C max and area under the plasma concentration–time curve [AUC]) increased dose proportionally over the dose range of 50 to 200 mg. Overall, inter-subject variability of AUC and C max at steady state was low to moderate (inter-subject coefficient of variation [CV%] range 16–44%) . With regards to the primary metabolite, plasma concentrations were detected within 30 min after single doses and reached a maximum 3 to 8 h postdose (Table ). As with the parent compound, C max and AUC increased dose proportionally within the 50 to 200 mg dose range. Inter-subject variability of AUC and C max of the metabolite at steady state was low (inter-subject CV% <26%). Primary metabolite exposures were on average 16- to 20-fold higher than filgotinib exposures . Various formulations have been used throughout the clinical development of filgotinib, all of which have demonstrated similar exposures, including the drug product used in the phase IIb studies (a hydrochloride trihydrate salt) and the maleate salt tablet formulation selected for use in the phase III studies and subsequently commercialized (Fig. ) .
Filgotinib can be administered without regard to food intake, as demonstrated in an evaluation of the commercialized maleate salt tablet formulation. Effects of both a high-fat (approximately 800 calories with 50% from fat) and a low-fat (approximately 400 calories with 20% from fat) meal on the pharmacokinetics of a single 200 mg dose of filgotinib were compared with those of fasting conditions in healthy subjects . Slight decreases were noted for filgotinib C max , with both the high-fat (20%) and the low-fat (11%) meal versus fasting (Fig. a). Food intake delayed absorption of filgotinib, with a median time to reach C max following drug administration ( t max ) of 1 h while fasting, which increased to 2 h with a low-fat meal and to 3 h with a high-fat meal. These effects were marginal and not considered clinically relevant. Neither the high-fat nor the low-fat meal affected the pharmacokinetics of the primary metabolite (Fig. b) . This lack of effect was consistent with the high solubility of the maleate salt formulation in simulated intestinal fluids (data on file). Based on results from this study, the filgotinib tablet formulation was administered without regard to food in phase III studies.
Binding of filgotinib or its primary metabolite to plasma proteins was determined in vitro by equilibrium dialysis using [ 14 C]-filgotinib or cold compound. Protein binding of filgotinib was low in all evaluated species, with a value of 55–59% in humans and 32–70% in animal species. Binding of the primary metabolite to plasma proteins was similar to that of filgotinib, with values of 39–44% in humans and 29–55% in animal species (data on file). After oral administration of [ 14 C]-filgotinib to healthy male subjects, geometric mean whole blood-to-plasma total radioactivity ratios fluctuated between 0.9 and 1.1, suggesting some affinity of the radioactivity for red blood cell elements . Protein binding of filgotinib and its primary metabolite were unchanged in patients with moderate hepatic impairment compared with healthy controls (bound fraction 56–59% and 39–45%, respectively) . Data from two studies in healthy subjects and a proof-of-concept study in patients with RA were used to develop estimates for several filgotinib pharmacokinetic parameters. The apparent volume of distribution of filgotinib was 3.1 L (95% confidence interval [CI] 2.2–3.6) in the central compartment and 4.7 L (95% CI 4.4–5.0) in the peripheral compartment. These findings were comparable to the apparent volume of distribution of the metabolite compartment (4.4 L [95% CI 4.3–4.4]) .
Carboxylesterases (CESs) are the enzymes responsible for the formation of the primary metabolite . CESs belong to the serine hydrolase superfamily of enzymes, of which there are six CES isoforms. The two most important forms of CESs for drug metabolism in humans are CES1 and CES2 . Human CES2 is found mainly in the intestine and, to a lesser extent, in the liver, whereas CES1 is found mainly in the liver . Following administration of 100 mg [ 14 C]-filgotinib to healthy subjects, two metabolites were identified and quantified in plasma: the primary metabolite and its N -glucuronide derivative (Fig. ). The primary metabolite accounted for approximately 92% of the total radioactivity in plasma, and both metabolites accounted for 68.6% of the total radioactivity in urine. Cleavage of filgotinib into its primary metabolite released cyclopropanecarboxylic acid (CPCA). CPCA was recorded at very low concentrations (close to the lower limit of quantification of 5.00 ng/mL) in plasma and urine but formed conjugates with endogenous amino acids, such as carnitine, taurine, and glycine . Other minor filgotinib metabolites included M1 and M3 (accounting for 1.55% and 0.294% of the total radioactivity in urine, respectively) (Fig. ).
Following administration of [ 14 C]-filgotinib 100 mg to healthy male subjects, 86.9% of the dose (filgotinib and its metabolites) was recovered from urine, which suggested that the majority of filgotinib and its metabolites were cleared by the kidneys . Urinary excretion was rapid, with approximately half of recovery achieved within 24 h of dosing. Eight metabolites were identified in the urine and seven in the feces. Similar to observations involving plasma, the primary metabolite and its N -glucuronide derivative were the most prevalent, representing 54.0% and 14.6% of the total radioactivity in urine, and 8.9% and 1.9% of the total radioactivity in feces, respectively. Each of the remaining metabolites identified in the urine and feces represented, on average, less than 2.2% and 0.3% of the total radioactivity, respectively. The identical metabolite pattern in urine and feces suggested some biliary excretion of filgotinib and its metabolites (data on file).
For filgotinib, no accumulation to steady state was observed after QD dosing, consistent with its relatively short half-life of 4.9 to 10.7 h (Tables and ). Steady state was reached on day 2 of dosing. After repeated dosing with filgotinib, plasma elimination of the primary metabolite displayed a monophasic pattern, with mean apparent terminal elimination half-life ( t ½ ) ranging between 19.6 and 27.3 h (Tables and ). Consistent with the terminal half-life, steady-state levels of the metabolite were achieved within 4 days, with an average 2-fold accumulation of the metabolite after QD dosing .
Age Age had a limited impact on filgotinib exposure. In a dedicated study of filgotinib pharmacokinetics in healthy elderly subjects, no pharmacokinetic differences were noted between subjects aged 65–74 years and those aged 40–50 years . In subjects aged ≥ 75 years, filgotinib exposure (AUC during a dosage interval [AUC τ ]) was 1.4-fold higher than in those aged 40−50 years, but C max , t ½ , and the amount of unchanged filgotinib excreted in urine were not altered. As with exposure to filgotinib, exposure (AUC τ ) to the primary metabolite was higher (by 1.3-fold) in subjects aged ≥ 75 years than in subjects aged 40–50 years after filgotinib 100 mg QD, with no change in the formation and elimination of the metabolite. This observation was supported by the constant metabolite-over-parent exposure ratio (from 18.4–19.4) over the entire age range. The relative differences in the pharmacokinetics of filgotinib and its primary metabolite between the two elderly age groups (65–75 and ≥ 75 years) are illustrated in Fig. c, d. Based on these data, it was concluded that age has no impact on the CESs involved in filgotinib metabolite formation ; however, because of a higher incidence of serious infections and limited clinical experience in patients with RA aged ≥ 75 years, caution is counselled when treating this population, and a starting filgotinib dose of 100 mg QD is recommended . Japanese Ethnicity The pharmacokinetics of filgotinib and its primary metabolite were compared between healthy Japanese and White subjects after 10 days of filgotinib 200 mg QD (Table ; Fig. c, d) . Filgotinib had a half-life of 6 h and 11 h in Japanese and White subjects, respectively, reaching steady-state plasma concentrations by day 2 in both populations. In Japanese and White subjects, the active metabolite reached higher plasma concentrations than filgotinib, consistent with its longer half-life (17–20 h). Overall exposures for filgotinib and its metabolite were similar in both groups. These data indicate that the pharmacokinetic/pharmacodynamic profile of filgotinib is comparable in Japanese and White subjects . Renal Impairment Both filgotinib and its primary metabolite contribute to the overall clinical efficacy of the molecule; therefore, in instances where intrinsic/extrinsic factors show some effect on filgotinib or metabolite pharmacokinetics, both parent and metabolite exposures can be combined into a single parameter, AUC eff (the sum of the AUC of filgotinib and its metabolite adjusted for their respective molecular weights and potencies), to more fully analyze whether dose adjustment is required . Mild renal impairment (estimated glomerular filtration rate [eGFR] 60 to <90 mL/min/1.73 m 2 ) had limited impact on filgotinib pharmacokinetics (AUC eff increased by 1.5-fold); thus, no dose adjustment was required in these patients . By contrast, moderate renal impairment (eGFR 30 to <60 mL/min/1.73 m 2 ) increased AUC eff by 2.0-fold, and severe renal impairment (eGFR 15 to <30 mL/min/1.73 m 2 ) increased AUC eff by 3.0-fold; as a result, a dose of filgotinib 100 mg QD is recommended in patients with RA with moderate or severe renal impairment . As shown in Fig. c, d, the observed differences in exposures across patients with renal impairment of varying severity were more pronounced for the metabolite, consistent with its primary excretion route (at least 50% in urine) . Filgotinib has not been studied in patients with end-stage renal disease (eGFR <15 mL/min/1.73 m 2 ), so its use is therefore not recommended in this population . Hepatic Impairment In a phase I study of patients with moderate hepatic impairment (Child–Pugh score B; n = 10), AUC from time zero to infinity (AUC ∞ ) for filgotinib and its primary metabolite increased by 1.6- and 1.2-fold, respectively, versus healthy controls ( n = 10), after a single dose of filgotinib 100 mg. Protein binding of filgotinib and its primary metabolite was unchanged (see Sect. 3.3). The absence of substantial differences in the pharmacokinetics of filgotinib and its primary metabolite was anticipated in individuals with mild-to-moderate hepatic impairment as filgotinib is metabolized by CES2, an enzyme located mainly in the intestine. Consequently, no dose adjustment of filgotinib is required in patients with mild-to-moderate hepatic impairment ; however, filgotinib has not been studied in patients with severe hepatic impairment (Child–Pugh score C) so is not recommended for use in this population . Disease State The steady-state pharmacokinetics of filgotinib and its primary metabolite after administration of filgotinib 200 mg QD were investigated using intensive pharmacokinetic analyses across phase III trials in patients with moderate-to-severe RA (FINCH 1–3; NCT02889796, NCT02873936, NCT02886728) and in patients with UC in the SELECTION study (NCT02914522) (Table 3). Summary pharmacokinetic parameters in patients with RA and UC were comparable to those reported for healthy subjects (Tables and ). Sex and Body Weight Although sex and weight have not been formally evaluated for potential effects on filgotinib pharmacokinetics in dedicated studies, they do not appear to have a clinically relevant effect on the pharmacokinetics of filgotinib or its metabolite .
Age had a limited impact on filgotinib exposure. In a dedicated study of filgotinib pharmacokinetics in healthy elderly subjects, no pharmacokinetic differences were noted between subjects aged 65–74 years and those aged 40–50 years . In subjects aged ≥ 75 years, filgotinib exposure (AUC during a dosage interval [AUC τ ]) was 1.4-fold higher than in those aged 40−50 years, but C max , t ½ , and the amount of unchanged filgotinib excreted in urine were not altered. As with exposure to filgotinib, exposure (AUC τ ) to the primary metabolite was higher (by 1.3-fold) in subjects aged ≥ 75 years than in subjects aged 40–50 years after filgotinib 100 mg QD, with no change in the formation and elimination of the metabolite. This observation was supported by the constant metabolite-over-parent exposure ratio (from 18.4–19.4) over the entire age range. The relative differences in the pharmacokinetics of filgotinib and its primary metabolite between the two elderly age groups (65–75 and ≥ 75 years) are illustrated in Fig. c, d. Based on these data, it was concluded that age has no impact on the CESs involved in filgotinib metabolite formation ; however, because of a higher incidence of serious infections and limited clinical experience in patients with RA aged ≥ 75 years, caution is counselled when treating this population, and a starting filgotinib dose of 100 mg QD is recommended .
The pharmacokinetics of filgotinib and its primary metabolite were compared between healthy Japanese and White subjects after 10 days of filgotinib 200 mg QD (Table ; Fig. c, d) . Filgotinib had a half-life of 6 h and 11 h in Japanese and White subjects, respectively, reaching steady-state plasma concentrations by day 2 in both populations. In Japanese and White subjects, the active metabolite reached higher plasma concentrations than filgotinib, consistent with its longer half-life (17–20 h). Overall exposures for filgotinib and its metabolite were similar in both groups. These data indicate that the pharmacokinetic/pharmacodynamic profile of filgotinib is comparable in Japanese and White subjects .
Both filgotinib and its primary metabolite contribute to the overall clinical efficacy of the molecule; therefore, in instances where intrinsic/extrinsic factors show some effect on filgotinib or metabolite pharmacokinetics, both parent and metabolite exposures can be combined into a single parameter, AUC eff (the sum of the AUC of filgotinib and its metabolite adjusted for their respective molecular weights and potencies), to more fully analyze whether dose adjustment is required . Mild renal impairment (estimated glomerular filtration rate [eGFR] 60 to <90 mL/min/1.73 m 2 ) had limited impact on filgotinib pharmacokinetics (AUC eff increased by 1.5-fold); thus, no dose adjustment was required in these patients . By contrast, moderate renal impairment (eGFR 30 to <60 mL/min/1.73 m 2 ) increased AUC eff by 2.0-fold, and severe renal impairment (eGFR 15 to <30 mL/min/1.73 m 2 ) increased AUC eff by 3.0-fold; as a result, a dose of filgotinib 100 mg QD is recommended in patients with RA with moderate or severe renal impairment . As shown in Fig. c, d, the observed differences in exposures across patients with renal impairment of varying severity were more pronounced for the metabolite, consistent with its primary excretion route (at least 50% in urine) . Filgotinib has not been studied in patients with end-stage renal disease (eGFR <15 mL/min/1.73 m 2 ), so its use is therefore not recommended in this population .
In a phase I study of patients with moderate hepatic impairment (Child–Pugh score B; n = 10), AUC from time zero to infinity (AUC ∞ ) for filgotinib and its primary metabolite increased by 1.6- and 1.2-fold, respectively, versus healthy controls ( n = 10), after a single dose of filgotinib 100 mg. Protein binding of filgotinib and its primary metabolite was unchanged (see Sect. 3.3). The absence of substantial differences in the pharmacokinetics of filgotinib and its primary metabolite was anticipated in individuals with mild-to-moderate hepatic impairment as filgotinib is metabolized by CES2, an enzyme located mainly in the intestine. Consequently, no dose adjustment of filgotinib is required in patients with mild-to-moderate hepatic impairment ; however, filgotinib has not been studied in patients with severe hepatic impairment (Child–Pugh score C) so is not recommended for use in this population .
The steady-state pharmacokinetics of filgotinib and its primary metabolite after administration of filgotinib 200 mg QD were investigated using intensive pharmacokinetic analyses across phase III trials in patients with moderate-to-severe RA (FINCH 1–3; NCT02889796, NCT02873936, NCT02886728) and in patients with UC in the SELECTION study (NCT02914522) (Table 3). Summary pharmacokinetic parameters in patients with RA and UC were comparable to those reported for healthy subjects (Tables and ).
Although sex and weight have not been formally evaluated for potential effects on filgotinib pharmacokinetics in dedicated studies, they do not appear to have a clinically relevant effect on the pharmacokinetics of filgotinib or its metabolite .
Carboxylesterase Inhibitors Human CES2 enzymes are the main isoforms responsible for the formation of the primary metabolite of filgotinib; they are localized mainly to the intestine and, to a lesser extent, the liver . These enzymes of the α / β -hydrolase family are abundant, with ubiquitous tissue-expression profiles. Filgotinib is also metabolized by CES1, which is predominantly expressed in the liver but to a lesser extent than by CES2 . In vitro inhibition of CES2 by medications including fenofibrate, carvedilol, diltiazem, and simvastatin has been demonstrated, although the clinical relevance of these interactions is currently unknown . In vitro characterization indicates that, even when CES2 is fully saturated, filgotinib metabolism is not completely abrogated, as CES1 can also form the primary metabolite . P-Glycoprotein Inhibitors and Inducers Both filgotinib and its primary metabolite are substrates of the xenobiotic compound transporter P-glycoprotein (P-gp) . The potential effect of the potent P-gp inhibitor itraconazole (200 mg single dose with 1 h pretreatment) on the pharmacokinetics of filgotinib (100 mg single dose) was evaluated in healthy subjects. Coadministration of filgotinib with itraconazole increased filgotinib AUC ∞ and C max by 45% and 64%, respectively but did not affect the AUC ∞ or C max of its primary metabolite (Fig. a, b). Itraconazole increased the combined AUC eff of filgotinib and its primary metabolite by 21%, so no dose adjustment of filgotinib was deemed necessary . The effect of the P-gp inducer rifampin (600 mg QD) on the pharmacokinetics of filgotinib (200 mg single dose) in healthy subjects was also evaluated. With coadministration, filgotinib AUC ∞ and C max were reduced by 27% and 26%, respectively, and the primary metabolite AUC ∞ and C max were reduced by 38% and 19%, respectively (Fig. a, b) . The combined AUC eff was reduced by 33%. Based on the combined AUC eff of filgotinib and its metabolite in the presence of these drugs, filgotinib dose adjustment was not deemed to be warranted with coadministration of P-gp inhibitors or inducers . Acid-Reducing Agents As patients with inflammatory diseases are likely to receive acid-reducing agents, such as histamine H 2 antagonists and proton pump inhibitors, potential drug interactions have been evaluated with representative medications in each of these classes . Administration of filgotinib (100 mg single dose) with simultaneously dosed omeprazole 40 mg (with 5-day pretreatment at 40 mg QD) or 2 h after famotidine 40 mg (with 4-day pretreatment at 40 mg twice daily [BID]) had no effect on the plasma exposure of filgotinib (AUC ∞ ) or its metabolite (AUC ∞ and C max ) (Fig. a, b) . Coadministration of omeprazole slightly reduced the C max of filgotinib by 27% (Fig. a) . The effects of acid-reducing agents on filgotinib pharmacokinetics are not considered clinically relevant, and filgotinib dose adjustment is not required when coadministering these agents .
Human CES2 enzymes are the main isoforms responsible for the formation of the primary metabolite of filgotinib; they are localized mainly to the intestine and, to a lesser extent, the liver . These enzymes of the α / β -hydrolase family are abundant, with ubiquitous tissue-expression profiles. Filgotinib is also metabolized by CES1, which is predominantly expressed in the liver but to a lesser extent than by CES2 . In vitro inhibition of CES2 by medications including fenofibrate, carvedilol, diltiazem, and simvastatin has been demonstrated, although the clinical relevance of these interactions is currently unknown . In vitro characterization indicates that, even when CES2 is fully saturated, filgotinib metabolism is not completely abrogated, as CES1 can also form the primary metabolite .
Both filgotinib and its primary metabolite are substrates of the xenobiotic compound transporter P-glycoprotein (P-gp) . The potential effect of the potent P-gp inhibitor itraconazole (200 mg single dose with 1 h pretreatment) on the pharmacokinetics of filgotinib (100 mg single dose) was evaluated in healthy subjects. Coadministration of filgotinib with itraconazole increased filgotinib AUC ∞ and C max by 45% and 64%, respectively but did not affect the AUC ∞ or C max of its primary metabolite (Fig. a, b). Itraconazole increased the combined AUC eff of filgotinib and its primary metabolite by 21%, so no dose adjustment of filgotinib was deemed necessary . The effect of the P-gp inducer rifampin (600 mg QD) on the pharmacokinetics of filgotinib (200 mg single dose) in healthy subjects was also evaluated. With coadministration, filgotinib AUC ∞ and C max were reduced by 27% and 26%, respectively, and the primary metabolite AUC ∞ and C max were reduced by 38% and 19%, respectively (Fig. a, b) . The combined AUC eff was reduced by 33%. Based on the combined AUC eff of filgotinib and its metabolite in the presence of these drugs, filgotinib dose adjustment was not deemed to be warranted with coadministration of P-gp inhibitors or inducers .
As patients with inflammatory diseases are likely to receive acid-reducing agents, such as histamine H 2 antagonists and proton pump inhibitors, potential drug interactions have been evaluated with representative medications in each of these classes . Administration of filgotinib (100 mg single dose) with simultaneously dosed omeprazole 40 mg (with 5-day pretreatment at 40 mg QD) or 2 h after famotidine 40 mg (with 4-day pretreatment at 40 mg twice daily [BID]) had no effect on the plasma exposure of filgotinib (AUC ∞ ) or its metabolite (AUC ∞ and C max ) (Fig. a, b) . Coadministration of omeprazole slightly reduced the C max of filgotinib by 27% (Fig. a) . The effects of acid-reducing agents on filgotinib pharmacokinetics are not considered clinically relevant, and filgotinib dose adjustment is not required when coadministering these agents .
Overall, filgotinib has a low drug–drug interaction potential . In vitro, filgotinib and its primary metabolite at clinically relevant concentrations did not meaningfully interact with most cytochrome P450 (CYP) enzymes or uridine 5ʹ-diphospho-glucuronosyltransferases and did not inhibit most key drug efflux transporters. One exception is organic cation transporter 2 (OCT2), which was inhibited by both filgotinib and its metabolite, with IC 50 values at least 11-fold higher than the C max values that filgotinib and its metabolite reached with 200 mg QD dosing . In vitro studies indicated that filgotinib and its metabolite may inhibit organic anion transporting polypeptides (OATP)1B1 and OATP1B3; however, a study in healthy subjects given filgotinib 200 mg QD alongside OATP probe substrates (atorvastatin and pravastatin/rosuvastatin) indicated that no dose adaptation was required for statins or other OATP substrates with filgotinib coadministration (see Sect. 6.4) . In vitro studies are inconclusive regarding the potential of filgotinib to induce CYP2B6 and induce or inhibit CYP1A2. Caution is therefore recommended when coadministering filgotinib with CYP1A2 substrates if the therapeutic index is narrow, such as is the case with warfarin . In the presence of a probe substrate at concentrations of up to 200 and 500 µM, maximum inhibition of P-gp was 3.0% and 0% with filgotinib and its primary metabolite, respectively (IC 50 >200 and >285 µM, respectively) [data on file]. At the same maximum concentrations, filgotinib and its primary metabolite inhibited the breast cancer resistance protein (BCRP) by 20.9% and 0%, respectively (IC 50 >200 and >285 µM, respectively) [data on file]. As such, filgotinib and its primary metabolite were deemed not to meaningfully inhibit P-gp or BCRP at clinically relevant concentrations. Overall, all in vitro and clinical data on drug-metabolizing enzymes and key drug transporters support coadministration of filgotinib with drugs commonly used for patients with inflammatory diseases without the need for dose adjustments. Midazolam To confirm and validate the in vitro results, the potential for interaction with CYP3A4 was investigated in healthy subjects using the sensitive CYP3A4 substrate midazolam (2 mg single dose) administered alone or after filgotinib 200 mg QD for 7 days. Neither the C max nor the AUC ∞ of midazolam were affected by coadministration of filgotinib (Fig. ). The point estimate pairwise comparisons for midazolam AUC ∞ and C max when coadministered with filgotinib versus midazolam alone were 1.1 (90% CI 1.0–1.2) and 1.0 (90% CI 0.9–1.1), respectively, meaning that no specific recommendation was warranted for coadministration of filgotinib with CYP3A4 substrates . Hormonal Contraceptives Since there may be mechanisms of enzyme induction/inhibition relevant to hormonal contraceptives that are presently poorly characterized (e.g., increased ethinyl estradiol due to sulfotransferase inhibition, which has been associated with thrombosis ), the effects of filgotinib (200 mg QD for 15 days) on hormonal contraceptives (single doses of 30 µg ethinyl estradiol/150 µg levonorgestrel before and after filgotinib) were evaluated in healthy female subjects . For both levonorgestrel and ethinyl estradiol, the percentage geometric least squares mean (GLSM) ratios and associated 90% CIs of C max and AUC ∞ were contained within the prespecified lack of interaction bounds (70–143%) when coadministered with filgotinib (Fig. ) . Based on these results, no dose adjustment was warranted for coadministration of filgotinib with hormonal contraceptives . Metformin The oral antidiabetic metformin is a substrate of OCT2 and of multidrug and toxin extrusion (MATE) transporter 1 (MATE1) and MATE2-K. As in vitro assessments indicated that filgotinib and its primary metabolite are OCT inhibitors (albeit at supratherapeutic concentrations) , potential drug interactions between metformin (850 mg single dose) and filgotinib 200 mg QD were assessed. Metformin percentage GLSM ratios and associated 90% CIs for AUC ∞ and C max were not altered by coadministration of filgotinib (Fig. ), meaning that filgotinib dose adjustment with concomitant use of metformin and other OCT2 and MATE transporter substrates was deemed unnecessary . Statins In vitro data indicated that filgotinib and its primary metabolite may be inhibitors of OATP1B1 and OATP1B3 at supratherapeutic filgotinib concentrations . A phase I study was conducted in healthy subjects given filgotinib 200 mg QD for 11 days alongside single doses of the OATP probe substrates atorvastatin (40 mg) and a combination of pravastatin (40 mg)/rosuvastatin (10 mg) . Of note, rosuvastatin is both an OATP and a BCRP substrate, and the combination of rosuvastatin and pravastatin as probes permits simultaneous measurement of the activity of the drug transporters OATP and BCRP in the presence of filgotinib . Although coadministration of filgotinib did not affect atorvastatin AUC ∞ , atorvastatin C max was reduced by 18% (Fig. ). Neither the AUC ∞ nor the C max of the atorvastatin metabolite, 2-hydroxy-atorvastatin, were affected by coadministration with filgotinib (Fig. ). Pravastatin AUC ∞ was unaffected by filgotinib; however, filgotinib increased pravastatin C max by 25% and increased the C max and AUC ∞ of rosuvastatin by 68% and 42%, respectively (Fig. ). None of these changes were considered to represent a clinically meaningful effect of filgotinib coadministration on statin exposure, and no dose adjustment is required for statins or other OATP or BCRP substrates . Methotrexate Methotrexate is an OAT1 and OAT3 substrate that is a standard of care for the treatment of RA. The potential for drug interaction with filgotinib was assessed during a phase IIa study that investigated filgotinib in the treatment of patients with active RA and an inadequate response to methotrexate . Patients received filgotinib 30, 75, 150, or 300 mg QD (or placebo) and continued their stable methotrexate dose (7.5–20 mg/kg weekly); pharmacokinetic data were pooled across filgotinib doses ( n = 17) because of the low patient numbers in individual dosing groups. The point estimate pairwise comparisons for methotrexate AUC τ and C max when coadministered with filgotinib versus methotrexate alone fell within the lack of interaction bounds (70–143%; Fig. ) . These data indicate that filgotinib does not significantly impact the pharmacokinetics of methotrexate and support the coadministration of filgotinib with methotrexate without dose adjustment.
To confirm and validate the in vitro results, the potential for interaction with CYP3A4 was investigated in healthy subjects using the sensitive CYP3A4 substrate midazolam (2 mg single dose) administered alone or after filgotinib 200 mg QD for 7 days. Neither the C max nor the AUC ∞ of midazolam were affected by coadministration of filgotinib (Fig. ). The point estimate pairwise comparisons for midazolam AUC ∞ and C max when coadministered with filgotinib versus midazolam alone were 1.1 (90% CI 1.0–1.2) and 1.0 (90% CI 0.9–1.1), respectively, meaning that no specific recommendation was warranted for coadministration of filgotinib with CYP3A4 substrates .
Since there may be mechanisms of enzyme induction/inhibition relevant to hormonal contraceptives that are presently poorly characterized (e.g., increased ethinyl estradiol due to sulfotransferase inhibition, which has been associated with thrombosis ), the effects of filgotinib (200 mg QD for 15 days) on hormonal contraceptives (single doses of 30 µg ethinyl estradiol/150 µg levonorgestrel before and after filgotinib) were evaluated in healthy female subjects . For both levonorgestrel and ethinyl estradiol, the percentage geometric least squares mean (GLSM) ratios and associated 90% CIs of C max and AUC ∞ were contained within the prespecified lack of interaction bounds (70–143%) when coadministered with filgotinib (Fig. ) . Based on these results, no dose adjustment was warranted for coadministration of filgotinib with hormonal contraceptives .
The oral antidiabetic metformin is a substrate of OCT2 and of multidrug and toxin extrusion (MATE) transporter 1 (MATE1) and MATE2-K. As in vitro assessments indicated that filgotinib and its primary metabolite are OCT inhibitors (albeit at supratherapeutic concentrations) , potential drug interactions between metformin (850 mg single dose) and filgotinib 200 mg QD were assessed. Metformin percentage GLSM ratios and associated 90% CIs for AUC ∞ and C max were not altered by coadministration of filgotinib (Fig. ), meaning that filgotinib dose adjustment with concomitant use of metformin and other OCT2 and MATE transporter substrates was deemed unnecessary .
In vitro data indicated that filgotinib and its primary metabolite may be inhibitors of OATP1B1 and OATP1B3 at supratherapeutic filgotinib concentrations . A phase I study was conducted in healthy subjects given filgotinib 200 mg QD for 11 days alongside single doses of the OATP probe substrates atorvastatin (40 mg) and a combination of pravastatin (40 mg)/rosuvastatin (10 mg) . Of note, rosuvastatin is both an OATP and a BCRP substrate, and the combination of rosuvastatin and pravastatin as probes permits simultaneous measurement of the activity of the drug transporters OATP and BCRP in the presence of filgotinib . Although coadministration of filgotinib did not affect atorvastatin AUC ∞ , atorvastatin C max was reduced by 18% (Fig. ). Neither the AUC ∞ nor the C max of the atorvastatin metabolite, 2-hydroxy-atorvastatin, were affected by coadministration with filgotinib (Fig. ). Pravastatin AUC ∞ was unaffected by filgotinib; however, filgotinib increased pravastatin C max by 25% and increased the C max and AUC ∞ of rosuvastatin by 68% and 42%, respectively (Fig. ). None of these changes were considered to represent a clinically meaningful effect of filgotinib coadministration on statin exposure, and no dose adjustment is required for statins or other OATP or BCRP substrates .
Methotrexate is an OAT1 and OAT3 substrate that is a standard of care for the treatment of RA. The potential for drug interaction with filgotinib was assessed during a phase IIa study that investigated filgotinib in the treatment of patients with active RA and an inadequate response to methotrexate . Patients received filgotinib 30, 75, 150, or 300 mg QD (or placebo) and continued their stable methotrexate dose (7.5–20 mg/kg weekly); pharmacokinetic data were pooled across filgotinib doses ( n = 17) because of the low patient numbers in individual dosing groups. The point estimate pairwise comparisons for methotrexate AUC τ and C max when coadministered with filgotinib versus methotrexate alone fell within the lack of interaction bounds (70–143%; Fig. ) . These data indicate that filgotinib does not significantly impact the pharmacokinetics of methotrexate and support the coadministration of filgotinib with methotrexate without dose adjustment.
Filgotinib has no prolongation effect on the corrected QT interval (calculated using Fridericia’s correction formula [QTcF] and an individual correction factor [QTcI]). A partially blinded, randomized, placebo- and positive-controlled (moxifloxacin 400 mg), four-period, multiple-dose, crossover study was conducted to evaluate the effect of filgotinib (at doses of 200 and 450 mg QD for 7 days) on placebo-corrected QTcF (ΔΔQTcF; primary endpoint) . Plasma exposures of filgotinib and its primary metabolite were expectedly higher following administration of filgotinib 450 versus 200 mg QD (mean C max increase of 2.1- and 1.9-fold for filgotinib and its primary metabolite, respectively, over a 2.3-fold dose range). No QT prolongation occurred based on by-timepoint analysis. No clinically relevant relationships were observed between time-matched, baseline-adjusted, placebo-corrected QTc interval and plasma concentrations of filgotinib or its primary metabolite . This represents a negative thorough QT study, as defined by International Conference on Harmonisation E14 guidance.
An analysis was performed based on early clinical data (studies in healthy subjects and a proof-of-concept study in patients with RA) to develop a preliminary population pharmacokinetic/pharmacodynamic model describing the time course of plasma concentrations of filgotinib and its primary active metabolite across the pharmacodynamic dose range of filgotinib (25–450 mg QD) . The pharmacokinetics of filgotinib and its primary metabolite were adequately described by a combined two-compartment and a one-compartment model, respectively, with complete conversion of filgotinib into metabolite at all except the highest filgotinib doses. Individual status (healthy subject vs. patient with RA) and sex were included as statistically significant covariates on filgotinib and primary metabolite plasma clearance and on filgotinib intercompartmental clearance, respectively. Since the phase I studies included in the model were conducted exclusively in healthy male subjects and the proof-of-concept study included 33/36 (92%) female patients with RA, sex was confounded with study and subject status in this pharmacokinetic model. The relative inhibition of pSTAT1 in healthy subjects was described by a combined direct-response model of the predicted plasma concentration of filgotinib and its primary metabolite; drug effect on pSTAT1 inhibition was implemented as a sigmoidal maximum effect ( E max ) model. No covariates were included in the model for biomarker. Figure shows the simulated steady-state inhibition of pSTAT1 for various filgotinib regimens . The biomarker–response curve over the dosing interval correlated with the filgotinib and primary metabolite time profiles, suggesting that the prolonged metabolite exposure resulted in the maintenance of inhibition over the dosing interval, whereas the peak filgotinib exposure contributed to the maximal inhibition . These data indicated that the maximum pharmacodynamic effect was achieved with a dose of filgotinib 200 mg QD and were used to support the dose and regimen selection for registration trials.
Filgotinib is a preferential JAK1 inhibitor that is metabolized to an active primary metabolite with similar JAK1 selectivity but with 10-fold lower potency and a relatively longer elimination half-life than the parent compound. Filgotinib 100 and 200 mg QD are approved for the treatment of RA in Europe and Japan. To date, filgotinib has been tested in various clinical studies in patients with RA , where doses of 100 to 200 mg QD or 50 to 100 mg BID were efficacious for signs and symptoms, with rapid-onset kinetics and a consistent safety profile. Of note, DARWIN 2 examined filgotinib monotherapy, whereas the other listed trials included filgotinib in combination with methotrexate. Filgotinib 200 mg QD is approved for the treatment of UC in Europe and Japan and is being evaluated (dosed at 100 and 200 mg QD) for the treatment of CD in phase III trials (NCT02914561, NCT02914600). Pharmacokinetic analysis following single and multiple ascending doses of filgotinib revealed rapid absorption following oral administration and conversion to its primary active metabolite. The pharmacokinetics of filgotinib and its metabolite were dose proportional over the dose range of 50 to 200 mg, with no notable food effect on the C max or AUC ∞ . As expected, based on its half-life of 4.9 to 10.7 h, filgotinib did not accumulate with QD or BID dosing. The metabolite has a longer half-life (19.6–27.3 h), reaching steady state within 4 days of chronic dosing and achieving a 2-fold accumulation in exposure after QD dosing. At steady state, exposures were approximately 16- to 20-fold higher for the primary metabolite than for the parent filgotinib . Filgotinib pharmacokinetics appear to be similar between healthy subjects and patients with RA or UC , and intrinsic factors such as age, mild renal impairment, and mild-to-moderate hepatic impairment have either no or minimal impact on the pharmacokinetics of filgotinib and its primary metabolite . Filgotinib has a low drug–drug interaction potential, without clinically significant interactions with commonly administered comedications, including methotrexate; oral contraceptives and other CYP3A4 substrates; statins; and acid-reducing agents, such as proton pump inhibitors and histamine antagonists . No dose adjustment is required when filgotinib is coadministered with other drugs . Both filgotinib and its primary metabolite are substrates of P-gp; however, coadministration with P-gp inhibitors or inducers does not affect filgotinib pharmacokinetics sufficiently to warrant dose adjustment . Neither filgotinib nor its primary metabolite have shown an effect on QTcF interval . In conclusion, the studies described herein support the use of filgotinib as a treatment for patients with RA and UC and potentially other inflammatory diseases, including CD, because of its pharmacokinetic/pharmacodynamic and efficacy profiles and acceptable tolerability .
|
Portal vein thrombosis complicating neonatal umbilical vein catheterization in a 3-month-old infant with coincidental extrahepatic biliary atresia: A case report | 5f5f4750-2275-4439-8218-aac6661e7dd7 | 11789905 | Surgical Procedures, Operative[mh] | The most common cause of vascular thrombosis in newborns is the use of central catheters. Neonatal portal vein thrombosis (PVT) is a rare event, occurring in approximately 1 in 100,000 live births, but is more frequently reported among graduates of neonatal intensive care units (NICU). PVT typically presents with portal hypertension (PH) and bleeding esophageal varices later in infancy or toddlerhood. Extrahepatic biliary atresia (EHBA) is a major cause of neonatal end-stage liver disease. Without timely diagnosis and treatment, most children with EHBA will develop irreversible liver fibrosis and PH within the first few months of life. To the best of our knowledge, coincidental PVT and EHBA have not been reported except after liver transplantation or Kasai portoenterostomy (KPE).
A 3-month-old female presented with jaundice, dark-colored urine, and clay-colored stools. She was born full term with a birth weight of 3.1 kg to a consanguineous, healthy couple. There was no maternal illness during pregnancy. The infant had a history of NICU admission for 10 days due to confirmed neonatal sepsis causing respiratory distress. An umbilical vein catheter (UVC) was inserted to provide the necessary treatment, which was frequently manipulated and repositioned. At 3 months of age, the infant weighed 4.4 kg and measured 53 cm in length. She was jaundiced and exhibited hepatosplenomegaly, with no other systemic abnormalities on examination. An abdominal ultrasound scan (USS) after a 4-hour fast revealed hepatosplenomegaly with mild ascites, and the gallbladder (GB) was not visualized. Color Doppler USS showed multiple collaterals at the site of the thrombosed portal vein, indicative of a portal cavernoma (Fig. ). A needle liver biopsy was performed, revealing distorted hepatic architecture and expanded portal areas with fibrous septa formation and proliferated bile ducts, some containing bile plugs. The hepatocytes exhibited mild ballooning, and there was evidence of intrahepatic and intracanalicular bile stasis (Fig. ). The patient underwent KPE. An intraoperative cholangiogram could not be performed through the atretic GB, confirming the diagnosis of EHBA. The procedure began with a rooftop incision followed by the mobilization and exteriorization of the liver through the release of the right and left triangular ligaments. The liver appeared cirrhotic, and an atretic GB was observed, along with ascitic fluid. At the porta hepatis, several peri-portal vein collaterals and a fibrotic main trunk (portal cavernoma) were noted. Lower peri-esophageal and gastric varices were also observed during the mobilization of the left lobe of the liver. The atretic GB was mobilized from its bed and excised along with the biliary remnants. The portal plate above the bifurcation of the portal vein, toward the corners of the hilar ductal plate, was meticulously dissected. The fibrous remnant in the ductal plate was excised using sharp dissection (Fig. ). A Roux-en-Y jejuno-jejunal anastomosis was performed 20 cm from the duodeno-jejunal junction, with an ascending limb of 40 cm. The anastomosis was completed side-to-end with 2-layer interrupted sutures. Pathological examination of the excised surgical specimen revealed a rudimentary GB with flattened mucosa and sparse chronic inflammatory infiltrate in the submucosa and muscle coat. The cystic duct and distal bile duct radicle showed marked lumen narrowing with sparse inflammatory infiltrate.
Postoperatively, the infant was transferred to the NICU. Her course was turbulent; however, she passed pigmented stools on day 2 and started oral fluids on day 3. Her total bilirubin levels decreased from 16 to 9.5 mg/dL. On day 8, she developed ascites with a drain output of 800 mL/d and started experiencing unexplained systemic hypertension, necessitating the addition of amlodipine, spironolactone, and alpha methyldopa. On day 10, the drain slipped and had to be reinserted. By day 12, the patient became hypoactive, required high nasal flow oxygen, began passing clay-colored stools, and had an average drain output of 400 to 500 mL/d. Her ammonia levels were elevated to 82 µmol/L. She died on day 15.
UVC in neonates can lead to several complications, including PVT, a significant cause of extrahepatic PH and upper gastrointestinal bleeding in children. EHBA is characterized by fibro-sclerosing obliteration of the extrahepatic bile ducts. Most infants with EHBA progress to chronic end-stage liver disease, PH, and its complications, such as massive bleeding from esophageal or gastric varices, ascites, hepato-renal syndrome, and hepatic encephalopathy. In this case, EHBA was associated with PVT and early PH with mild ascites. The French National Study reported a definite advantage of early KPE. Our 3-month-old patient developed early manifestations of PH, evidenced by esophageal varices detected during surgery, likely due to PVT induced by UVC. This challenges the notion that esophageal varices usually appear after 1 year of age. The coexistence of EHBA with other congenital anomalies has been reported, including malrotation, liver asymmetry, situs inversus, absent inferior vena cava, preduodenal portal vein, and polysplenia. However, the association of PVT and EHBA in our patient exacerbated their pathological impacts.
PVT can occur as a complication of neonatal UVC. It is often asymptomatic in the neonatal period and not clinically recognizable but can be easily detected by duplex Doppler USS. We strongly recommend routine checking of UVC position immediately after placement and during hospitalization to prevent future PVT in NICUs. Neonatologists and pediatricians should be aware of this potential complication to ensure proper catheter placement. Additionally, we recommend follow-up of NICU graduates with abdominal USS to diagnose early PVT development.
The authors thank their patient and his family.
Conceptualization: Mortada H.F. El-Shabrawi, Fetouh Hassanin. Data curation: Mortada H.F. El-Shabrawi, Fetouh Hassanin. Formal analysis: Mortada H.F. El-Shabrawi, Fetouh Hassanin. Investigation: Mortada H.F. El-Shabrawi, Ayman Hussein Abdel Sattar, Fetouh Hassanin, Maha Fathy Sheba, Ahmed Elhennawy, Seham Anwar Emam Marzouk. Methodology: Mortada H.F. El-Shabrawi, Ayman Hussein Abdel Sattar, Fetouh Hassanin, Maha Fathy Sheba, Ahmed Elhennawy, Seham Anwar Emam Marzouk. Supervision: Mortada H.F. El-Shabrawi. Validation: Mortada H.F. El-Shabrawi. Visualization: Mortada H.F. El-Shabrawi. Writing—original draft: Mortada H.F. El-Shabrawi, Fetouh Hassanin, Naglaa M. Kamal, Mohammed A.M. Oshi, Ali Algarni. Writing—review & editing: Mortada H.F. El-Shabrawi, Ayman Hussein Abdel Sattar, Fetouh Hassanin, Maha Fathy Sheba, Ahmed Elhennawy, Naglaa M. Kamal, Mohammed A.M. Oshi, Ali Algarni, Seham Anwar Emam Marzouk. Project administration: Ayman Hussein Abdel Sattar, Maha Fathy Sheba, Ahmed Elhennawy, Seham Anwar Emam Marzouk.
|
Should China be closed forever? A | 7770036b-18f7-48e6-89cd-674813b97c6b | 9162881 | Preventive Medicine[mh] | China has now 3 options to deal with the pandemic, each with some pros and cons . And option 2 seems the best strategy as Omicron and its variants are much harder to contain. What China should do in the coming months is to: • Ramp up vaccination rates using the highest efficacious mRNA vaccines, particularly targeting those 60+ and those with underlying high-risk conditions. • Apply different levels of restrictions according to the level of infections in a community. The focus should be on preventing severe outcomes while minimizing hospitalizations and excess deaths from other medical problems. Indeed, when key medical resources are protected, both population and individual health are protected. .) China could apply different levels of restrictions according to different degrees and stages of the outbreak in different communities. The level of restrictions should match the specific situation on the ground that aiming to maximize the pros and minimize the cons . While recognizing evolving scientific understanding and discoveries, we recommend 4 levels of restricted measures to inform prevention and control measures for specific local communities: ■ Level-1: Community transmission (< 10 per 100,000 population in one week) and no evidence of hospitalization of COVID-19 patients - Minimal requirement for mask wearing to get access to medical facilities without the need for quarantine. ■ Level-2: Community transmission (>10 per 100,000 population in one week) with new COVID-19 hospitalization - Mask wearing, and COVID-dedicated medical facilities (including a Fangcang shelter hospital for care). ■ Level-3: Community transmission (> 100 per 100,000 population in one week) – A comprehensive protection of critical or intensive care resources in all medical facilities. A complete closed loop between critical care resources and dedicated COVID facilities is established, and patients can only access critical care resources through dedicated COVID facilities. ■ Level 4: Community transmission (> 200 per 100,000 population in one week) - A complete closed loop between medical facilities and/or Fangcang shelter hospitals, like during the 2022 Beijing Winter Olympics. At the same time, China should double down on public education campaign about the status of COVID pandemic with timely dissemination of effective self-guided care for early detection and control measures. This could be achieved in 3 steps: 1. Ramp up education campaign to relieve public fear and anxiety with special emphasis on risk perception and the key differences between the current circulating Omicron variant and the original COVID variant that started the pandemic. In doing so, death due to specific causes from other medical problems should be provided for public awareness of excess deaths from chronic disease which could be reduced via risk factors modification and control measures. 2. Control “the epidemic of misinformation” from social media by establishing One Single Comprehensive and Authoritative Data Source for COVID-19 Outbreak for public use. Precise and authoritative definitions of COVID-19 cases and deaths should be communicated and used in data comparison. 3. Revise protocols for prophylactic treatment, stock antiviral drugs, teach basic self-care, and reduce public resistance of COVID-specific restriction. On the long run, the country should continue support innovative surveillance and screening research, as well as pharmaceutical R&D following an adaptive multistage policy which include: • Emergency Use Authorization (EUA) for rapid Point of Care Tests such as the NAAT-POCT • Increase production and stockpile the oral antiviral Paxlovid. In the randomized intervention trial supporting the EUA, Paxlovid was shown to reduce hospitalizations and deaths by 88% amongst those who tested positive for COVID-19 and are at high risk for severe outcomes ( https://www.fda.gov/news-events/press-announcements/coronavirus-covid-19-update-fda-authorizes-first-oral-antiviral-treatment-covid-19 ) . • Support new vaccine development. • Support COVID research through data sharing, including contact tracing data. Research areas should include how to reduce the socio-economic impacts of the COVID-19 pandemic and ensuing lockdowns. In dealing with this once a century pandemic, the difficulties faced by our Chinese colleagues is unprecedented. In adherence to the zero COVID policy, the enormous resources and effort devoted to implementing mass PCR testing and stringent quarantine measures to hundreds of millions of people for early detection and control has also no precedent in contemporary public health response to any epidemics. We salute the public and medical health professionals in China and the world should stand in their support. Throughout this COVID19 pandemic, health professionals and scientists throughout the world often experienced push back from politicians who decided on policy that also reflect their value-based constituents. We are cognizant of political pressure that may ensure on those who propose a policy change in China. Yet, difficult as it may seem, the middle option#2 represents the best path forward, as preponderance of the scientific evidence now indicates that covid19 will be around the world for a long time. China should not be closed forever. Dr. Liu reports consulting payments and honoraria or promises of the same for scientific presentations or reviews at numerous venues, including but not limited to Barilla, Johns Hopkins University, Fred Hutchinson Cancer Center, Harvard University, University of Buffalo, Guangdong General Hospital, Fuwai Hospital, and Chinese Academy of Medical Sciences, and the National Institutes of Health. He is also a member of the Data Safety and Monitoring Board for several trials, including the SELECT Trial -Semaglutide Effects on Cardiovascular Outcomes in People With Overweight or Obesity sponsored by Novo Nordisk and a trial of pulmonary hypertension in diabetes patients sponsored by at Massachusetts General Hospital. He receives royalties from UpToDate. Dr. Liu receives an honorarium from the American Society for Nutrition for his duties as Associate Editor. |
Evaluating surgical outcomes: robotic-assisted vs. conventional total knee arthroplasty | 8e38d216-62bc-4dc1-8076-ea0ceb8d62e9 | 11829363 | Surgical Procedures, Operative[mh] | Total knee arthroplasty (TKA) is a commonly used orthopedic procedure for treating severe knee osteoarthritis. Globally, knee osteoarthritis is prevalent, particularly among individuals aged 65 and older, with estimates suggesting a prevalence of approximately 22.9% . For patients who do not respond adequately to medication and physical therapy and experience significant impairment in their quality of life, TKA is an effective option for pain relief. With the aging population and changes in lifestyle, the demand for TKA has been increasing annually . However, traditional TKA is complex, requires high precision, and is associated with variable postoperative complications and recovery outcomes . In recent years, robotic-assisted surgery (RAS) has been increasingly adopted in TKA, aiming to enhance surgical precision and improve patient recovery . The integration of robotic technologies into orthopedic surgery has led to significant advancements in the field. Recent studies have highlighted the growing potential of robotic-assisted TKA (RA-TKA). For instance, RA-TKA has been shown to significantly improve surgical accuracy, which is crucial for achieving optimal postoperative outcomes . Additionally, it has been associated with faster recovery times and lower complication rates, further supporting its expanding role in clinical practice . Previous studies have suggested that robotic-assisted total knee arthroplasty (RA-TKA) may offer improvements in surgical precision, reducing complications and potentially enhancing postoperative outcomes. For instance, in a randomized controlled trial, Simcox et al. (2022) reported that RA-TKA significantly reduced postoperative infection rates and blood loss . Similarly, Gavin (2023) highlighted in a comparative study that RA-TKA patients recovered more quickly than those undergoing conventional surgery . However, despite the promising results, there are still unresolved issues. Most existing studies focus on short-term outcomes, with a lack of systematic evaluations on long-term recovery effects . Many studies emphasize basic metrics such as surgical duration and hospital stay, while comprehensive assessments of postoperative recovery, Complication Rate (CR), and long-term functional recovery are scarce . Furthermore, most research relies on traditional statistical methods, which may be insufficient for fully exploring valuable insights from the complex postoperative recovery process and individual variability. In particular, extracting critical features from large clinical datasets that are essential for assessing surgical outcomes and recovery remains a challenge . Therefore, more advanced analytical approaches, such as machine learning algorithms, are urgently needed to conduct a more comprehensive and in-depth analysis of the differences between RA-TKA and conventional TKA across various dimensions . This study aims to systematically evaluate the differences in surgical outcomes and postoperative recovery between RA-TKA and Conventional Total Knee Arthroplasty (C-TKA) through large-scale data analysis. The objective is to construct a predictive model based on machine learning to compare the effectiveness and recovery outcomes of RA-TKA and C-TKA. This systematic evaluation of RA-TKA’s surgical outcomes and recovery differences will provide scientific evidence and guidance for clinical practice. These findings may provide valuable insights that could inform efforts to optimize TKA procedures and postoperative management, with the potential to improve patient quality of life and promote the broader application of RAS in orthopedic operations. Collection of clinical data for robotic-assisted and C-TKA patients This study aims to systematically evaluate the differences in surgical outcomes and postoperative recovery between RA-TKA and C-TKA using machine learning algorithms. The study’s workflow is illustrated in Figure S1. We collected relevant data from the American College of Surgeons National Surgical Quality Improvement Program (NSQIP) clinical database ( https://www.facs.org/quality-programs/data-and-registries/ ), which included 17,266 patients—8,633 RA-TKA patients and 8,633 C-TKA patients. Each patient’s data consisted of 13 clinical features such as gender, age, Body Mass Index (BMI), operative time, and recovery information. To ensure the quality and integrity of the data, we conducted rigorous data cleaning and annotation. During the data cleaning process, we excluded samples that did not meet the standards, particularly those missing feature values due to technical issues. All data were standardized and normalized to a scale of 0 to 1. The NSQIP data used in this study is publicly accessible and includes detailed surgical and 30-day postoperative complication data from patients undergoing surgery at approximately 700 participating hospitals in the United States. Trained reviewers at each institution prospectively collected and clinically validated the data, with each entry marked by NSQIP identifiers. The visualization of the data used in this study is shown in Fig. . Feature extraction and selection in TKA patients Accurate feature extraction and selection are crucial for constructing an effective machine-learning model in TKA research. This process aims to identify the most informative variables from surgical records and postoperative recovery data to accurately predict and assess patient outcomes. To compare RA-TKA and C-TKA more precisely, data were collected from three perspectives: (1) Basic Information Features: This includes age, gender, and BMI. Age and BMI are critical factors influencing surgery and recovery. Older patients may have slower recovery rates, while those with higher BMI may face more postoperative complications. Gender differences can also affect surgical outcomes and recovery processes, making these basic features essential components of the dataset. (2) Surgical Information Features: These are extracted from surgical records and include operative time and surgical method. Operative time is a key variable reflecting the complexity of the surgery and potentially impacting the speed of postoperative recovery. The surgical method (robot-assisted or conventional) and the surgeon’s experience level can also influence surgical outcomes. These features provide insights into the details of the surgical process and help predict recovery outcomes. (3) Recovery Information Features: This mainly includes postoperative recovery time, whether the patient was discharged directly home, complication occurrence, and indicators of physical condition such as the modified frailty index (mFI) and functional status. To effectively identify outliers and key features from the extensive surgical and rehabilitation data, this study employs the Recursive Feature Elimination (RFE) algorithm to select the most informative features. The specific steps are as follows: (1) Train the dataset using a base model. (2) Gradually eliminate the least important features based on the model’s importance scores. (3) Iterate this process until the optimal feature set is obtained. This approach not only enhances the precision of feature selection but also reduces data dimensionality, thereby improving the training speed and predictive performance of the machine learning model. The process of feature extraction using the RFE algorithm is illustrated in Fig. . Dataset partitioning and model validation using cross-validation (CV) To ensure that the constructed model maintains good predictive performance on unknown data and effectively prevents overfitting, the dataset was divided into training, validation, and test sets. The sample sizes for these sets were in the ratio of 8:1:1 using random sampling. CV was then employed to validate the model’s performance. Specifically, a 10-fold CV was used, where the dataset is split into 10 subsets. In each iteration, one subset is used as the validation set, and the remaining nine subsets are used as the training set. This process is repeated 10 times, with each subset used as the validation set once. The final result is the average of all validation outcomes, providing a more stable and reliable model evaluation (Fig. ). Construction of the predictive model for surgical outcomes and postoperative recovery In this study, we aim to evaluate the differences in surgical outcomes and postoperative recovery between RA-TKA and C-TKA by constructing a Random Forest model. Random Forest is a robust machine learning algorithm known for its effectiveness in handling high-dimensional data and preventing overfitting, making it the preferred method for this research. The Random Forest model used in this study is an ensemble learning method based on decision trees. It enhances prediction performance and stability by averaging or voting on the predictions of multiple decision trees. The specific construction steps are as follows: (1) Construction of Multiple Decision Trees: During training, Random Forest employs bootstrapping to randomly extract multiple subsamples from the training set, each used to train a decision tree. (2) Random Feature Selection: At each node split in a decision tree, a random subset of features is selected from all features, from which the best-split feature is chosen. This not only reduces the impact of correlations but also improves the model’s generalization ability. (3) Ensemble Learning: By aggregating the predictions of all decision trees, the Random Forest model achieves more accurate and robust predictions. In our study, the Random Forest model determines the final classification outcome—whether the patient has recovered or not—through majority voting. The complete predictive model for evaluating surgical outcomes and postoperative recovery is illustrated in Fig. . Model performance evaluation metrics include accuracy, precision, recall, and F1 score. Statistical analysis In this study, we employed various software and statistical methods to construct predictive models for evaluating surgical outcomes and postoperative recovery. Programming and data processing were primarily conducted in Jupyter Notebook using Python, with data visualization performed using Visio software. Data manipulation and visualization were achieved with the pandas and Matplotlib libraries in Python. The statistical methods included t-tests, analysis of variance (ANOVA), and gradient backpropagation. Initially, we calculated the mean and variance of each feature and used box plots to display the distribution of features across different surgical methods. Model performance was assessed using Receiver Operating Characteristic (ROC) curves and confusion matrices. At each research stage, it was necessary to organize various datasets into arrays for subsequent feature extraction and model training. We utilized pandas to manage and analyze these arrays, normalizing each feature to eliminate differences in scale, thus enhancing model stability and efficiency. Matplotlib was employed to visually represent sample features and model performance. For the comparative analysis of robotic-assisted and C-TKA patients, we calculated the variance of each feature and visualized them using box plots to facilitate multi-angle comparisons. Statistical significance was determined using gradient descent for backpropagation to ascertain the weights of each gene, thus identifying the expression of specific genes. This study aims to systematically evaluate the differences in surgical outcomes and postoperative recovery between RA-TKA and C-TKA using machine learning algorithms. The study’s workflow is illustrated in Figure S1. We collected relevant data from the American College of Surgeons National Surgical Quality Improvement Program (NSQIP) clinical database ( https://www.facs.org/quality-programs/data-and-registries/ ), which included 17,266 patients—8,633 RA-TKA patients and 8,633 C-TKA patients. Each patient’s data consisted of 13 clinical features such as gender, age, Body Mass Index (BMI), operative time, and recovery information. To ensure the quality and integrity of the data, we conducted rigorous data cleaning and annotation. During the data cleaning process, we excluded samples that did not meet the standards, particularly those missing feature values due to technical issues. All data were standardized and normalized to a scale of 0 to 1. The NSQIP data used in this study is publicly accessible and includes detailed surgical and 30-day postoperative complication data from patients undergoing surgery at approximately 700 participating hospitals in the United States. Trained reviewers at each institution prospectively collected and clinically validated the data, with each entry marked by NSQIP identifiers. The visualization of the data used in this study is shown in Fig. . Accurate feature extraction and selection are crucial for constructing an effective machine-learning model in TKA research. This process aims to identify the most informative variables from surgical records and postoperative recovery data to accurately predict and assess patient outcomes. To compare RA-TKA and C-TKA more precisely, data were collected from three perspectives: (1) Basic Information Features: This includes age, gender, and BMI. Age and BMI are critical factors influencing surgery and recovery. Older patients may have slower recovery rates, while those with higher BMI may face more postoperative complications. Gender differences can also affect surgical outcomes and recovery processes, making these basic features essential components of the dataset. (2) Surgical Information Features: These are extracted from surgical records and include operative time and surgical method. Operative time is a key variable reflecting the complexity of the surgery and potentially impacting the speed of postoperative recovery. The surgical method (robot-assisted or conventional) and the surgeon’s experience level can also influence surgical outcomes. These features provide insights into the details of the surgical process and help predict recovery outcomes. (3) Recovery Information Features: This mainly includes postoperative recovery time, whether the patient was discharged directly home, complication occurrence, and indicators of physical condition such as the modified frailty index (mFI) and functional status. To effectively identify outliers and key features from the extensive surgical and rehabilitation data, this study employs the Recursive Feature Elimination (RFE) algorithm to select the most informative features. The specific steps are as follows: (1) Train the dataset using a base model. (2) Gradually eliminate the least important features based on the model’s importance scores. (3) Iterate this process until the optimal feature set is obtained. This approach not only enhances the precision of feature selection but also reduces data dimensionality, thereby improving the training speed and predictive performance of the machine learning model. The process of feature extraction using the RFE algorithm is illustrated in Fig. . To ensure that the constructed model maintains good predictive performance on unknown data and effectively prevents overfitting, the dataset was divided into training, validation, and test sets. The sample sizes for these sets were in the ratio of 8:1:1 using random sampling. CV was then employed to validate the model’s performance. Specifically, a 10-fold CV was used, where the dataset is split into 10 subsets. In each iteration, one subset is used as the validation set, and the remaining nine subsets are used as the training set. This process is repeated 10 times, with each subset used as the validation set once. The final result is the average of all validation outcomes, providing a more stable and reliable model evaluation (Fig. ). In this study, we aim to evaluate the differences in surgical outcomes and postoperative recovery between RA-TKA and C-TKA by constructing a Random Forest model. Random Forest is a robust machine learning algorithm known for its effectiveness in handling high-dimensional data and preventing overfitting, making it the preferred method for this research. The Random Forest model used in this study is an ensemble learning method based on decision trees. It enhances prediction performance and stability by averaging or voting on the predictions of multiple decision trees. The specific construction steps are as follows: (1) Construction of Multiple Decision Trees: During training, Random Forest employs bootstrapping to randomly extract multiple subsamples from the training set, each used to train a decision tree. (2) Random Feature Selection: At each node split in a decision tree, a random subset of features is selected from all features, from which the best-split feature is chosen. This not only reduces the impact of correlations but also improves the model’s generalization ability. (3) Ensemble Learning: By aggregating the predictions of all decision trees, the Random Forest model achieves more accurate and robust predictions. In our study, the Random Forest model determines the final classification outcome—whether the patient has recovered or not—through majority voting. The complete predictive model for evaluating surgical outcomes and postoperative recovery is illustrated in Fig. . Model performance evaluation metrics include accuracy, precision, recall, and F1 score. In this study, we employed various software and statistical methods to construct predictive models for evaluating surgical outcomes and postoperative recovery. Programming and data processing were primarily conducted in Jupyter Notebook using Python, with data visualization performed using Visio software. Data manipulation and visualization were achieved with the pandas and Matplotlib libraries in Python. The statistical methods included t-tests, analysis of variance (ANOVA), and gradient backpropagation. Initially, we calculated the mean and variance of each feature and used box plots to display the distribution of features across different surgical methods. Model performance was assessed using Receiver Operating Characteristic (ROC) curves and confusion matrices. At each research stage, it was necessary to organize various datasets into arrays for subsequent feature extraction and model training. We utilized pandas to manage and analyze these arrays, normalizing each feature to eliminate differences in scale, thus enhancing model stability and efficiency. Matplotlib was employed to visually represent sample features and model performance. For the comparative analysis of robotic-assisted and C-TKA patients, we calculated the variance of each feature and visualized them using box plots to facilitate multi-angle comparisons. Statistical significance was determined using gradient descent for backpropagation to ascertain the weights of each gene, thus identifying the expression of specific genes. Analysis and optimization of patient characteristics for robotic-assisted and C-TKA based on the NSQIP database Clinical data from the NSQIP database were analyzed, including 13 features related to patient demographics, surgical details, and recovery outcomes. Using the Recursive Feature Elimination (RFE) algorithm, eight key features were identified: Age, Body Mass Index (BMI), Modified Frailty Index (mFI), Operative Time, Length of Stay (LOS), Functional Status, Complication Rate (CR), and Surgical Type (Case). The distribution of these features is detailed in Fig. . Hyperparameter optimization and evaluation analysis of the model for predicting post-TKA recovery outcomes To develop the optimal analysis model, this study iteratively optimized the two most critical hyperparameters of the Random Forest model: the number of decision trees (N-Estimators) and the maximum depth of each tree (Max-Depth). Through continuous iteration, the optimal values were determined to be 40 for N-Estimators and 7 for Max-Depth, as illustrated in Fig. . A predictive model for evaluating surgical outcomes and postoperative recovery was constructed based on the Random Forest algorithm. Through hyperparameter optimization, the optimal parameters were determined (N-Estimators = 40, Max-Depth = 7), as illustrated in Fig. . The model’s ROC curve (Fig. A) indicates an AUC value of 0.74, suggesting moderate predictive capability for patient recovery. The confusion matrix (Fig. B) shows that the model achieves high accuracy in predicting recovered patients but requires improvement in predicting non-recovered cases. Evaluation and comparison of key features in RA-TKA and C-TKA surgery outcomes The results of the model’s weight analysis (Fig. ) revealed that the three most critical features were Functional Status, Complication Rate (CR), and Modified Frailty Index (mFI). Variance analysis and box plots (Fig. ) demonstrated that RA-TKA showed better performance than C-TKA across these key features. Patients undergoing RA-TKA showed better postoperative functional status compared to those undergoing C-TKA, suggesting improved joint function and mobility. Additionally, the incidence of postoperative complications was lower among RA-TKA patients, which may reflect the enhanced precision and reduced invasiveness of the robotic-assisted procedure. Furthermore, RA-TKA patients showed better mFI values, indicating faster and more effective recovery outcomes compared to C-TKA patients. Specific recommendations for optimizing TKA procedures and postoperative recovery management TKA is a common orthopedic surgery with outcomes that significantly impact patients’ quality of life. The analysis identified three critical indicators of postoperative success: functional status, CR, and the mFI. Based on these findings, this study offers specific recommendations in three areas to optimize surgical procedures and postoperative recovery management, thereby enhancing both surgical outcomes and recovery quality. Standardization of Surgical Procedures. (1) Promotion of RAS: RA-TKA, due to its precision and consistency, demonstrates superior performance in reducing complications and promoting postoperative recovery. RAS may benefit from further promotion and adoption, helping more surgeons become proficient with the technology and potentially improving surgical outcomes. (2) Standardized Surgical Procedures: Detailed operational standards for both RA-TKA and C-TKA should be established, ensuring that every step strictly adheres to the guidelines. This includes preoperative preparation, prosthesis selection, surgical incisions, bone cutting, and prosthesis implantation. Such standardized procedures help reduce intraoperative errors and postoperative complications. (3) Intraoperative Monitoring Systems: Advanced intraoperative monitoring and navigation systems should be introduced to monitor key parameters in real time during surgery. This allows for the timely detection and correction of potential operational deviations, thereby enhancing the safety and accuracy of surgeries. Personalized Postoperative Rehabilitation Plans. (1) Development of Personalized Rehabilitation Plans: Tailor rehabilitation plans based on the patient’s specific conditions, including age, functional status, and mFI. For patients with a higher mFI, provide more aggressive postoperative rehabilitation measures to promote rapid recovery. (2) Stage-specific Rehabilitation Assessments: Conduct regular assessments of patients during the rehabilitation process to monitor their progress and functional status. Use objective data analysis to promptly adjust rehabilitation measures, ensuring the effectiveness and relevance of the rehabilitation plan. (3) Optimization of Postoperative Care: Develop detailed postoperative care plans to address the high risk of complications, including wound care, pain management, and infection prevention. For patients with a high mFI, enhance postoperative care to reduce the incidence of complications and facilitate recovery. Other optimization measures include the following: (1) Preoperative optimization strategies: Prior to surgery, implement measures targeting the patient’s mFI and other health indicators, such as nutritional support and improving cardiopulmonary function, to enhance surgical tolerance and postoperative recovery. (2) Data analysis and feedback: Establish a postoperative recovery data management system to collect and analyze patient recovery data. Identify issues promptly and provide feedback to both the surgical and recovery teams. Use data-driven decision-making to continuously refine surgical and recovery processes, thereby improving overall treatment outcomes. Clinical data from the NSQIP database were analyzed, including 13 features related to patient demographics, surgical details, and recovery outcomes. Using the Recursive Feature Elimination (RFE) algorithm, eight key features were identified: Age, Body Mass Index (BMI), Modified Frailty Index (mFI), Operative Time, Length of Stay (LOS), Functional Status, Complication Rate (CR), and Surgical Type (Case). The distribution of these features is detailed in Fig. . To develop the optimal analysis model, this study iteratively optimized the two most critical hyperparameters of the Random Forest model: the number of decision trees (N-Estimators) and the maximum depth of each tree (Max-Depth). Through continuous iteration, the optimal values were determined to be 40 for N-Estimators and 7 for Max-Depth, as illustrated in Fig. . A predictive model for evaluating surgical outcomes and postoperative recovery was constructed based on the Random Forest algorithm. Through hyperparameter optimization, the optimal parameters were determined (N-Estimators = 40, Max-Depth = 7), as illustrated in Fig. . The model’s ROC curve (Fig. A) indicates an AUC value of 0.74, suggesting moderate predictive capability for patient recovery. The confusion matrix (Fig. B) shows that the model achieves high accuracy in predicting recovered patients but requires improvement in predicting non-recovered cases. The results of the model’s weight analysis (Fig. ) revealed that the three most critical features were Functional Status, Complication Rate (CR), and Modified Frailty Index (mFI). Variance analysis and box plots (Fig. ) demonstrated that RA-TKA showed better performance than C-TKA across these key features. Patients undergoing RA-TKA showed better postoperative functional status compared to those undergoing C-TKA, suggesting improved joint function and mobility. Additionally, the incidence of postoperative complications was lower among RA-TKA patients, which may reflect the enhanced precision and reduced invasiveness of the robotic-assisted procedure. Furthermore, RA-TKA patients showed better mFI values, indicating faster and more effective recovery outcomes compared to C-TKA patients. TKA is a common orthopedic surgery with outcomes that significantly impact patients’ quality of life. The analysis identified three critical indicators of postoperative success: functional status, CR, and the mFI. Based on these findings, this study offers specific recommendations in three areas to optimize surgical procedures and postoperative recovery management, thereby enhancing both surgical outcomes and recovery quality. Standardization of Surgical Procedures. (1) Promotion of RAS: RA-TKA, due to its precision and consistency, demonstrates superior performance in reducing complications and promoting postoperative recovery. RAS may benefit from further promotion and adoption, helping more surgeons become proficient with the technology and potentially improving surgical outcomes. (2) Standardized Surgical Procedures: Detailed operational standards for both RA-TKA and C-TKA should be established, ensuring that every step strictly adheres to the guidelines. This includes preoperative preparation, prosthesis selection, surgical incisions, bone cutting, and prosthesis implantation. Such standardized procedures help reduce intraoperative errors and postoperative complications. (3) Intraoperative Monitoring Systems: Advanced intraoperative monitoring and navigation systems should be introduced to monitor key parameters in real time during surgery. This allows for the timely detection and correction of potential operational deviations, thereby enhancing the safety and accuracy of surgeries. Personalized Postoperative Rehabilitation Plans. (1) Development of Personalized Rehabilitation Plans: Tailor rehabilitation plans based on the patient’s specific conditions, including age, functional status, and mFI. For patients with a higher mFI, provide more aggressive postoperative rehabilitation measures to promote rapid recovery. (2) Stage-specific Rehabilitation Assessments: Conduct regular assessments of patients during the rehabilitation process to monitor their progress and functional status. Use objective data analysis to promptly adjust rehabilitation measures, ensuring the effectiveness and relevance of the rehabilitation plan. (3) Optimization of Postoperative Care: Develop detailed postoperative care plans to address the high risk of complications, including wound care, pain management, and infection prevention. For patients with a high mFI, enhance postoperative care to reduce the incidence of complications and facilitate recovery. Other optimization measures include the following: (1) Preoperative optimization strategies: Prior to surgery, implement measures targeting the patient’s mFI and other health indicators, such as nutritional support and improving cardiopulmonary function, to enhance surgical tolerance and postoperative recovery. (2) Data analysis and feedback: Establish a postoperative recovery data management system to collect and analyze patient recovery data. Identify issues promptly and provide feedback to both the surgical and recovery teams. Use data-driven decision-making to continuously refine surgical and recovery processes, thereby improving overall treatment outcomes. This study systematically evaluates the differences in surgical outcomes and postoperative recovery between RA-TKA and C-TKA using machine learning algorithms. RA-TKA has gained increasing popularity, particularly due to its significant advantages in improving surgical precision and reducing complications . Previous studies have shown that RA-TKA significantly reduces prosthetic implantation errors and enhances postoperative joint function. Although C-TKA remains widely used, its accuracy and recovery outcomes are relatively inferior . By employing machine learning algorithms, this study not only provides a comprehensive evaluation of the two surgical techniques but also quantifies subtle differences through multidimensional feature analysis. The innovation of this research lies in the use of more sophisticated data analysis methods, yielding scientifically robust conclusions that build upon previous literature. RA-TKA typically adopts the principles of functional alignment or kinematic alignment, allowing for personalized adjustments to prosthesis implantation that better accommodate the patient’s anatomical and biomechanical characteristics. These approaches have been demonstrated in multiple studies to optimize postoperative knee joint dynamics, thereby improving functional outcomes and reducing complication rates. In contrast, C-TKA predominantly relies on mechanical alignment, which standardizes prosthesis implantation based on unified axes such as the tibiofemoral mechanical axis. However, this method often overlooks individual anatomical variations, potentially leading to lower postoperative satisfaction for certain patients. The superior outcomes associated with RA-TKA may partially stem from advancements in alignment strategies rather than the robotic technology alone. The high precision of robotic systems enables intraoperative adjustments to prosthesis positioning based on the patient’s specific anatomy, achieving more consistent and personalized implantation results. In contrast, C-TKA relies heavily on the surgeon’s experience, offering relatively less personalization, which may contribute to the observed differences in surgical outcomes between the two techniques. This finding highlights the need for future studies to explicitly document alignment strategies and investigate the roles of different alignment principles in RA-TKA and C-TKA. Furthermore, exploring ways to optimize alignment strategies based on individual patient characteristics will be crucial for improving outcomes in total knee arthroplasty. In terms of surgical outcomes, our study demonstrates that RA-TKA is significantly superior to C-TKA, particularly in terms of surgery duration, CR, and postoperative functional recovery. Previous research has supported this conclusion, suggesting that RA-TKA, through precise intraoperative navigation and robotic assistance, reduces prosthesis misalignment and consequently decreases surgical complications . However, unlike some studies, we found that RA-TKA did not result in significantly longer surgery times, a finding that contrasts with reports suggesting prolonged operative times for RA-TKA. Feature selection and analysis using machine learning indicated that variations in surgery duration are more likely attributed to the surgical team’s proficiency rather than limitations inherent to the surgical method itself. Postoperative recovery is a crucial indicator for evaluating surgical outcomes. Our study demonstrates that patients undergoing RA-TKA exhibit faster recovery and more significant postoperative functional improvement. This finding aligns with previous research, which has highlighted that RA-TKA enhances joint function, reduces pain, and improves patients’ quality of life through precise bone cutting and prosthesis positioning . Additionally, our study assessed the recovery of frail patients using the mFI and found that RA-TKA was particularly effective in this group. Frail patients typically recover more slowly after surgery, but the high precision and minimally invasive nature of RA-TKA significantly accelerated their recovery . In contrast, traditional C-TKA performed relatively poorly in these patients, further highlighting the greater potential of RA-TKA in postoperative recovery management. The application of machine learning algorithms is another highlight of this study. Unlike traditional statistical methods, machine learning algorithms can handle multidimensional and complex features while capturing nonlinear relationships. In this study, a random forest algorithm was employed to build a model using a large training dataset, allowing for a systematic prediction and analysis of the differences in surgical outcomes and recovery between RA-TKA and C-TKA. The strength of the random forest algorithm lies in its robust feature selection capability, accurately identifying key factors that influence surgical outcomes and recovery, thus providing more scientifically sound guidance for clinical practice. Compared to traditional approaches, this study offers a more comprehensive analysis, and the use of machine learning algorithms enhances the accuracy of the predictive model while minimizing potential biases. Based on the findings of this study, several recommendations for optimizing clinical practice are proposed: (1) RA-TKA demonstrates superior performance in reducing complications and promoting postoperative recovery, and should be widely adopted in clinical practice. (2) Detailed procedural standards for both RA-TKA and C-TKA should be established to ensure that every step strictly follows standardized protocols, minimizing intraoperative errors and postoperative complications. (3) Personalized rehabilitation plans should be developed based on the specific conditions of each patient, with regular assessments to ensure the effectiveness and relevance of the rehabilitation process. Although this study provides a scientific evaluation of the surgical outcomes of RA-TKA and C-TKA using big data and machine learning algorithms, several limitations remain. First, the data were primarily sourced from the NSQIP clinical database, which may not fully represent clinical practices across all regions, especially outside the U.S., potentially affecting the generalizability of the results. Second, certain influencing factors, such as patients’ psychological conditions, social support, and postoperative rehabilitation environments, were not included in the analysis, which could limit the comprehensiveness of the findings. Additionally, the retrospective nature of the data may introduce selection bias and hinder the ability to establish causality, as the model’s predictions rely heavily on the quality and completeness of the data. While the findings align with those of Simcox and Gavin , future research should incorporate a broader range of patient data and prioritize multicenter randomized controlled trials to reduce bias, improve the generalizability of the results, and further validate the effectiveness of RA-TKA. In conclusion, this study systematically evaluated the differences in surgical outcomes and postoperative recovery between RA-TKA and C-TKA using machine learning algorithms. The findings demonstrate that RA-TKA outperforms C-TKA across several key metrics, including functional status, complication rate, and modified frailty index. Notably, RA-TKA showed greater potential benefits in postoperative recovery for frail patients. These results provide valuable guidance for clinical practice, supporting the broader application of RA-TKA in personalized surgical planning and rehabilitation management. Furthermore, this study highlights the combined advantages of robotic technology and advanced alignment philosophies, such as functional or kinematic alignment, and offers specific recommendations for optimizing TKA procedures and postoperative care. Despite certain limitations, such as the lack of explicit alignment documentation in the database and the exclusion of other potential influencing factors, the results offer important insights for future research. Future studies should explore how alignment strategies can be optimized based on individual patient characteristics and validate these strategies in diverse patient populations to further enhance the clinical benefits of RA-TKA. |
Distinct effects of phyllosphere and rhizosphere microbes on invader | 6ad7c81b-edb9-4050-a655-10396ecb77ca | 11186635 | Microbiology[mh] | Plant-modified soil properties affect the performance of plants, which are termed ‘plant-soil feedbacks (PSFs)’. PSFs can affect species coexistence and local plant community composition and dynamics . For plant invasion, PSFs are usually positive because of escaping soil pathogens and recruiting some beneficial microbes or negative effects because of accumulating local pathogens . Similar to soil, leaf litter can also affect plant growth, species diversity, and community structure , thus playing important roles in population establishment and community dynamics . However, related research has focused mainly on physical (e.g. maintaining soil moisture and temperature, increasing nutrition and reducing light) or chemical effects (e.g. releasing allelochemicals) but has rarely focused on leaf microbial effects. Until 2017, , extended the PSF to aboveground tissues (including leaf, stem, and floral tissues), termed ‘plant-phyllosphere feedbacks (PPFs)’, and found that all four Asteraceae species experienced stronger negative PPFs than PSFs. Subsequently, this team further verified that all 10 tested Asteraceae plants experienced negative PPFs . The lack of strong mutualists and relatively high abundance of pathogens in the phyllosphere may account for the negative PPFs. In addition to microbial sources (i.e. soil vs leaf litter), ontogeny (seedling growth stage) and soil nutrient levels can affect plant-microbe interactions. For example, seedlings showed distinct sensitivity during the growth stage, and younger seedlings were more susceptible to infection by soil microbes because of fewer defense resources . Interestingly, leaf litter has an adverse effect on seedling emergence but a positive effect on later plant growth ; litter also has a stronger negative effect on earlier vegetation growth than on the elder . Moreover, plants enrich distinct microbes under different nutrient conditions and affect plant performance . For example, the bacterial diversity in duckweed plants was reduced under nutrient-deficient conditions, but the abundance of Firmicutes increased , and members of Firmicutes have been reported to promote plant stress tolerance . , reported that nutrient additions cause crops to enrich some bacteria and fungi from soil and increase yield. Ageratina adenophora (Sprengel) RM King and H Robinson (Asteraceae), known as Crofton weed or Mexican devil weed, has invaded more than 30 countries and areas, including South Africa, Australia, New Zealand, Hawaii, India, and China . It is a perennial weed and can produce high yields of seeds with a high germination rate (GR) . This weed commonly grows in monocultures, but a high density of seedlings is not common in the wild. Previous studies have shown that A. adenophora can enrich the beneficial soil microbial community to facilitate invasion ; in contrast, A. adenophora leaves harbor diverse fungal pathogens that can cause adverse effects on itself seed germination and growth . Thus, it is interesting to determine whether leaf microbes play a distinct role from soil microbes in regulating A. adenophora seedling density and whether these effects change with the A. adenophora growth stage and soil nutrition level. In this study, we inoculated A. adenophora with soil or leaf litter at three stages, 0 day, 21 days, and 28 days after sowing, and transplanted seedlings to grow in soils with high or low nutritional levels. We first determined the germination, seedling survival, and growth of the A. adenophora plants. Then, we characterized the bacterial and fungal communities of the soils and leaf litter as inoculation sources, as well as the microbial communities enriched in the leaves and roots of the A. adenophora seedlings after growing; we also isolated the fungi associated with the dead seedlings and tested their seedling-killing effects on A. adenophora . Finally, we correlated the microbial community with A. adenophora seedling mortality and growth. We hypothesized that the microbial communities associated with leaf litter and rhizosphere soils can account for the differential effects on A. adenophora seedling mortality and growth during different growth stages when growing under different nutrient conditions. We expected that (1) leaf litter would have more adverse effects on seed germination, seedling survival, and growth than soil, as leaf litter often harbors more plant pathogens, and (2) inoculation at different growth stages would change the microbial community enriched by seedlings and thus affect seedling growth. Earlier inoculation had a greater adverse impact on seedling growth than later inoculation, as younger seedlings were more sensitive to pathogen infection than older seedlings. (3) The nutrient level influences seedlings to recruit microbes and thus affects seedling growth.
Effects of leaf litter and rhizosphere soil on the mortality and growth of A. adenophora seedlings At the G0 timepoint, sterile leaf inoculation significantly delayed germination time more than did soil and non-sterile leaf inoculation, as well as the control (nothing inoculated) ( , p<0.05). Leaf and soil inoculation had no distinct effects on the GR ( , p>0.05). In addition, the inoculation of sterile and non-sterile leaves at G0 caused a high mortality rate (MR) (19.7% vs 96.7%) for seedlings growing in Petri dishes. Only non-sterile leaves caused a low percentage of seedling death (8.4%) when the seedlings were inoculated at G21 . Two weeks after transplanting these seedlings into the cups, leaf inoculation caused significantly greater seedling mortality than did soil inoculation (p<0.001); the non-sterile sample caused greater seedling mortality than did the sterile sample, especially leaf inoculation during the G0 and G21 periods. Moreover, non-sterile leaf inoculation at earlier stages significantly increased seedling mortality compared with that at later stages ( , p<0.05). However, seedling mortality did not differ between the high and low nutrient conditions, regardless of leaf or soil inoculation ( , both p>0.05). With the exception of nutrient level, inoculation source and time, as well as their interaction with nutrient level, significantly affected the microbial role in the total dry biomass of A. adenophora (p<0.05, ). Soil and leaf microbial effects on seedling biomass interacted with soil nutrient level and inoculation time: when inoculated at the G0 timepoint, both soil and leaf microbes had adverse effects on A. adenophora growth, as all seedlings inoculated with non-sterile leaves died, and those inoculated with non-sterile soil grew poorer than those inoculated with sterile soil under both low- and high nutrition conditions; when inoculated at the G21 timepoint, both soil and leaf microbes had significantly positive effects only under high nutrition conditions; when inoculated at the G28 timepoint, only soil microbes promoted seedling growth under high nutrition conditions; and when inoculated at the G21+28 timepoint, only leaf microbes had a significantly positive effect on seedling growth under low nutrition conditions . Correlations of the microbial community and potential functions of inocula with A. adenophora seedling mortality at the early stage The soil and leaf inocula had distinct microbial diversity, community compositions, and potential functions. The microbial diversity and richness were greater in the soil than in the leaf litter . Top three core bacteria were Rhodoplanes (5.65%), Bradyrizhobium (4.80%), and some unclassified Alphaproteobacteria (4.56%) for soil and Pseudomonas (30.33%), Massilia (17.45%), and Sphingomonas (16.35%) for leaf . Top three core fungal genera were Mortierella (7.00%), Inocybe (5.36%), and Neonectria (3.63%) for soil and Didymella (27.30%), Alternaria (8.95%), and Cryptococcus (4.44%) for leaf . Plant bacterial pathogens (2.29%) were the potential function of the core bacterial taxa in leaves but not in soil; soil had a greater abundance of nitrogen circle-related function (50.98%) than did leaves (17.55%) . The abundance of plant fungal pathogens was greater in leaves (68.20%) than in soil (33.93%) . We further correlated the top 30 microbial genera of leaf and soil inocula with seed germination and seedling mortality in response to inoculation with non-sterile inocula at G0. The abundances of both the soil and leaf microbial genera were related to the seedling mortality rate (MR) but not to the germination time (GT) or GR . Specifically, the leaf core bacterial genera Pseudomonas , Sphingomonas , Massilia , Variovorax , Aureimonas , and Agrobacterium were positively correlated with MR, but the soil core bacteria, including Alphaproteobacteria_unclassified, Rhodoplanes , Vicinamibacteraceae_unclassified, and Pedomicrobium , were negatively correlated with MR . The leaf core fungal genera Didymella , Pleosporales_unclassified, Subplenodomus , and Bulleromyces were positively correlated with MR, but the soil core fungal genera Mortierella , Hypocrea , Pochonia, and Volutella were negatively correlated with MR . We obtained 192 cultivable fungal isolates from 40 dead seedlings, with an average of 4.825 isolates per dead seedling . Based on the ITS genes of the representative strains , they were divided into seven families. The dominant family was Didymellaceae (relative abundance = 66.15%), and the numerically dominant genera were Allophoma (50.52%), Alternaria (26.04%), and Epicoccum (5.73%) . The seedling-killing effects of these strains on A. adenophora exhibited a significant phylogenetic signal (Pagel’s λ=0.82, p=0.0002). Overall, numerically dominant Allophoma (Didymellaceae) and Alternaria (Pleosporaceae) had high seedling mortality (54–100%) . Enrichment of microbial community and function by A. adenophora seedlings under different treatments Nonmetric multidimensional scaling (NMDS) and permutational analysis of variance (PERMANOVA) revealed that all four factors significantly affected the bacterial community and functional assembly of the seedlings, and the greatest effects were inoculation time and compartment (all p<0.05, R 2 : 0.102–0.138), followed by inoculation source and nutrition (all p<0.05, R 2 : 0.024–0.082, ). Additionally, compartment, inoculation source, and time significantly affected the fungal community and functional assembly (all p<0.05, R 2 : 0.054–0.102), but nutrition affected only the fungal community (p=0.001, R 2 =0.031) . Further analysis for each inoculation time treatment showed that compartment and inoculation source mainly affected the microbial community and functional assembly and explained a greater proportion of the variation in bacteria than in fungi and in function than in the community. The nutrient level mainly affected the bacterial community at certain inoculation time (p<0.05 for G0, G21, and G28, , , ). Correlations of the enriched microbial community and function with A. adenophora seedling growth We further analyzed the correlation of microbial abundance and putative functions enriched by seedlings with the microbial effect on seedling growth (response index [RI]). We identified 47 root endophytic genera that were significantly correlated with A. adenophora growth. Among these genera, seven negative genera were less abundant in seedlings treated by leaf inoculation than in those treated by soil inoculation but were similar in abundance in seedlings treated by different inoculation time, and in seedlings grown at different nutrient levels. In contrast, 40 positive genera, e.g., the fungi Duganella and Mortierella and the bacteria Massilia, Pseudomonas, and Sphingomonas , were more abundant in seedlings treated by leaf inoculation than by soil inoculation but less abundant in seedlings inoculated at G0 than at the other three inoculation time treatments . Eighteen leaf endophytic genera with significant correlations with the RI were identified, of which three negative genera, namely, the bacteria Tardiphaga and Brevundimonas and the fungus Microsphaera , were more abundant in seedlings inoculated at G0 than at the other three inoculation time treatments and were slightly more abundant in the seedlings inoculated with soil than in those inoculated with leaf; in contrast, 15 positive genera, e.g., the fungi Hypocrea and Pleosporales_unclassified, were more abundant in the seedlings inoculated with leaf than in those inoculated with soil and were more abundant in the seedlings inoculated at G21 than in the seedlings inoculated at the other three inoculation time treatments . We identified several bacterial functions in the roots and fungal function guilds enriched in the roots and leaves of A. adenophora seedlings that were significantly correlated with the RI . Two bacterial functions involved in the N cycle (nitrate ammonification and nitrite ammonification) in roots showed a significant positive correlation with RI . The fungal guild Ectomycorrhizal showed a significantly positive correlation. Unexpectedly, the putative plant pathogen guild showed a significant positive correlation with seedling growth, while the arbuscular mycorrhizal guild showed a negative correlation . The positive bacterial functions in roots and fungal guilds in leaves had greater relative abundances in the seedlings after leaf inoculation than those after soil inoculation ; additionally, the positive fungal guilds in roots and leaves had significantly greater abundances in the G21 seedlings than in those at the other three inoculation time treatments. Surprisingly, there was no difference in the relative abundance of microbes or functions involved in seedling growth under different nutrient levels .
A. adenophora seedlings At the G0 timepoint, sterile leaf inoculation significantly delayed germination time more than did soil and non-sterile leaf inoculation, as well as the control (nothing inoculated) ( , p<0.05). Leaf and soil inoculation had no distinct effects on the GR ( , p>0.05). In addition, the inoculation of sterile and non-sterile leaves at G0 caused a high mortality rate (MR) (19.7% vs 96.7%) for seedlings growing in Petri dishes. Only non-sterile leaves caused a low percentage of seedling death (8.4%) when the seedlings were inoculated at G21 . Two weeks after transplanting these seedlings into the cups, leaf inoculation caused significantly greater seedling mortality than did soil inoculation (p<0.001); the non-sterile sample caused greater seedling mortality than did the sterile sample, especially leaf inoculation during the G0 and G21 periods. Moreover, non-sterile leaf inoculation at earlier stages significantly increased seedling mortality compared with that at later stages ( , p<0.05). However, seedling mortality did not differ between the high and low nutrient conditions, regardless of leaf or soil inoculation ( , both p>0.05). With the exception of nutrient level, inoculation source and time, as well as their interaction with nutrient level, significantly affected the microbial role in the total dry biomass of A. adenophora (p<0.05, ). Soil and leaf microbial effects on seedling biomass interacted with soil nutrient level and inoculation time: when inoculated at the G0 timepoint, both soil and leaf microbes had adverse effects on A. adenophora growth, as all seedlings inoculated with non-sterile leaves died, and those inoculated with non-sterile soil grew poorer than those inoculated with sterile soil under both low- and high nutrition conditions; when inoculated at the G21 timepoint, both soil and leaf microbes had significantly positive effects only under high nutrition conditions; when inoculated at the G28 timepoint, only soil microbes promoted seedling growth under high nutrition conditions; and when inoculated at the G21+28 timepoint, only leaf microbes had a significantly positive effect on seedling growth under low nutrition conditions .
A. adenophora seedling mortality at the early stage The soil and leaf inocula had distinct microbial diversity, community compositions, and potential functions. The microbial diversity and richness were greater in the soil than in the leaf litter . Top three core bacteria were Rhodoplanes (5.65%), Bradyrizhobium (4.80%), and some unclassified Alphaproteobacteria (4.56%) for soil and Pseudomonas (30.33%), Massilia (17.45%), and Sphingomonas (16.35%) for leaf . Top three core fungal genera were Mortierella (7.00%), Inocybe (5.36%), and Neonectria (3.63%) for soil and Didymella (27.30%), Alternaria (8.95%), and Cryptococcus (4.44%) for leaf . Plant bacterial pathogens (2.29%) were the potential function of the core bacterial taxa in leaves but not in soil; soil had a greater abundance of nitrogen circle-related function (50.98%) than did leaves (17.55%) . The abundance of plant fungal pathogens was greater in leaves (68.20%) than in soil (33.93%) . We further correlated the top 30 microbial genera of leaf and soil inocula with seed germination and seedling mortality in response to inoculation with non-sterile inocula at G0. The abundances of both the soil and leaf microbial genera were related to the seedling mortality rate (MR) but not to the germination time (GT) or GR . Specifically, the leaf core bacterial genera Pseudomonas , Sphingomonas , Massilia , Variovorax , Aureimonas , and Agrobacterium were positively correlated with MR, but the soil core bacteria, including Alphaproteobacteria_unclassified, Rhodoplanes , Vicinamibacteraceae_unclassified, and Pedomicrobium , were negatively correlated with MR . The leaf core fungal genera Didymella , Pleosporales_unclassified, Subplenodomus , and Bulleromyces were positively correlated with MR, but the soil core fungal genera Mortierella , Hypocrea , Pochonia, and Volutella were negatively correlated with MR . We obtained 192 cultivable fungal isolates from 40 dead seedlings, with an average of 4.825 isolates per dead seedling . Based on the ITS genes of the representative strains , they were divided into seven families. The dominant family was Didymellaceae (relative abundance = 66.15%), and the numerically dominant genera were Allophoma (50.52%), Alternaria (26.04%), and Epicoccum (5.73%) . The seedling-killing effects of these strains on A. adenophora exhibited a significant phylogenetic signal (Pagel’s λ=0.82, p=0.0002). Overall, numerically dominant Allophoma (Didymellaceae) and Alternaria (Pleosporaceae) had high seedling mortality (54–100%) .
A. adenophora seedlings under different treatments Nonmetric multidimensional scaling (NMDS) and permutational analysis of variance (PERMANOVA) revealed that all four factors significantly affected the bacterial community and functional assembly of the seedlings, and the greatest effects were inoculation time and compartment (all p<0.05, R 2 : 0.102–0.138), followed by inoculation source and nutrition (all p<0.05, R 2 : 0.024–0.082, ). Additionally, compartment, inoculation source, and time significantly affected the fungal community and functional assembly (all p<0.05, R 2 : 0.054–0.102), but nutrition affected only the fungal community (p=0.001, R 2 =0.031) . Further analysis for each inoculation time treatment showed that compartment and inoculation source mainly affected the microbial community and functional assembly and explained a greater proportion of the variation in bacteria than in fungi and in function than in the community. The nutrient level mainly affected the bacterial community at certain inoculation time (p<0.05 for G0, G21, and G28, , , ).
A. adenophora seedling growth We further analyzed the correlation of microbial abundance and putative functions enriched by seedlings with the microbial effect on seedling growth (response index [RI]). We identified 47 root endophytic genera that were significantly correlated with A. adenophora growth. Among these genera, seven negative genera were less abundant in seedlings treated by leaf inoculation than in those treated by soil inoculation but were similar in abundance in seedlings treated by different inoculation time, and in seedlings grown at different nutrient levels. In contrast, 40 positive genera, e.g., the fungi Duganella and Mortierella and the bacteria Massilia, Pseudomonas, and Sphingomonas , were more abundant in seedlings treated by leaf inoculation than by soil inoculation but less abundant in seedlings inoculated at G0 than at the other three inoculation time treatments . Eighteen leaf endophytic genera with significant correlations with the RI were identified, of which three negative genera, namely, the bacteria Tardiphaga and Brevundimonas and the fungus Microsphaera , were more abundant in seedlings inoculated at G0 than at the other three inoculation time treatments and were slightly more abundant in the seedlings inoculated with soil than in those inoculated with leaf; in contrast, 15 positive genera, e.g., the fungi Hypocrea and Pleosporales_unclassified, were more abundant in the seedlings inoculated with leaf than in those inoculated with soil and were more abundant in the seedlings inoculated at G21 than in the seedlings inoculated at the other three inoculation time treatments . We identified several bacterial functions in the roots and fungal function guilds enriched in the roots and leaves of A. adenophora seedlings that were significantly correlated with the RI . Two bacterial functions involved in the N cycle (nitrate ammonification and nitrite ammonification) in roots showed a significant positive correlation with RI . The fungal guild Ectomycorrhizal showed a significantly positive correlation. Unexpectedly, the putative plant pathogen guild showed a significant positive correlation with seedling growth, while the arbuscular mycorrhizal guild showed a negative correlation . The positive bacterial functions in roots and fungal guilds in leaves had greater relative abundances in the seedlings after leaf inoculation than those after soil inoculation ; additionally, the positive fungal guilds in roots and leaves had significantly greater abundances in the G21 seedlings than in those at the other three inoculation time treatments. Surprisingly, there was no difference in the relative abundance of microbes or functions involved in seedling growth under different nutrient levels .
Leaf litter microbes had more adverse effects on A. adenophora seed germination and seedling survival than soil microbes In support of our first expectation, leaf litter had more adverse effects on seed germination and seedling survival than soil . Leaf litter has been previously reported to have adverse effects on seedling emergence and population establishment by physical barriers and reducing light and releasing allelochemicals . Our study did not directly test the allelopathic effects of leaf litter. However, leaf litter possibly produces allelochemicals that adversely impact A. adenophora seed GT and seedling survival. We observed that sterile leaf litter inoculation caused longer GTs than sterile soil and the control (nothing inoculated) . Interestingly, sterile leaf litter inoculation also caused longer GTs than non-sterile leaf litter inoculation, suggesting that some pathways through which leaf microbes alleviate the adverse effects of leaf allelopathy on GTs are unknown. Moreover, sterile leaf inoculation at G0 caused a 19.7% MR for seedlings growing in Petri dishes , but no dead seedlings were observed when the plants were not inoculated . Nonetheless, our study highlighted the adverse microbial role of leaf litter in seedling mortality because non-sterile leaves have significantly greater seedling mortality (96.7%) than sterile leaves (19.7%) when inoculated at G0 . Indeed, we found that leaf litter harbored more abundant bacterial and fungal genera positively related to seedling mortality, as well as a greater proportion of plant pathogens, than soil . The results implied that the litter harbored many plant pathogens and thus played an essential role in mediating A. adenophora population density by killing conspecific seedlings. These findings provide novel insights for understanding plant invasion. Invasive plants are commonly characterized by rapid growth and a high yield of seeds . These traits are beneficial for rapid population establishment and range expansion at newly invaded sites. However, these invasive plant species commonly form high-density monocultures once the population is established, e.g., A. adenophora . For those perennial invasive species, e.g., A. adenophora , self-limiting of the population elicited by leaf microbes may alleviate intraspecific competition thus in turn help to maintain monocultures at old established sites. Thus, it is highly interesting to determine whether leaf microbe-mediated self-limitation at an early life stage is common and important in other invasive systems. Peripheral microbial sources had more adverse effects on seedling survival and growth when inoculating at the early growth stage than at the later stage Consistent with our second expectation, inoculation time significantly affected seedling survival and growth; in particular, seedling mortality was greater and seedling growth was poorer when inoculated at G0 than at later growth stages . The plant growth stage could change the impact of plant host-associated microbes, and such an impact is always strongest in early plant growth stages . One potential reason is that small seedlings usually allocate most of their resources to survive and grow, while older seedlings have relatively more resources to defend against pathogen infection . For example, smaller seedlings were more sensitive to inoculated individual fungi, soil microbiota, or litter addition than older seedlings due to fewer defense resources and little chance of recovering from biomass loss . Another possible reason is related to the interaction between seed-borne microbes and peripheral microbial sources in young seedlings. There is evidence that seed-borne endophytes are likely to be beneficial for seedling growth and stress resistance . These seed endophytes might be inhibited or even excluded from young seedlings by external sources of microbes inoculated at G0, when seedlings are highly sensitive to inoculated microbes . Therefore, determining how peripheral microbial sources interact with seed-borne endophytes in seedlings across ontogeny is highly valuable. We did not observe an adverse effect of leaf litter microbes on A. adenophora growth, as observed previously by , who inoculated A. adenophora at G0 (sowing seeds), because all seedlings inoculated with leaf litter at G0 died after transplantation into soils in this study. In contrast, both the microbial community and function were significantly positively correlated with seedling growth and had a greater relative abundance in seedlings inoculated with leaf litter than in those inoculated with soil, while those with a negative correlation showed the opposite trend . This finding suggested that leaf litter microbes might have a more positive effect on A. adenophora growth than soil microbes if inoculated during the latter growth stage, such as at G21 and G21+28 . Interestingly, we found that N circle-related bacterial functions in seedling roots were positively correlated with seedling growth . Similarly, , showed that A. adenophora invasion increased N circle-related bacterial functional genes in soil and subsequently directly promoted plant growth and invasion. , also reported that several root endophytic nitrogen-fixing bacteria of A. adenophora could significantly promote its growth. Interestingly, the abundance of these N circle-related bacterial functions was greater in seedling roots inoculated with leaf litter than in those inoculated with soil , which suggests a possible way in which some N circle-related bacteria associated with leaf litter may migrate from leaves into roots after leaf litter inoculation. Nutrient levels did not affect seedling mortality and microbe-mediated A. adenophora growth Nutrient addition can promote severe invasion, as invasive plants often have greater nutrient availability than do native plants . However, it is unclear whether such an advantage is involved in a change in the microbial community-driven host growth effect. We also found that A. adenophora grew larger and more rapidly in the high nutrient treatment than in the low nutrient treatment ; moreover, the nutrient level significantly changed the microbial community and bacterial function . Nonetheless, in contrast to our third expectation, seedling mortality was not affected by different nutrient levels, and nutrient levels negligently affected overall microbe-mediated A. adenophora growth and the relative abundance of microbes and functions correlated with seedling growth . Previously, , reported that nutrient additions cause crops to enrich some bacteria and fungi from soil and increase yield; however, there is no evidence that the increased yield effect is due to enriched microbial communities in this study. Our data indicated that the invasion advantage driven by high nutrient availability may be driven primarily by plant physiological traits, such as rapid nutrient absorption and growth strategies, rather than by enriched microbes. Alternatively, it is possible that our delayed harvest of seedlings under low nutrient levels may cover the distinct microbial role in seedling growth between the two nutrient levels (see Materials and methods). However, there was an interaction effect between nutrient level and inoculation time on seedling growth . For example, a high nutrient level resulted in a more significant positive microbial effect on seedling growth than a low nutrient level when inoculated at G21, regardless of leaf litter or soil inoculation. It is unclear whether, during the first 21 days before inoculation, more beneficial seed endophytes are enriched to produce a more positive effect on seedling growth under high nutrition conditions than under low nutrition conditions, as seed endophytes can facilitate nutrient acquisition and subsequently promote plant growth . The same microbial genera had distinct effects on A. adenophora seedling survival versus growth Correlation analysis of the microbial community and function with seedling survival and growth revealed that several genera showed distinct correlations with seedling survival and growth. For example, the bacterial genera Pseudomonas , Sphingomonas, and Massilia are positively correlated with seedling mortality and subsequent seedling growth ( and ). Many strains belonging to these genera have been reported to promote the growth of many plant species , including A. adenophora , because they are commonly involved in N 2 fixation . However, some microbes, e.g., the fungi Mortierella and Hypocrea , are negatively correlated with early seedling mortality but positively correlated with later seedling growth ( and ). These fungi, as plant growth-promoting fungi, have been widely reported . These findings suggested that microbial interactions are highly complicated during the early life stage of A. adeonophora . On the one hand, there may be sequential effects for some plant growth-promoting microbial groups. For example, the bacteria Massilia, Pseudomonas, and Sphingomonas may negatively affect seedling growth and even kill seedlings if the arrival time is too early after germination. On the other hand, such distinct effects of these bacterial groups on seedling survival versus growth may result from different species from the same genus or even from genetically distinct strains from one species. Interestingly, we found that most Pseudomonas and Sphingomonas amplicon sequence variants (ASVs) enriched in the seedlings (>80%) were not associated with the inoculum source . This suggested that most of the bacterial ASVs positively correlated with growth might be from seeds rather than from the inocula. It is necessary to isolate these enriched microbes to test their interactions with the early life stage of A. adeonophora . Surprisingly, related plant pathogen guilds showed a positive correlation with A. adenophora seedling growth . Because these putative plant pathogens were classified as plant pathogens based on the current database, it remains to determine whether such putative plant pathogens for most native plant species are not detrimental to invader A. adenophora growth. Indeed, plant pathogens often can switch from a beneficial endophyte to a pathogen or vice versa depending on different host plant species . Seedling-killing microbes were those associated with leaf litter We found that most seedling-killing microbes isolated from dead seedlings were previously reported to be leaf spot pathogens. For example, Alternaria (Pleosporaceae) and several genera belonging to the family Didymellaceae, such as Allophoma , Stagonosporopsis , Didymella , Boeremia , and Epicoccum , caused high seedling mortality . Alternaria sp. is often pathogenic to a large variety of plants, such as those causing stem cancer, leaf blight, or leaf spot , and members of Allophoma have also been reported to cause dieback and leaf spot . All these fungi are leaf spot pathogens of A. adenophora and its neighboring native plant . In particular, the numerically dominant Allophoma strains obtained in this study had the same ITS genes as the leaf endophyte and leaf spot pathogen Allophoma associated with A. adenophora . Interestingly, a previous report revealed that the dominant genera in healthy seedlings inoculated with leaf litter were Didymella and Alternaria . We did not isolate fungi from healthy seedlings to determine whether the live seedlings indeed lacked or accumulated a lower abundance of the seedling-killing strains than did the dead seedlings in this study. We could assume that these fungal genera likely exist in A. adenophora mature individual experiencing a lifestyle switch from endophytic to pathogenic and play an essential role in limiting the population density of A. adenophora monocultures by killing seedlings only at very early stages. Thus, it is worth exploring the dynamic abundance of these strains and host resistance variation during A. adenophora seedling development. Implications for developing biocontrol agents for A. adenophora invasion Our data also have implications for the development of biocontrol agents for A. adenophora invasion. Currently, several leaf spot fungi, such as the leaf spot fungus Phaeoramularia sp., which is released against A. adenophora ; the white smut fungus Entyloma ageratinae against A. riparia ; the rust fungus Uromycladium tepperianum against the weed Acacia saligna ; and the rust fungus Puccinia spegazzinii against Mikania micrantha , have been used as biological agents for the control of plant invasion. These agents mainly control weeds by damaging the leaves, stems, and petioles and reducing growth rates, flowering, percentage cover, and population density. In this study, the strains associated with leaf litter, such as Allophoma sp. and Alternaria sp., caused high seedling mortality and thus could control A. adenophora invasion at the seedling establishment stage. On the other hand, we found that an external source of microbes had a greater adverse effect on seedling survival and growth when inoculated at G0 than at the later growth stage. Therefore, prevention and control measures by microbial agents taken at the early seedling stage of invasive plants may be more effective than at the mature stage.
A. adenophora seed germination and seedling survival than soil microbes In support of our first expectation, leaf litter had more adverse effects on seed germination and seedling survival than soil . Leaf litter has been previously reported to have adverse effects on seedling emergence and population establishment by physical barriers and reducing light and releasing allelochemicals . Our study did not directly test the allelopathic effects of leaf litter. However, leaf litter possibly produces allelochemicals that adversely impact A. adenophora seed GT and seedling survival. We observed that sterile leaf litter inoculation caused longer GTs than sterile soil and the control (nothing inoculated) . Interestingly, sterile leaf litter inoculation also caused longer GTs than non-sterile leaf litter inoculation, suggesting that some pathways through which leaf microbes alleviate the adverse effects of leaf allelopathy on GTs are unknown. Moreover, sterile leaf inoculation at G0 caused a 19.7% MR for seedlings growing in Petri dishes , but no dead seedlings were observed when the plants were not inoculated . Nonetheless, our study highlighted the adverse microbial role of leaf litter in seedling mortality because non-sterile leaves have significantly greater seedling mortality (96.7%) than sterile leaves (19.7%) when inoculated at G0 . Indeed, we found that leaf litter harbored more abundant bacterial and fungal genera positively related to seedling mortality, as well as a greater proportion of plant pathogens, than soil . The results implied that the litter harbored many plant pathogens and thus played an essential role in mediating A. adenophora population density by killing conspecific seedlings. These findings provide novel insights for understanding plant invasion. Invasive plants are commonly characterized by rapid growth and a high yield of seeds . These traits are beneficial for rapid population establishment and range expansion at newly invaded sites. However, these invasive plant species commonly form high-density monocultures once the population is established, e.g., A. adenophora . For those perennial invasive species, e.g., A. adenophora , self-limiting of the population elicited by leaf microbes may alleviate intraspecific competition thus in turn help to maintain monocultures at old established sites. Thus, it is highly interesting to determine whether leaf microbe-mediated self-limitation at an early life stage is common and important in other invasive systems.
Consistent with our second expectation, inoculation time significantly affected seedling survival and growth; in particular, seedling mortality was greater and seedling growth was poorer when inoculated at G0 than at later growth stages . The plant growth stage could change the impact of plant host-associated microbes, and such an impact is always strongest in early plant growth stages . One potential reason is that small seedlings usually allocate most of their resources to survive and grow, while older seedlings have relatively more resources to defend against pathogen infection . For example, smaller seedlings were more sensitive to inoculated individual fungi, soil microbiota, or litter addition than older seedlings due to fewer defense resources and little chance of recovering from biomass loss . Another possible reason is related to the interaction between seed-borne microbes and peripheral microbial sources in young seedlings. There is evidence that seed-borne endophytes are likely to be beneficial for seedling growth and stress resistance . These seed endophytes might be inhibited or even excluded from young seedlings by external sources of microbes inoculated at G0, when seedlings are highly sensitive to inoculated microbes . Therefore, determining how peripheral microbial sources interact with seed-borne endophytes in seedlings across ontogeny is highly valuable. We did not observe an adverse effect of leaf litter microbes on A. adenophora growth, as observed previously by , who inoculated A. adenophora at G0 (sowing seeds), because all seedlings inoculated with leaf litter at G0 died after transplantation into soils in this study. In contrast, both the microbial community and function were significantly positively correlated with seedling growth and had a greater relative abundance in seedlings inoculated with leaf litter than in those inoculated with soil, while those with a negative correlation showed the opposite trend . This finding suggested that leaf litter microbes might have a more positive effect on A. adenophora growth than soil microbes if inoculated during the latter growth stage, such as at G21 and G21+28 . Interestingly, we found that N circle-related bacterial functions in seedling roots were positively correlated with seedling growth . Similarly, , showed that A. adenophora invasion increased N circle-related bacterial functional genes in soil and subsequently directly promoted plant growth and invasion. , also reported that several root endophytic nitrogen-fixing bacteria of A. adenophora could significantly promote its growth. Interestingly, the abundance of these N circle-related bacterial functions was greater in seedling roots inoculated with leaf litter than in those inoculated with soil , which suggests a possible way in which some N circle-related bacteria associated with leaf litter may migrate from leaves into roots after leaf litter inoculation.
A. adenophora growth Nutrient addition can promote severe invasion, as invasive plants often have greater nutrient availability than do native plants . However, it is unclear whether such an advantage is involved in a change in the microbial community-driven host growth effect. We also found that A. adenophora grew larger and more rapidly in the high nutrient treatment than in the low nutrient treatment ; moreover, the nutrient level significantly changed the microbial community and bacterial function . Nonetheless, in contrast to our third expectation, seedling mortality was not affected by different nutrient levels, and nutrient levels negligently affected overall microbe-mediated A. adenophora growth and the relative abundance of microbes and functions correlated with seedling growth . Previously, , reported that nutrient additions cause crops to enrich some bacteria and fungi from soil and increase yield; however, there is no evidence that the increased yield effect is due to enriched microbial communities in this study. Our data indicated that the invasion advantage driven by high nutrient availability may be driven primarily by plant physiological traits, such as rapid nutrient absorption and growth strategies, rather than by enriched microbes. Alternatively, it is possible that our delayed harvest of seedlings under low nutrient levels may cover the distinct microbial role in seedling growth between the two nutrient levels (see Materials and methods). However, there was an interaction effect between nutrient level and inoculation time on seedling growth . For example, a high nutrient level resulted in a more significant positive microbial effect on seedling growth than a low nutrient level when inoculated at G21, regardless of leaf litter or soil inoculation. It is unclear whether, during the first 21 days before inoculation, more beneficial seed endophytes are enriched to produce a more positive effect on seedling growth under high nutrition conditions than under low nutrition conditions, as seed endophytes can facilitate nutrient acquisition and subsequently promote plant growth .
A. adenophora seedling survival versus growth Correlation analysis of the microbial community and function with seedling survival and growth revealed that several genera showed distinct correlations with seedling survival and growth. For example, the bacterial genera Pseudomonas , Sphingomonas, and Massilia are positively correlated with seedling mortality and subsequent seedling growth ( and ). Many strains belonging to these genera have been reported to promote the growth of many plant species , including A. adenophora , because they are commonly involved in N 2 fixation . However, some microbes, e.g., the fungi Mortierella and Hypocrea , are negatively correlated with early seedling mortality but positively correlated with later seedling growth ( and ). These fungi, as plant growth-promoting fungi, have been widely reported . These findings suggested that microbial interactions are highly complicated during the early life stage of A. adeonophora . On the one hand, there may be sequential effects for some plant growth-promoting microbial groups. For example, the bacteria Massilia, Pseudomonas, and Sphingomonas may negatively affect seedling growth and even kill seedlings if the arrival time is too early after germination. On the other hand, such distinct effects of these bacterial groups on seedling survival versus growth may result from different species from the same genus or even from genetically distinct strains from one species. Interestingly, we found that most Pseudomonas and Sphingomonas amplicon sequence variants (ASVs) enriched in the seedlings (>80%) were not associated with the inoculum source . This suggested that most of the bacterial ASVs positively correlated with growth might be from seeds rather than from the inocula. It is necessary to isolate these enriched microbes to test their interactions with the early life stage of A. adeonophora . Surprisingly, related plant pathogen guilds showed a positive correlation with A. adenophora seedling growth . Because these putative plant pathogens were classified as plant pathogens based on the current database, it remains to determine whether such putative plant pathogens for most native plant species are not detrimental to invader A. adenophora growth. Indeed, plant pathogens often can switch from a beneficial endophyte to a pathogen or vice versa depending on different host plant species .
We found that most seedling-killing microbes isolated from dead seedlings were previously reported to be leaf spot pathogens. For example, Alternaria (Pleosporaceae) and several genera belonging to the family Didymellaceae, such as Allophoma , Stagonosporopsis , Didymella , Boeremia , and Epicoccum , caused high seedling mortality . Alternaria sp. is often pathogenic to a large variety of plants, such as those causing stem cancer, leaf blight, or leaf spot , and members of Allophoma have also been reported to cause dieback and leaf spot . All these fungi are leaf spot pathogens of A. adenophora and its neighboring native plant . In particular, the numerically dominant Allophoma strains obtained in this study had the same ITS genes as the leaf endophyte and leaf spot pathogen Allophoma associated with A. adenophora . Interestingly, a previous report revealed that the dominant genera in healthy seedlings inoculated with leaf litter were Didymella and Alternaria . We did not isolate fungi from healthy seedlings to determine whether the live seedlings indeed lacked or accumulated a lower abundance of the seedling-killing strains than did the dead seedlings in this study. We could assume that these fungal genera likely exist in A. adenophora mature individual experiencing a lifestyle switch from endophytic to pathogenic and play an essential role in limiting the population density of A. adenophora monocultures by killing seedlings only at very early stages. Thus, it is worth exploring the dynamic abundance of these strains and host resistance variation during A. adenophora seedling development.
A. adenophora invasion Our data also have implications for the development of biocontrol agents for A. adenophora invasion. Currently, several leaf spot fungi, such as the leaf spot fungus Phaeoramularia sp., which is released against A. adenophora ; the white smut fungus Entyloma ageratinae against A. riparia ; the rust fungus Uromycladium tepperianum against the weed Acacia saligna ; and the rust fungus Puccinia spegazzinii against Mikania micrantha , have been used as biological agents for the control of plant invasion. These agents mainly control weeds by damaging the leaves, stems, and petioles and reducing growth rates, flowering, percentage cover, and population density. In this study, the strains associated with leaf litter, such as Allophoma sp. and Alternaria sp., caused high seedling mortality and thus could control A. adenophora invasion at the seedling establishment stage. On the other hand, we found that an external source of microbes had a greater adverse effect on seedling survival and growth when inoculated at G0 than at the later growth stage. Therefore, prevention and control measures by microbial agents taken at the early seedling stage of invasive plants may be more effective than at the mature stage.
Sample collection and preparation All seeds, rhizosphere soil (AAS) and leaf litter (AAL) of A. adenophora were collected from Xishan Forest Park, Kunming city, Yunnan Province (25°55′34″N; 102°38′30″E, 1890 m), on April 9, 2022. We collected dead leaves (litters) from the stems as inoculated leaves to avoid contamination by soil microbes; moreover, dead leaves could better represent litter in natural systems than fresh leaves. All leaf litter and soil samples were collected from five A. adenophora populations ∼200 m away from each other and treated as independent biological replicates. These A. adenophora plants had been grown in monoculture for more than 10 years; thus, their rhizosphere soils and leaf litters were used in our feedback experiment rather than via a typical two-phase approach (the first conditioning phase and the second testing phase) . The collected soil and litter samples were naturally dried in a clean room and weighed. The soil was ground to a 2 mm sieve before weighing. For convenience in the inoculation application, we prepared these samples in leaf litter bags (each containing 2 g of leaf litter) and soil bags (each containing 5 g of soil) as well as 0.1 g of soil or litter (cut into pieces smaller than 2×2 mm 2 ) in centrifuge tubes. All sample bags and tubes were divided into non-sterile and sterile groups and stored at 4°C until inoculation. The sterile groups (soils or litter) were sterilized by gamma irradiation (30 kGy, 30 hr, Huayuan Nuclear Radiation Technology Co., Ltd., Kunming, China), which can kill all microorganisms because no colonies were formed after 7 days of inoculation on PDA media for gamma-irradiated samples (see ); moreover, there was no evidence that this irradiation method changed the chemistry of the samples. The non-sterile groups (soils or litter) were natural samples containing live microorganisms. The natural soil or litter (0.3 g) was weighed into tubes and placed at –80°C until DNA extraction. Experimental design In May 2022, all the seeds were surface sterilized, germinated, and subsequently grown in RXZ-380D growth chambers (Ningbo Southeast Instrument Co., Ltd., Ningbo, China) at a temperature of 25/20°C (day/night), a light intensity of 12 000 lux, a 12 hr photoperiod, and a humidity of 65%. The experimental design is shown in . (1) We inoculated A. adenophora with non-sterile rhizosphere soil and leaf litter of A. adenophora and sterile groups as control groups to distinguish the effects of aboveground microbes from those of underground microbes. (2) We inoculated the soils or leaf litters of A. adenophora at three growth stages (0 day, 21 days, 28 days) to explore the susceptibility of the seedlings to microbial infection and growth effects. We exposed 16 surface-sterilized seeds to litter or rhizosphere soil from A. adenophora when the seeds were sown on a water ager plate (named G0 inoculation). Soils or leaves (0.1 g) were distributed in a thin layer on a plate in 90 mm Petri dishes. Each treatment was replicated five times, resulting in a total of 20 plates (2 inoculum sources × 2 microbial treatments × 5 replicates). We recorded germinated seeds every day from the first to 14th day and on the 21st day after sowing to calculate the GT, GR, and number of dead seedlings on the 28th day. The GT was calculated by the formula GT = Σ(Gi×i)/ΣGi (i: number of days between seed sowing (day 0) and seed germination; Gi: number of seeds germinated on day i), referring to . The GR was calculated as the proportion of germinated seeds on the 21st day after sowing relative to the total number of sown seeds. The seedling MR=the number of dead seedlings/the total number of germinated seedlings. Sixteen surface-sterilized seeds were germinated for 21 days in a water agar plate without soil or litter, after which the number of germinated seedlings was recorded. Plates with more than 10 germinated seedlings were chosen for inoculation of 0.1 g of litter or rhizosphere soil (named G21 inoculation), resulting in a total of 20 plates (2 inoculum sources × 2 microbial treatments × 5 replicates). After 7 days of inoculation, germinated and dead seedlings were recorded to calculate the MR. Sixteen surface-sterilized seeds were germinated for 28 days on a water agar plate without soil or litter, and germinated seeds every day from the first to 14th day and on the 21st day after sowing were recorded to calculate the GT, GR as the control (nothing inoculated). Litter and rhizosphere soil were inoculated when similar-sized seedlings were transplanted into cups after 28 days of germination without soil or litter (named G28 inoculation). For leaf inoculation, 2 g of litter bags was suspended above the plants inside individual cups using twine when three plants were transplanted into the soil in a cup, referring to the study of , such litter had no direct contact with the soils or plants. Litter (2 g) was placed into individual mesh bags made from 10×15 cm 2 cheese cloth squares. Litter bags were removed after 4 weeks to avoid the effects of extra allelochemicals. For soil inoculation, 5 g of rhizosphere soil was added to 65 g of sterile background soil when three seedlings were transplanted into the soil in a cup. Leaf inoculation at G28 was performed to simulate natural microbial spread from the leaf litter to the above part of the seedlings by suspending the leaf bag over the transplanted seedlings without direct contact all the time (see ). This method may result in only microbial species with easy air transmission to infect seedlings. Thus, an additional combination inoculation (named G21+28) was performed to ensure that most leaf microbes had the opportunity to reach the seedlings. Briefly, 0.1 g of litter or rhizosphere soil was inoculated in a plate after 21 days of surface-sterilized seed germination without soil or litter for a total of 20 plates (2 inoculum sources × 2 microbial treatments × 5 replicates). Seedlings were transplanted into the soil in the cup after 7 days of inoculation, and the litter bags were suspended above the plants inside individual cups as part of continuous leaf inoculation. Litter bags were removed after 4 weeks to avoid the effects of extra allelochemicals. Rhizosphere soil (5 g) inoculum was added to 65 g of sterile background soil for soil inoculation. (3) Seedlings from the four inoculation treatments were respectively transplanted into 1000 mL sterile polypropylene cups containing 65 g of sterile background soil (made of Pindstrup substrate, pearlite, and vermiculite at a volume ratio of 8:1:1, and nutrient content of Pindstrup substrate, see ) and 120 mL of sterile tap water after 28 days of growth in Petri dishes after sowing. The background soil was sterilized by autoclaving three times for 2 hr, with a 1 day rest period in-between. Three similar-sized seedlings (∼1 cm high with 4 small leaves) were transplanted into each cup and subsequently thinned to one seedling per cup 2 weeks after transplantation to avoid intraspecific competition. We recorded dead seedlings in each cup before thinning. The cups were sealed with PTFE microbial filter membranes to prevent airborne microbe infection and minimize cross-contamination between treatments and randomly placed in the growth chamber and rearranged randomly every week to mitigate potential positional effects. The same volume of water (as low nutrition) or Hoagland nutrient solution (as high nutrition) was added after seedlings were transplanted into cups if needed until seedling harvesting. In total, our experimental design included 4 inoculation time (G0, G21, G28, G21+28) × 2 inoculum sources (leaf litter, rhizosphere soil) × 2 microbial treatments (sterile, non-sterile) × 2 nutrient levels (high, low) × 5 replicates = 160 cups. Seedlings were harvested after 8 weeks of growth under high nutrient conditions because they grew too fast and touched the PTFE cover; however, we harvested those plants grown under low nutritional conditions after another 4 weeks of growth due to their very small size (see ). No seedlings survived at the G0 inoculation of non-sterile leaf litters when harvested. Stem height, dry aboveground biomass, and underground biomass were measured at harvest. Fresh seedling leaves and roots (0.3 g) from three seedlings per treatment as three replicates were harvested and surface-sterilized and then stored at –80°C until total DNA was extracted for microbial community detection. The aboveground and underground dry biomasses for seedlings with less than 0.3 g fresh weight were obtained by linear regression (see ). Molecular sequencing of the microbial community To link microbial sources (leaf litter and soil) with seed germination, seedling mortality, and subsequent seedling biomass, we sequenced the microbial community associated with inoculum samples (natural AAS and AAL), as well as fresh leaves and roots of A. adenophora seedlings grown in the non-sterile treatments. Total DNA of soil and plant tissue was extracted using the cetyltrimethylammonium bromide (CTAB) method . The quality of the extracted DNA was assessed via electrophoresis in a 1.5% agarose gel using an ND-1000 spectrophotometer (NanoDrop Technology, Wilmington, DE, USA). A Qubit dsDNA HS assay kit (Invitrogen, USA) was used to quantify the DNA concentration. We amplified the bacterial 16S rDNA V4 region and fungal ITS2 region with the primer sets 515F/806R and ITS1F/ITS4, respectively. PCR amplification was performed in a 50 μL mixture containing 12.5 μL of 2× Phanta Max master mix (Thermo Scientific), 2.5 μL of forward primer, 2.5 μL of reverse primer, 50 ng of DNA as a template, and 25 μL of sterile ddH 2 O. The PCR conditions for the bacterial 16S rRNA genes were as follows: 98°C for 30 s, 98°C for 10 s; 35 cycles of 54°C for 30 s, and 72°C for 45 s; and 72°C for 10 min. For the fungal ITS2 region, PCR amplification is performed twice to minimize host plant contamination as much as possible, the first time: 94°C for 5 min, 94°C for 1 min, 20 cycles of 50°C for 50 s, 68°C for 1 min, and 68°C for 10 min; as for the twice: 98°C for 1 min, 98°C for 10 s, 19 cycles of 50°C for 30 s, 72°C for 45 s, and 72°C for 10 min. The PCR products were purified with AMPure XT beads (Beckman Coulter Genomics, Danvers, MA, USA) and quantified with a Qubit (Invitrogen, USA). The amplicon pools were prepared for sequencing, and the size and quantity of the amplicon library were assessed on an Agilent 2100 Bioanalyzer (Agilent, USA) and with the Library Quantification Kit for Illumina (Kapa Biosciences, Woburn, MA, USA), respectively. The libraries were sequenced on a NovaSeq 6000 platform at LC-BIO Biotech Ltd. (Hangzhou, China). High-quality sequences were obtained after removal of low-quality sequences (quality score <20 and sequence length <100 bp). Chimeric sequences were filtered using Vsearch software (v2.3.4). After dereplication using DADA2, we obtained an ASV feature table and feature sequence. The ASV sequences with poor alignment performance and singleton ASVs were discarded. Taxonomic identification of bacteria and fungi was performed against the SILVA (v138) and UNITE (v8.0) databases , respectively. Alpha diversity was calculated by QIIME2, in which the same number of sequences was extracted randomly by reducing the number of sequences to the minimum of some samples. All the sequences obtained in this study have been deposited in the National Center for Biotechnology Information (NCBI) GenBank under SRA accession number PRJNA1008375 for the bacterial 16S rRNA genes and PRJNA1008403 for the fungal ITS2 genetic region. Seedling-killing fungus experiment No dead seedlings were observed from Petri dishes inoculated with non-sterile soils at G0. Thus, we used 40 dead seedlings obtained from Petri dishes inoculated with non-sterile leaf litter at G0 to isolate fungi. Each dead seedling was cut into 1×1 mm 2 pieces, and three tissues were placed on each PDA Petri dish and incubated at ambient temperature (20–25°C) for 6–8 days or until mycelia grew. Hyphal tip cultures were subsequently transferred onto new PDA plates and incubated until pure colonies appeared. Fungal mycelia DNA was also extracted using the CTAB method . We amplified the ITS region of the fungal DNA with the primers ITS4 and ITS5. PCR was performed in a Veriti 96 Well Thermal Cycler (Applied Biosystems Inc, Foster City, CA, USA) in a 50 reactions volume composed of 25 μL of 2× PCR Master Mix, 1 μL of each primer (10 μM), 22 μL of ddH 2 O, and 1 μL of template DNA. The PCR conditions consisted of an initial denaturation at 94°C for 1 min; 35 cycles of denaturation at 94°C for 1 min, annealing at 54°C for 1 min, and extension at 72°C for 1 min; and a final extension at 72°C for 10 min. PCR products were purified, and forward amplicons were sequenced by Sangon Biotech Co., Ltd. (Shanghai, China). The obtained sequences were edited using EDITSEQ and SEQMAN software in the DNASTAR package (DnaStar Inc, Madison, WI, USA). We aligned sequences in MEGA v6.0 using MUSCLE with default parameters , followed by manual checking of alignments. Taxonomic identification was performed against the UNITE (v8.0) database, and BLASTN analyses were performed against the GenBank database. The ITS sequences reported in this study have been deposited in the GenBank database (for accession numbers, see ). BEAST v1.10.4 was used to build a Bayesian phylogenetic tree . The resulting tree was visualized in FigTree v1.4.3. To test the seedling-killing effects of these fungal strains on A. adenophora , 16 surface-sterilized A. adenophora seeds were sown in a water agar plate in a Petri dish. Ten similar-sized seedlings in one Petri dish 21 days after sowing were selected for fungal inoculation. Five Petri dishes were used as five replicates for each strain. Fungi were grown on PDA for 7 days in an incubator at 25°C, after which 3 mm diameter agar discs with fungal mycelia were inoculated into seedlings by touching the leaves or stems (see ). Seedlings were regarded as dead when the leaf and stem became brown and rotten. We recorded the number of dead seedlings after 14 days of inoculation with agar discs and then calculated the MR (=the number of dead seedlings/10). Statistical analysis It is unreasonable to directly compare seedling biomass among treatments because of different harvest time under high or low soil nutrition conditions (see Materials and methods description above), the RI was calculated to evaluate the feedback intensity (or direction) of microbes in the inocula soil or leaf on seedling growth: RI = (variable non-sterile – variable sterile )/variable sterile , a one-sample t test was used to determine the significance between the RI value and zero, where RI >0 and <0 represent microbes that promote or inhibit seedling growth, respectively. Because the MRs of some sterile groups were zero and their RIs were impossible to calculate, we had to directly compare the seedling mortality caused by non-sterile with by sterile samples and perform the analysis of correlation between the MR and microbial composition. Generalized linear models (GLMs) with Gaussian error distributions (identity link) generated by the ‘lme4’ package were used to identify the effects of inoculation source, time, nutrient level treatments, and their interaction on the RIs of plant growth. The R 2 values of the models were obtained by the ‘piecewiseSEM’ package, and p values were estimated using the ANOVA function via chi-squared (χ 2 ) tests in GLMs. The nonparametric Mann-Whitney U test was used to perform all two-group comparisons, and the Kruskal-Wallis test was performed to compare the differences in seedling MR, RI, or microbial relative abundances among the four inoculation time treatments. Nonparametric Mann-Whitney U tests, Kruskal-Wallis tests, and one-sample t tests were performed using SPSS v22.0 (SPSS, Inc, Chicago, IL, USA). NMDS analysis was used to visualize the similarities in bacterial and fungal composition and function among the treatments. PERMANOVA was performed with the ADONIS function in the R (v.4.2.0) package ‘vegan’ to test the differences in the bacterial and fungal communities and functions among the treatment groups. Bacterial functional profiles were predicted using functional annotation of prokaryotic taxa . Fungal functional guilds were inferred using the program FUNGuild, and guild assignments with confidence rankings of ‘Highly probable’ and ‘Probable’ were retained . The core microbial taxa were primarily selected from the ASVs that appeared (100% prevalence) among all the samples. Spearman’s correlation analysis was used to link microbial communities in inoculation sources (leaf litter and soils) with seed germination and seedling mortality in Petri dishes of the non-sterile G0 treatment, as well as to link the RI of seedling growth with microbial communities or functions in seedling leaves or roots by pooling data from all levels of sources, time periods, and nutrients. A correlation was considered significant when the p<0.05. Heatmap plotting was performed in R 4.2.0 with the ‘pheatmap’ package. To examine the phylogenetic signal of the seedling-killing of fungal strains on A. adenophora , we calculated Pagel’s λ with the R package ‘phytools’, which measures the distribution of a trait across a phylogeny. A Pagel’s λ closer to 1 indicated a stronger phylogenetic signal . The remaining figures were visualized in GraphPad Prism v7.0 (GraphPad Software, Inc, San Diego, CA, USA).
All seeds, rhizosphere soil (AAS) and leaf litter (AAL) of A. adenophora were collected from Xishan Forest Park, Kunming city, Yunnan Province (25°55′34″N; 102°38′30″E, 1890 m), on April 9, 2022. We collected dead leaves (litters) from the stems as inoculated leaves to avoid contamination by soil microbes; moreover, dead leaves could better represent litter in natural systems than fresh leaves. All leaf litter and soil samples were collected from five A. adenophora populations ∼200 m away from each other and treated as independent biological replicates. These A. adenophora plants had been grown in monoculture for more than 10 years; thus, their rhizosphere soils and leaf litters were used in our feedback experiment rather than via a typical two-phase approach (the first conditioning phase and the second testing phase) . The collected soil and litter samples were naturally dried in a clean room and weighed. The soil was ground to a 2 mm sieve before weighing. For convenience in the inoculation application, we prepared these samples in leaf litter bags (each containing 2 g of leaf litter) and soil bags (each containing 5 g of soil) as well as 0.1 g of soil or litter (cut into pieces smaller than 2×2 mm 2 ) in centrifuge tubes. All sample bags and tubes were divided into non-sterile and sterile groups and stored at 4°C until inoculation. The sterile groups (soils or litter) were sterilized by gamma irradiation (30 kGy, 30 hr, Huayuan Nuclear Radiation Technology Co., Ltd., Kunming, China), which can kill all microorganisms because no colonies were formed after 7 days of inoculation on PDA media for gamma-irradiated samples (see ); moreover, there was no evidence that this irradiation method changed the chemistry of the samples. The non-sterile groups (soils or litter) were natural samples containing live microorganisms. The natural soil or litter (0.3 g) was weighed into tubes and placed at –80°C until DNA extraction.
In May 2022, all the seeds were surface sterilized, germinated, and subsequently grown in RXZ-380D growth chambers (Ningbo Southeast Instrument Co., Ltd., Ningbo, China) at a temperature of 25/20°C (day/night), a light intensity of 12 000 lux, a 12 hr photoperiod, and a humidity of 65%. The experimental design is shown in . (1) We inoculated A. adenophora with non-sterile rhizosphere soil and leaf litter of A. adenophora and sterile groups as control groups to distinguish the effects of aboveground microbes from those of underground microbes. (2) We inoculated the soils or leaf litters of A. adenophora at three growth stages (0 day, 21 days, 28 days) to explore the susceptibility of the seedlings to microbial infection and growth effects. We exposed 16 surface-sterilized seeds to litter or rhizosphere soil from A. adenophora when the seeds were sown on a water ager plate (named G0 inoculation). Soils or leaves (0.1 g) were distributed in a thin layer on a plate in 90 mm Petri dishes. Each treatment was replicated five times, resulting in a total of 20 plates (2 inoculum sources × 2 microbial treatments × 5 replicates). We recorded germinated seeds every day from the first to 14th day and on the 21st day after sowing to calculate the GT, GR, and number of dead seedlings on the 28th day. The GT was calculated by the formula GT = Σ(Gi×i)/ΣGi (i: number of days between seed sowing (day 0) and seed germination; Gi: number of seeds germinated on day i), referring to . The GR was calculated as the proportion of germinated seeds on the 21st day after sowing relative to the total number of sown seeds. The seedling MR=the number of dead seedlings/the total number of germinated seedlings. Sixteen surface-sterilized seeds were germinated for 21 days in a water agar plate without soil or litter, after which the number of germinated seedlings was recorded. Plates with more than 10 germinated seedlings were chosen for inoculation of 0.1 g of litter or rhizosphere soil (named G21 inoculation), resulting in a total of 20 plates (2 inoculum sources × 2 microbial treatments × 5 replicates). After 7 days of inoculation, germinated and dead seedlings were recorded to calculate the MR. Sixteen surface-sterilized seeds were germinated for 28 days on a water agar plate without soil or litter, and germinated seeds every day from the first to 14th day and on the 21st day after sowing were recorded to calculate the GT, GR as the control (nothing inoculated). Litter and rhizosphere soil were inoculated when similar-sized seedlings were transplanted into cups after 28 days of germination without soil or litter (named G28 inoculation). For leaf inoculation, 2 g of litter bags was suspended above the plants inside individual cups using twine when three plants were transplanted into the soil in a cup, referring to the study of , such litter had no direct contact with the soils or plants. Litter (2 g) was placed into individual mesh bags made from 10×15 cm 2 cheese cloth squares. Litter bags were removed after 4 weeks to avoid the effects of extra allelochemicals. For soil inoculation, 5 g of rhizosphere soil was added to 65 g of sterile background soil when three seedlings were transplanted into the soil in a cup. Leaf inoculation at G28 was performed to simulate natural microbial spread from the leaf litter to the above part of the seedlings by suspending the leaf bag over the transplanted seedlings without direct contact all the time (see ). This method may result in only microbial species with easy air transmission to infect seedlings. Thus, an additional combination inoculation (named G21+28) was performed to ensure that most leaf microbes had the opportunity to reach the seedlings. Briefly, 0.1 g of litter or rhizosphere soil was inoculated in a plate after 21 days of surface-sterilized seed germination without soil or litter for a total of 20 plates (2 inoculum sources × 2 microbial treatments × 5 replicates). Seedlings were transplanted into the soil in the cup after 7 days of inoculation, and the litter bags were suspended above the plants inside individual cups as part of continuous leaf inoculation. Litter bags were removed after 4 weeks to avoid the effects of extra allelochemicals. Rhizosphere soil (5 g) inoculum was added to 65 g of sterile background soil for soil inoculation. (3) Seedlings from the four inoculation treatments were respectively transplanted into 1000 mL sterile polypropylene cups containing 65 g of sterile background soil (made of Pindstrup substrate, pearlite, and vermiculite at a volume ratio of 8:1:1, and nutrient content of Pindstrup substrate, see ) and 120 mL of sterile tap water after 28 days of growth in Petri dishes after sowing. The background soil was sterilized by autoclaving three times for 2 hr, with a 1 day rest period in-between. Three similar-sized seedlings (∼1 cm high with 4 small leaves) were transplanted into each cup and subsequently thinned to one seedling per cup 2 weeks after transplantation to avoid intraspecific competition. We recorded dead seedlings in each cup before thinning. The cups were sealed with PTFE microbial filter membranes to prevent airborne microbe infection and minimize cross-contamination between treatments and randomly placed in the growth chamber and rearranged randomly every week to mitigate potential positional effects. The same volume of water (as low nutrition) or Hoagland nutrient solution (as high nutrition) was added after seedlings were transplanted into cups if needed until seedling harvesting. In total, our experimental design included 4 inoculation time (G0, G21, G28, G21+28) × 2 inoculum sources (leaf litter, rhizosphere soil) × 2 microbial treatments (sterile, non-sterile) × 2 nutrient levels (high, low) × 5 replicates = 160 cups. Seedlings were harvested after 8 weeks of growth under high nutrient conditions because they grew too fast and touched the PTFE cover; however, we harvested those plants grown under low nutritional conditions after another 4 weeks of growth due to their very small size (see ). No seedlings survived at the G0 inoculation of non-sterile leaf litters when harvested. Stem height, dry aboveground biomass, and underground biomass were measured at harvest. Fresh seedling leaves and roots (0.3 g) from three seedlings per treatment as three replicates were harvested and surface-sterilized and then stored at –80°C until total DNA was extracted for microbial community detection. The aboveground and underground dry biomasses for seedlings with less than 0.3 g fresh weight were obtained by linear regression (see ).
To link microbial sources (leaf litter and soil) with seed germination, seedling mortality, and subsequent seedling biomass, we sequenced the microbial community associated with inoculum samples (natural AAS and AAL), as well as fresh leaves and roots of A. adenophora seedlings grown in the non-sterile treatments. Total DNA of soil and plant tissue was extracted using the cetyltrimethylammonium bromide (CTAB) method . The quality of the extracted DNA was assessed via electrophoresis in a 1.5% agarose gel using an ND-1000 spectrophotometer (NanoDrop Technology, Wilmington, DE, USA). A Qubit dsDNA HS assay kit (Invitrogen, USA) was used to quantify the DNA concentration. We amplified the bacterial 16S rDNA V4 region and fungal ITS2 region with the primer sets 515F/806R and ITS1F/ITS4, respectively. PCR amplification was performed in a 50 μL mixture containing 12.5 μL of 2× Phanta Max master mix (Thermo Scientific), 2.5 μL of forward primer, 2.5 μL of reverse primer, 50 ng of DNA as a template, and 25 μL of sterile ddH 2 O. The PCR conditions for the bacterial 16S rRNA genes were as follows: 98°C for 30 s, 98°C for 10 s; 35 cycles of 54°C for 30 s, and 72°C for 45 s; and 72°C for 10 min. For the fungal ITS2 region, PCR amplification is performed twice to minimize host plant contamination as much as possible, the first time: 94°C for 5 min, 94°C for 1 min, 20 cycles of 50°C for 50 s, 68°C for 1 min, and 68°C for 10 min; as for the twice: 98°C for 1 min, 98°C for 10 s, 19 cycles of 50°C for 30 s, 72°C for 45 s, and 72°C for 10 min. The PCR products were purified with AMPure XT beads (Beckman Coulter Genomics, Danvers, MA, USA) and quantified with a Qubit (Invitrogen, USA). The amplicon pools were prepared for sequencing, and the size and quantity of the amplicon library were assessed on an Agilent 2100 Bioanalyzer (Agilent, USA) and with the Library Quantification Kit for Illumina (Kapa Biosciences, Woburn, MA, USA), respectively. The libraries were sequenced on a NovaSeq 6000 platform at LC-BIO Biotech Ltd. (Hangzhou, China). High-quality sequences were obtained after removal of low-quality sequences (quality score <20 and sequence length <100 bp). Chimeric sequences were filtered using Vsearch software (v2.3.4). After dereplication using DADA2, we obtained an ASV feature table and feature sequence. The ASV sequences with poor alignment performance and singleton ASVs were discarded. Taxonomic identification of bacteria and fungi was performed against the SILVA (v138) and UNITE (v8.0) databases , respectively. Alpha diversity was calculated by QIIME2, in which the same number of sequences was extracted randomly by reducing the number of sequences to the minimum of some samples. All the sequences obtained in this study have been deposited in the National Center for Biotechnology Information (NCBI) GenBank under SRA accession number PRJNA1008375 for the bacterial 16S rRNA genes and PRJNA1008403 for the fungal ITS2 genetic region.
No dead seedlings were observed from Petri dishes inoculated with non-sterile soils at G0. Thus, we used 40 dead seedlings obtained from Petri dishes inoculated with non-sterile leaf litter at G0 to isolate fungi. Each dead seedling was cut into 1×1 mm 2 pieces, and three tissues were placed on each PDA Petri dish and incubated at ambient temperature (20–25°C) for 6–8 days or until mycelia grew. Hyphal tip cultures were subsequently transferred onto new PDA plates and incubated until pure colonies appeared. Fungal mycelia DNA was also extracted using the CTAB method . We amplified the ITS region of the fungal DNA with the primers ITS4 and ITS5. PCR was performed in a Veriti 96 Well Thermal Cycler (Applied Biosystems Inc, Foster City, CA, USA) in a 50 reactions volume composed of 25 μL of 2× PCR Master Mix, 1 μL of each primer (10 μM), 22 μL of ddH 2 O, and 1 μL of template DNA. The PCR conditions consisted of an initial denaturation at 94°C for 1 min; 35 cycles of denaturation at 94°C for 1 min, annealing at 54°C for 1 min, and extension at 72°C for 1 min; and a final extension at 72°C for 10 min. PCR products were purified, and forward amplicons were sequenced by Sangon Biotech Co., Ltd. (Shanghai, China). The obtained sequences were edited using EDITSEQ and SEQMAN software in the DNASTAR package (DnaStar Inc, Madison, WI, USA). We aligned sequences in MEGA v6.0 using MUSCLE with default parameters , followed by manual checking of alignments. Taxonomic identification was performed against the UNITE (v8.0) database, and BLASTN analyses were performed against the GenBank database. The ITS sequences reported in this study have been deposited in the GenBank database (for accession numbers, see ). BEAST v1.10.4 was used to build a Bayesian phylogenetic tree . The resulting tree was visualized in FigTree v1.4.3. To test the seedling-killing effects of these fungal strains on A. adenophora , 16 surface-sterilized A. adenophora seeds were sown in a water agar plate in a Petri dish. Ten similar-sized seedlings in one Petri dish 21 days after sowing were selected for fungal inoculation. Five Petri dishes were used as five replicates for each strain. Fungi were grown on PDA for 7 days in an incubator at 25°C, after which 3 mm diameter agar discs with fungal mycelia were inoculated into seedlings by touching the leaves or stems (see ). Seedlings were regarded as dead when the leaf and stem became brown and rotten. We recorded the number of dead seedlings after 14 days of inoculation with agar discs and then calculated the MR (=the number of dead seedlings/10).
It is unreasonable to directly compare seedling biomass among treatments because of different harvest time under high or low soil nutrition conditions (see Materials and methods description above), the RI was calculated to evaluate the feedback intensity (or direction) of microbes in the inocula soil or leaf on seedling growth: RI = (variable non-sterile – variable sterile )/variable sterile , a one-sample t test was used to determine the significance between the RI value and zero, where RI >0 and <0 represent microbes that promote or inhibit seedling growth, respectively. Because the MRs of some sterile groups were zero and their RIs were impossible to calculate, we had to directly compare the seedling mortality caused by non-sterile with by sterile samples and perform the analysis of correlation between the MR and microbial composition. Generalized linear models (GLMs) with Gaussian error distributions (identity link) generated by the ‘lme4’ package were used to identify the effects of inoculation source, time, nutrient level treatments, and their interaction on the RIs of plant growth. The R 2 values of the models were obtained by the ‘piecewiseSEM’ package, and p values were estimated using the ANOVA function via chi-squared (χ 2 ) tests in GLMs. The nonparametric Mann-Whitney U test was used to perform all two-group comparisons, and the Kruskal-Wallis test was performed to compare the differences in seedling MR, RI, or microbial relative abundances among the four inoculation time treatments. Nonparametric Mann-Whitney U tests, Kruskal-Wallis tests, and one-sample t tests were performed using SPSS v22.0 (SPSS, Inc, Chicago, IL, USA). NMDS analysis was used to visualize the similarities in bacterial and fungal composition and function among the treatments. PERMANOVA was performed with the ADONIS function in the R (v.4.2.0) package ‘vegan’ to test the differences in the bacterial and fungal communities and functions among the treatment groups. Bacterial functional profiles were predicted using functional annotation of prokaryotic taxa . Fungal functional guilds were inferred using the program FUNGuild, and guild assignments with confidence rankings of ‘Highly probable’ and ‘Probable’ were retained . The core microbial taxa were primarily selected from the ASVs that appeared (100% prevalence) among all the samples. Spearman’s correlation analysis was used to link microbial communities in inoculation sources (leaf litter and soils) with seed germination and seedling mortality in Petri dishes of the non-sterile G0 treatment, as well as to link the RI of seedling growth with microbial communities or functions in seedling leaves or roots by pooling data from all levels of sources, time periods, and nutrients. A correlation was considered significant when the p<0.05. Heatmap plotting was performed in R 4.2.0 with the ‘pheatmap’ package. To examine the phylogenetic signal of the seedling-killing of fungal strains on A. adenophora , we calculated Pagel’s λ with the R package ‘phytools’, which measures the distribution of a trait across a phylogeny. A Pagel’s λ closer to 1 indicated a stronger phylogenetic signal . The remaining figures were visualized in GraphPad Prism v7.0 (GraphPad Software, Inc, San Diego, CA, USA).
|
Identification of susceptibility variants to benign childhood epilepsy with centro-temporal spikes (BECTS) in Chinese Han population | e2ab23b7-c0c6-470c-8540-0dfe89046a5c | 7317238 | Pediatrics[mh] | Introduction Benign Childhood Epilepsy with Centro-temporal Spikes (BECTS) is the most common form of idiopathic epilepsy in children, accounting for 8 to 23% of epilepsy in children less than 16 years of age . The typical age of onset is 3 to 14 years, and the condition typically resolves by the early teenage years. Whilst affected children are usually neurodevelopmentally normal, BECTS has been associated with varying degrees of neuropsychological damage, and can be associated with sociological and behavioral problems in adulthood . Research in context Evidence before this study We investigated all the manuscripts on BECTS in PubMed and China National Knowledge Infrastructure (CNKI in Chinese language). BECTS is the most common form of idiopathic epilepsy in children. Whilst increased familiality of BECTS has raised the hypothesis of BECTS to be a genetic disorder, it remains elusive whether there is a major genetic contribution to susceptibility. To date, formal heritability studies in twins have not supported the hypothesis, and it has not been investigated in a GWAS study to pinpoint the genetic predisposition and further the disease causality. Added value of this study Here, we identified 12 loci suggestively associated with BECTS loci in a case-control GWAS study of Chinese Han population. The post-GWAS analysis indicates BECTS associated variant rs1948 is significantly associated with BECTS through effects on expression of CHRNA5 in brain tissue. Using a generalized SMR approach we demonstrate that maternal smoking around birth is significantly associated with increased risk of BECTS. Further, significant common variant heritability of BECTS was observed in the combined phase 1 and 2 datasets which was > 10% even for assumed prevalence as low as 0.00025. Implications of all the available evidence It suggests the involvements of multiple susceptible genes in pathogenesis of BECTS. BECTS risk is at least partially heritable and explained by common genetic variants. Additionally, maternal smoking around birth is casually associated with onset of BECTS in offspring. Alt-text: Unlabelled box Increased familiality of BECTS has raised the hypothesis that the condition has genetic underpinnings (reviewed in ). However familial recurrence may occur because of shared environmental or genetic factors, and to date formal heritability studies in twins have not supported a major genetic contribution to susceptibility . To better characterize the genetic architecture of BECTS, a large two-stage case-control GWAS was conducted in the Han Chinese population of 1800 BECTS cases and 7090 healthy ethnically matched controls. A significant common variant heritability of the disease was demonstrated, and several loci identified with suggestive association with BECTS risk.
Materials and methods 2.1 Subjects Patients with BECTS were recruited from outpatient clinics at 30 hospitals in Beijing, Shanghai, and 26 other Chinese provincial capital cities. Diagnosis was determined according to the International League Against Epilepsy (Commission on Classification and Terminology of the International League Against Epilepsy, 1989) definition for BECTS by pediatric neurologists. A patient was diagnosed with epilepsy if he/she had at least two unprovoked epileptic seizures. Healthy, unrelated adult blood donors, from the Beijing and Shanghai were included as controls. Only self-reported Han Chinese ethnicity cases and controls were recruited, and additional ethnicity checks were performed as described below. Written informed consent was obtained from all the parents or guardians, or directly from adult participants, and the study was approved by the relevant ethics committees of the hospitals and institutions involved. 2.2 Genotyping and quality control DNA was isolated from venous blood samples of study participants. Genome-wide genotyping was performed with Illumina OmniZhongHua-8 version1.0 BeadChips (Stage 1) and Illumina Human CoreExome-24 version 1.0 BeadChips (Stage 2) on an Illumina iScan array scanner at the Laboratory of Department of Rheumatology and Immunology, Changzheng Hospital (Shanghai, China). Genotype calls were made using the Illumina BeadStudio software; all SNPs with quality scores <0·15 were excluded. The cluster plots of the top-associated single nucleotide polymorphisms (SNPs) were inspected manually. Genome-wide association analysis was performed using PLINK. We excluded individuals with call rates below 98% and heterozygosity rates >3 standard deviations from the mean. Duplicate subjects or probable relatives were identified by identify-by-descent (IBD) analysis (PI_HAT>0·1875) and excluded. SNP markers were excluded if they had a minor allele frequency (MAF) below 0·01, a genotype distribution out of Hardy-Weinberg equilibrium ( P< 10 −7 ), or had a high rate of missing genotype calls (missing genotype call rate>0·02). Outliers on heterozygosity vs missingness plots were also excluded . Sex chromosomes were excluded from the analysis; only autosomal SNPs were analyzed. After these quality control steps, to detect and correct for population stratification we used the Shellfish software ( http://www.stats.ox.ac.uk/~davison/software/shellfish/shellfish.php ), having first excluded regions of long range LD. To confirm ethnicity, we performed a continental PCA on the Han Chinese dataset, merged with available data from 51 available populations genotyped by Illumina 650Y from the Human Genome Diversity Panel (HGDP-CEPH) . Continental PCA indicated that all the samples came from subjects of Han Chinese descent (East Asian) . Cases or controls lying more than 6 standard deviations from the population mean on principal components (PCs) 1–10 were then excluded . 2.3 Data analysis Imputation was performed separately according to SNP microarray used, using the 1000 Genome reference through the Sanger Imputation Service (imputation.sanger.ac.uk/). BECTS associations of all markers with a MAF>0.01 and imputation quality INFO>0·8 (6,563,936 SNPs in the OmniZhonghua set and 5,742,369 SNPs in the CoreExome set) were analyzed by using logistic regression (PLINK) with dosage output, adding the top PCs as covariates (4 PCs for the OmniZhonghua set; 3 PCs for the CoreExome set). The selection of the numbers of PCs that used to control for population stratification was based on Scree plot findings and calculation of lambda was performed after controlling for top PCs. A meta-analysis was then performed combining SNPs genotyped or imputed in both datasets using METAL software . Heritability was calculated by GCTA software using both observed and liability scales, the latter for population prevalences of 0·00025, 0·0005, 0·001, 0·002, 0·003, 0·004, and 0·005. Because small errors for each SNP can accumulate to give incorrect estimates for genetic variance, for this analysis additional, extremely stringent, QC steps were employed. We excluded SNPs whose p values were <0·05 for either the Hardy-Weinberg equilibrium test or for missingness-difference between cases and controls. To keep individuals who were only distantly related, both individuals from a pair with an estimated relationship statistic PI_HAT>0·05 were excluded. The Summary-data-based Mendelian Randomization (SMR) analysis method was used to further analysis the suggestive GWAS hits obtained in our study. This approach analyses the intersection of genetic effects on gene-expression (eQTLs) and on the trait/disease, to determine the most likely candidate gene at individual loci. We used SMR analysis to prioritize these candidate genes on the basic of the GWAS and eQTL data (BRAINEAC eQTL data from brain tissue samples. http://www.braineac.org/ ). A generalised SMR (GSMR) analysis was performed to investigate the relationship between smoking and risk BECTS . Summary-level GWAS data from the current study and from analysis of “maternal smoking around birth” from the UK Biobank (Data-Field 1787, N case/control = 95,182/214,760; total N = 309,942, http://www.nealelab.is/uk-biobank/ ) were analysed, using the Complex Traits Genetics Virtual Lab platform ( https://vl.genoma.io ) . Outlier SNPs that have apparent pleiotropic effects on both maternal smoking around birth and BECTS and could therefore bias the GSMR findings were identified and excluded using the heterogeneity in dependent instruments (HEIDI)-outlier method.
Subjects Patients with BECTS were recruited from outpatient clinics at 30 hospitals in Beijing, Shanghai, and 26 other Chinese provincial capital cities. Diagnosis was determined according to the International League Against Epilepsy (Commission on Classification and Terminology of the International League Against Epilepsy, 1989) definition for BECTS by pediatric neurologists. A patient was diagnosed with epilepsy if he/she had at least two unprovoked epileptic seizures. Healthy, unrelated adult blood donors, from the Beijing and Shanghai were included as controls. Only self-reported Han Chinese ethnicity cases and controls were recruited, and additional ethnicity checks were performed as described below. Written informed consent was obtained from all the parents or guardians, or directly from adult participants, and the study was approved by the relevant ethics committees of the hospitals and institutions involved.
Genotyping and quality control DNA was isolated from venous blood samples of study participants. Genome-wide genotyping was performed with Illumina OmniZhongHua-8 version1.0 BeadChips (Stage 1) and Illumina Human CoreExome-24 version 1.0 BeadChips (Stage 2) on an Illumina iScan array scanner at the Laboratory of Department of Rheumatology and Immunology, Changzheng Hospital (Shanghai, China). Genotype calls were made using the Illumina BeadStudio software; all SNPs with quality scores <0·15 were excluded. The cluster plots of the top-associated single nucleotide polymorphisms (SNPs) were inspected manually. Genome-wide association analysis was performed using PLINK. We excluded individuals with call rates below 98% and heterozygosity rates >3 standard deviations from the mean. Duplicate subjects or probable relatives were identified by identify-by-descent (IBD) analysis (PI_HAT>0·1875) and excluded. SNP markers were excluded if they had a minor allele frequency (MAF) below 0·01, a genotype distribution out of Hardy-Weinberg equilibrium ( P< 10 −7 ), or had a high rate of missing genotype calls (missing genotype call rate>0·02). Outliers on heterozygosity vs missingness plots were also excluded . Sex chromosomes were excluded from the analysis; only autosomal SNPs were analyzed. After these quality control steps, to detect and correct for population stratification we used the Shellfish software ( http://www.stats.ox.ac.uk/~davison/software/shellfish/shellfish.php ), having first excluded regions of long range LD. To confirm ethnicity, we performed a continental PCA on the Han Chinese dataset, merged with available data from 51 available populations genotyped by Illumina 650Y from the Human Genome Diversity Panel (HGDP-CEPH) . Continental PCA indicated that all the samples came from subjects of Han Chinese descent (East Asian) . Cases or controls lying more than 6 standard deviations from the population mean on principal components (PCs) 1–10 were then excluded .
Data analysis Imputation was performed separately according to SNP microarray used, using the 1000 Genome reference through the Sanger Imputation Service (imputation.sanger.ac.uk/). BECTS associations of all markers with a MAF>0.01 and imputation quality INFO>0·8 (6,563,936 SNPs in the OmniZhonghua set and 5,742,369 SNPs in the CoreExome set) were analyzed by using logistic regression (PLINK) with dosage output, adding the top PCs as covariates (4 PCs for the OmniZhonghua set; 3 PCs for the CoreExome set). The selection of the numbers of PCs that used to control for population stratification was based on Scree plot findings and calculation of lambda was performed after controlling for top PCs. A meta-analysis was then performed combining SNPs genotyped or imputed in both datasets using METAL software . Heritability was calculated by GCTA software using both observed and liability scales, the latter for population prevalences of 0·00025, 0·0005, 0·001, 0·002, 0·003, 0·004, and 0·005. Because small errors for each SNP can accumulate to give incorrect estimates for genetic variance, for this analysis additional, extremely stringent, QC steps were employed. We excluded SNPs whose p values were <0·05 for either the Hardy-Weinberg equilibrium test or for missingness-difference between cases and controls. To keep individuals who were only distantly related, both individuals from a pair with an estimated relationship statistic PI_HAT>0·05 were excluded. The Summary-data-based Mendelian Randomization (SMR) analysis method was used to further analysis the suggestive GWAS hits obtained in our study. This approach analyses the intersection of genetic effects on gene-expression (eQTLs) and on the trait/disease, to determine the most likely candidate gene at individual loci. We used SMR analysis to prioritize these candidate genes on the basic of the GWAS and eQTL data (BRAINEAC eQTL data from brain tissue samples. http://www.braineac.org/ ). A generalised SMR (GSMR) analysis was performed to investigate the relationship between smoking and risk BECTS . Summary-level GWAS data from the current study and from analysis of “maternal smoking around birth” from the UK Biobank (Data-Field 1787, N case/control = 95,182/214,760; total N = 309,942, http://www.nealelab.is/uk-biobank/ ) were analysed, using the Complex Traits Genetics Virtual Lab platform ( https://vl.genoma.io ) . Outlier SNPs that have apparent pleiotropic effects on both maternal smoking around birth and BECTS and could therefore bias the GSMR findings were identified and excluded using the heterogeneity in dependent instruments (HEIDI)-outlier method.
Results 3.1 Genome-wide association analysis The first stage of GWAS included individuals of 997 BECTS cases and 3115 healthy controls. 972 cases and 2916 controls genotyped using the Illumina OmiZhongua SNP microarray for 805,150 SNPs were retained after a series of stringent quality control (QC) procedures. Minimal residual genomic inflation was observed (genomic inflation factor after correcting top 4 PCs=1·033) . The second stage GWAS was performed with an independent cohort of 803 BECTS cases and 3975 controls. 777 cases and 3768 controls were genotyped using the Illumina Core-Exome SNP microarray for 257,007 SNPs and left/remained for association analysis after QC procedures. Again, minimal residual genomic inflation was observed (genomic inflation factor=1·034) after correcting for the top 3 principal components (PC). SNP imputation was performed on each stage (imputed lambda of OmniZhonghua and CoreExome cohorts are 1·042 and 1·039 respectively) and imputed genotype data was combined by meta-analysis (final analysis 5,352,724 SNPs in 1738 cases and 6592 controls). Whilst no locus in this analysis achieved genome-wide significance ( P <5 × 10 −8 ), 12 independent loci reached suggestive significance (5 × 10 −8 < P <10 −5 , , ), the direction of association being consistent between datasets for each locus ( P = 2·44 × 10 −4 ). The strongest associations observed were multiple SNPs in regions on chromosome 3, 15 and 10, located in or nearby the genes KALRN, CHRNB4 , and PTCHD3/RAB18 (zoom plots, see -B and , respectively). The most strongly associated SNP, rs1561578 , is within an intron of KALRN , which encodes the protein kalirin. The second associated signal rs1948 was encompassed by several genes that encode nicotinic cholinergic receptor subunits ( CHRNB4/CHRNA5/CHRNA3 ¸ ). The third locus on chromosome 10p12.1 (peak SNP rs139905806, ) is a region which has been previously associated with lamotrigine-induced skin rash in a previous GWAS study in Korean patients with epilepsy . 3.2 Summary-data-based mendelian randomization analysis eQTL analysis of the probes of the 12 most strongly-associated loci demonstrated that 9 were associated with transcriptional levels of the most proximal gene (P eqtl <0·05). To investigate whether the observed genetic associations operated through these transcriptional effects, a Summary-data-based Mendelian Randomization (SMR) analysis was performed, using brain tissue gene-expression data. Significant association was observed at CHRNA5 locus, tagged by rs76712448 (P smr =0·028, ). A HEIDI test supported this SNP being directly associated with BECTS and CHRNA5 expression. The most strongly associated GWAS hit, rs1948, shows significant association at the CHRNA5 loci in the SMR analysis (P smr =7·9 × 10 −5 ). The association suggests that more than one SNP on the haplotype tagged by rs1948 is associated with BECTS susceptibility through effects on central nervous system (CNS) CHRNA5 expression. 3.3 Generalised summary-data-based mendelian randomisation analysis To test for potential causal association between maternal smoking around birth and BECTS, we conducted a GSMR analysis. GSMR algorithm selected 5 independent SNPs as instruments from the GWAS of maternal smoking at birth at a genome-wide significance level ( P GWAS < 5 × 10 –8 , ). We identified a risk effect of maternal smoking around birth for BECTS ( b xy =1.36 for BECTS is approximately equivalent to odds ratio = 3·90, P = 0·0099) . 3.4 Heritability of BECTS Common variant heritability was tested using the unrelated cases and controls from the combined imputed dataset. The h 2 estimates were calculated for a range of BECTS prevalences , showing that h 2 captured by these SNPs was > 0·10 for prevalence 0·00,025.
Genome-wide association analysis The first stage of GWAS included individuals of 997 BECTS cases and 3115 healthy controls. 972 cases and 2916 controls genotyped using the Illumina OmiZhongua SNP microarray for 805,150 SNPs were retained after a series of stringent quality control (QC) procedures. Minimal residual genomic inflation was observed (genomic inflation factor after correcting top 4 PCs=1·033) . The second stage GWAS was performed with an independent cohort of 803 BECTS cases and 3975 controls. 777 cases and 3768 controls were genotyped using the Illumina Core-Exome SNP microarray for 257,007 SNPs and left/remained for association analysis after QC procedures. Again, minimal residual genomic inflation was observed (genomic inflation factor=1·034) after correcting for the top 3 principal components (PC). SNP imputation was performed on each stage (imputed lambda of OmniZhonghua and CoreExome cohorts are 1·042 and 1·039 respectively) and imputed genotype data was combined by meta-analysis (final analysis 5,352,724 SNPs in 1738 cases and 6592 controls). Whilst no locus in this analysis achieved genome-wide significance ( P <5 × 10 −8 ), 12 independent loci reached suggestive significance (5 × 10 −8 < P <10 −5 , , ), the direction of association being consistent between datasets for each locus ( P = 2·44 × 10 −4 ). The strongest associations observed were multiple SNPs in regions on chromosome 3, 15 and 10, located in or nearby the genes KALRN, CHRNB4 , and PTCHD3/RAB18 (zoom plots, see -B and , respectively). The most strongly associated SNP, rs1561578 , is within an intron of KALRN , which encodes the protein kalirin. The second associated signal rs1948 was encompassed by several genes that encode nicotinic cholinergic receptor subunits ( CHRNB4/CHRNA5/CHRNA3 ¸ ). The third locus on chromosome 10p12.1 (peak SNP rs139905806, ) is a region which has been previously associated with lamotrigine-induced skin rash in a previous GWAS study in Korean patients with epilepsy .
Summary-data-based mendelian randomization analysis eQTL analysis of the probes of the 12 most strongly-associated loci demonstrated that 9 were associated with transcriptional levels of the most proximal gene (P eqtl <0·05). To investigate whether the observed genetic associations operated through these transcriptional effects, a Summary-data-based Mendelian Randomization (SMR) analysis was performed, using brain tissue gene-expression data. Significant association was observed at CHRNA5 locus, tagged by rs76712448 (P smr =0·028, ). A HEIDI test supported this SNP being directly associated with BECTS and CHRNA5 expression. The most strongly associated GWAS hit, rs1948, shows significant association at the CHRNA5 loci in the SMR analysis (P smr =7·9 × 10 −5 ). The association suggests that more than one SNP on the haplotype tagged by rs1948 is associated with BECTS susceptibility through effects on central nervous system (CNS) CHRNA5 expression.
Generalised summary-data-based mendelian randomisation analysis To test for potential causal association between maternal smoking around birth and BECTS, we conducted a GSMR analysis. GSMR algorithm selected 5 independent SNPs as instruments from the GWAS of maternal smoking at birth at a genome-wide significance level ( P GWAS < 5 × 10 –8 , ). We identified a risk effect of maternal smoking around birth for BECTS ( b xy =1.36 for BECTS is approximately equivalent to odds ratio = 3·90, P = 0·0099) .
Heritability of BECTS Common variant heritability was tested using the unrelated cases and controls from the combined imputed dataset. The h 2 estimates were calculated for a range of BECTS prevalences , showing that h 2 captured by these SNPs was > 0·10 for prevalence 0·00,025.
Discussion BECTS is the prototypic illness of a group of epileptiform diseases characterized by the electroencephalogram (EEG) finding of rolandic spike and wave discharges (predominantly centro-temporal spikes). Despite findings from twin studies not supporting a genetic etiology, the tendency of BECTS to run in families has spurred efforts to identify genetic variants associated with the disease including candidate gene association studies, and sequencing studies aimed at identifying rare variant associations. To date though no robust genetic association has been reported with the condition. This study demonstrates that BECTS does have significant common variant heritability. The cumulative incidence studies indicate that up to the age of 15 years, 1·0–1·7% of children will have at least one unprovoked seizure, and 0·7–0·8% repeated seizures , and BECTS represents ~15% of childhood epilepsy . This study demonstrates that even assuming a prevalence or a lifetime risk of BECTS as low as 2·5/10,000, the common variant heritability of the condition remains > 0·10. Assuming a disease prevalence of 0·1%, the liability scale heritability in the combined phase 1 and 2 datasets was 0·123 (standard error 0·020, P = 3·74 × 10 −11 ), reflecting the heritability captured by the SNPs genotyped on the two microarrays used and imputed from them. While conducting GWAS analysis, rare SNPs with MAF<1% were excluded, and thus this heritability figure does not include rare variants effects. Nor does the association analyses include heritability of other forms of genetic variation that may be either missed or incompletely captured by the SNPs studied, such as copy number variation, chromosomal rearrangements or epigenetic changes. Unlike family or twin studies which make assumptions regarding sharing of environmental risk factors, the GCTA approach employed here makes no such assumption, confirming that the risk of this trait is significantly determined by common genetic variants. The absence of major gene effects in the GWAS conducted in this study indicates that the disease is likely to be largely polygenic in pathogenesis. These findings are consistent with common variant heritability reported for epilepsy overall (26%, standard deviation 5%), for focal seizure epilepsy (27%, standard deviation 5%) , or for epilepsy subtypes (4·1–107%) , though neither of these studies specifically included BECTS patients. Suggestive association of BECTS was observed with 12 SNPs with concordant association in both datasets. These associations are novel, and no association was observed with genes previously implicated in BECTS and related forms of epilepsy, including ELP4 , GRIN2A , , and RBFOX1/RBFOX3 . Because the SNP microarrays used in the current study do not detect the rare variants previously reported in these genes, this study simply shows that common variants in these genes (MAF>1%) do not have major influences on the risk of BECTS. A combination of suggestive genetic association and strong transcriptomic effect supports the involvement of the gene CHRNA5, encoding the cholinergic receptor nicotinic alpha 5 subunit, in BECTS aetiopathogenesis. While the most strongly associated SNP at this locus (rs1948 at chromosome 15q24) is located within the gene CHRNB4 , SMR analysis suggests that the associated gene at this locus is the neighbouring gene CHRNA5 . The rs1948 is strongly associated with expression of the t3603436 transcript of the CHRNA5 genes (P eqtl =2·10 × 10 −12 , P smr =7·9 × 10 −5 ) which encodes a subunit of cholinergic receptors. Acetylcholine is an important excitatory CNS neurotransmitter. Association has previously been reported between multiple SNPs in the CHRNA5-CHRNA3-CHRNB4 gene cluster and cigarette smoking, nicotine dependence, and smoking associated lung diseases , , , as well as with cognitive measures . There is suggestive evidence that smoking increases the risk of epilepsy overall , but whether smoking influences the risk of BECTS specifically is unknown. It was reported that BECTS risk allele rs1948-A is also associated with higher Fagerström Test for Nicotine Dependence (FTND) score in the Chinese Han population . Although it is unlikely the patients themselves were smoking at their age of onset, they could have been exposed to secondhand smoke. If the offspring is a carrier of the risk allele, then at least one of the parents must be a carrier too, which means they are more likely to develop tobacco addiction or become a heavier smoker. Studies have shown that prenatal maternal cigarette smoking is associated with febrile seizure . It is possible that CHRNA5 influences the onset of BECTS by another mechanism independently of smoking (including maternal or paternal perinatal, or postnatal passive, smoking). The GSMR analysis reported here though demonstrates using data from the UK Biobank that maternal smoking around birth is associated with 3·9x increased risk of BECTS. Unfortunately data is not available in UK Biobank about other forms of perinatal or antenatal smoking exposure, and therefore we are unable for example to determine the effect of paternal or other sources of passive smoking on BECTS risk. Linkage of chromosome 15 (15q24), at which CHRNA3, CHRNA5 , and CHRNB4 are encoded, has also been reported in autosomal dominant nocturnal frontal lobe epilepsy (ADNFLE) , but this finding has not been confirmed in other families , and the mutations identified to date in CHRN genes in ADNFLE have been in other receptor subunits (reviewed in ). Of interest, as with ADNFLE, in BECT seizures occur more commonly at night and during sleep, suggesting overlapping pathogenesis. Studies in rodents have described the anatomical localization and function of the nicotinic acetylcholine receptors (nAChRs) formed by the subunits encoded by this gene cluster. Animal experiments also have shown that microinjection of acetylcholine into the brain can cause seizures in animals, suggesting that acetylcholine, as a neurotransmitter, may play an important role in the development of epilepsy . While not definitive, this data supports a role for this locus and a possible link between nicotine/cholinergic neurostimulation and BECTS, raising the hypothesis that anticholinergic therapies may be effective in BECTS. Further research will be required to test this. KALRN , associated with Heschl's gyrus (temporal lobe) morphology in GWAS, induces various signaling mechanisms that regulate neuronal shape, growth, and plasticity, through their effects on the actin cytoskeleton . This locus has not previously been associated with epilepsy but its known association with brain morphology is consistent with a role in a CNS disease like this. Another eight loci achieved suggestive association, and further research will be required such as expanding the GWAS dataset, and/or functional genomic analyses, to determine if they have a role in BECTS. In conclusion, this study shows that significant common variant heritability contributes to the development of BECTS, and provides evidence that the cholinergic receptor subunit gene CHRNA5 , is involved in its pathogenesis. This raises the hypothesis that anticholinergic therapies may be effective in BECTS; further research will be required to test this. Importantly, the study demonstrates through Mendelian randomization approaches that maternal smoking around birth is associated with increased risk of BECTS, thereby identifying to our knowledge the first environmental risk factor for the disease.
Huji Xu, Li-Ping Zou, Perry Bartlett, David Reutens and Matthew A Brown made substantial contributions to conception and design of the study. Geng Wang, Zhixiu Li, Paul Leo, Gabriel Cuellar Partida, Mischa Lundberg and Matthew A Brown contributed to statistical analysis and interpretation of data. Huji Xu, Li-Ping Zou, Xiuyu Shi, Zhisheng Liu, Gefei Wu, Hongmin Zhu, Yuqin Zhang, Dong Li, Li Gao, Liu Yang, Wei Wang, Jianxiang Liao, Jiwen Wang, Shuizhen Zhou, Hua Wang, Xiaojing Li, Jingyun Gao, Li Zhang, Xiaomei Shu, Dan Li, Yan Li, Chunhong Chen, and Xiuju Zhang were responsible for case diagnosis, subject recruitment and the collection of blood samples. David Reutens are responsible for reviewing the diagnosis of the patients. Geng Wang, Matthew Brown, and Huji Xu were primarily responsible for drafting the manuscript and revising it. All authors approved the final version of the manuscript.
The authors have no conflict of interest.
|
Tandem metabolic reaction–based sensors unlock in vivo metabolomics | f8fa5452-d6dd-4a19-b230-4b9ec9f1909b | 11892595 | Biochemistry[mh] | TMR Integrates and Self-Mediates Cofactor Redox Reactions at the Enzymatic Limit. The electrode’s support for cofactor-assisted enzymatic reactions and electrochemical analysis of modified cofactor end-products is fundamental to versatile metabolite intermediation and detection. To put it in perspective, considering cofactor NAD alone: It facilitates roughly ten times more reactions than cofactor-less oxidase reactions, leading to a proportional increase in directly detectable metabolites (~800, Dataset S2 ). However, the involvement of cofactor molecules, which expands reaction versatility, also increases reaction complexity, posing two key challenges. First, the essential cofactor molecules required to drive reactions are scarce in vivo, necessitating integration into the sensor for in vivo operations . This contrasts with oxidase-based reactions, which benefit from naturally abundant oxygen . Second, the larger end-product size in cofactor-based sensing hinders the application of size-exclusion methods commonly used in oxidase sensing for interference mitigation. Coupled with high overpotential requirements, this increases the sensor’s susceptibility to interfering molecules in electrochemical analysis . The TMR electrode’s unique features overcome these challenges from multiple aspects, as we describe below. Cofactors such as NAD + and ATP (adenosine triphosphate) can be directly adsorbed onto the electrode’s SWCNT framework through π–π stacking interactions. Utilizing scanning transmission electron microscopy (S/TEM) in conjunction with energy-dispersive X-ray spectroscopy (EDS), we visualized this integration. and SI Appendix , Fig. S1 depict the morphology of the cofactor-integrated SWCNT for the cases of NAD + and ATP, respectively. The corresponding EDS images in and SI Appendix , Fig. S1 indicate a consistent and concentrated cofactor distribution across the SWCNT framework. The successful immobilization of NAD + can be further verified via cyclic voltammetry (CV), which captures the distinctive redox signatures of this electroactive cofactor . The TMR electrode can electrochemically analyze the electroactive end-product of cofactor-assisted enzymatic reactions with exceptionally high efficiency. We studied its performance in the context of NAD redox reactions, chosen as a model due to their broad applicability to dehydrogenase-based systems within the oxidoreductase library. In this context, the electrochemical oxidation of NADH theoretically involves a two-electron transfer process, represented by the electron transfer number, n e , of 2 . Nonetheless, empirically, due to NADH’s inherent instability resulting in irreversible decomposition and electrode fouling, the effective electron transfer number is reduced . For conventional electrodes such as glassy carbon (GC) and carbon paste incorporating 1,10-phenanthroline-5,6,-dione-based mediators (PD/CP), n e is close to 1 . Our TMR electrode, based on acid-treated SWCNT, demonstrates significantly higher NADH oxidation efficiency with an n e as high as 1.90 ± 0.07 (obtained via rotating disk electrode, RDE, analysis, SI Appendix , Fig. S2 ), while also exhibiting high NADH sensitivity at 0 V (vs. Ag/AgCl) oxidation potential (i.e., self-mediating reactions, ). This enhanced performance can be attributed to the large specific area of the SWCNT substrate ( and SI Appendix , Fig. S3 ), the presence of quinone-based groups on the acid-treated SWCNT, and the catalytic property of the SWCNT’s edge plane . The high oxidation efficiency of the TMR electrode also contributes to its superior antifouling performance. The results of our antifouling study (1-h oxidation at 150 μM NADH) illustrate that the degradation in the TMR electrode’s NADH response is less than one percent, which is over 10-fold smaller than that observed with alternative electrodes all operated at their intended voltages for NADH oxidation ( , Inset ). To further assess its long-term stability, the TMR sensor was subjected to continuous NADH oxidation for 10 h—10 times the duration of the initial test—where it still exhibited minimal signal degradation (1.2%, SI Appendix , Fig. S4 ). The TMR’s self-mediation not only enhances the signal generated by electroactive end-product redox reactions but also minimizes noise from electroactive interference, thus addressing a fundamental challenge in enzymatic sensing. To illustrate this benefit, we characterized the TMR’s sensitivity to NADH and a panel of electroactive molecules commonly found in biofluids . For comparison, we conducted the same procedure using GC and PD/CP electrodes and defined the ratio of the electrodes’ reaction sensitivity to NADH vs. interfering analytes as a measure of SNR . This definition of SNR offers comprehensive coverage across a spectrum of target and interference concentrations, ensuring applicability to diverse sensing scenarios. It is distinct from traditional single-point interference characterization methods, which often use large concentration differences between the target and interfering molecules . demonstrates that for all interference cases, the TMR exhibited significantly higher SNR compared to the alternatives ( SI Appendix , Fig. S5 ). The TMR architecture also enhances both the mass transport of the enzymatic reaction product and the associated electrochemical reaction kinetics, ultimately achieving optimal overall reaction rates. In TMR, cofactor molecules (NAD + ) are directly immobilized onto the porous electrode substrate with a large specific area ( A 0,TMR ~ 10 8 /m). This design reduces the diffusion distance of the enzyme product (NADH) to the substrate electrode, down to a few nanometers. In contrast, conventional cofactor-based enzymatic sensors superficially immobilize cofactors atop the electrode substrate, such as GC and PD/CP, within the enzyme layer of approximately a few micrometers thickness . Consequently, their diffusion length scales are on the order of micrometer scale. Furthermore, the TMR drives NADH oxidation at a substantially higher intrinsic rate than conventional electrodes, as evidenced by its large exchange current density ( i 0,TMR ~ 17.4 ± 0.3 A/m 2 , SI Appendix , Fig. S2 , as compared to i 0,GC ~ 0.06 A/m 2 and i 0,PD/CP ~ 0.35 A/m 2 ) . This together with TMR’s extremely large specific area ( and SI Appendix , Fig. S3 ), allows TMR to drive overall catalytic reactions (within a fixed footprint) at a dramatically higher capacity than alternative electrodes. To study the TMR’s enhanced enzymatic/electrochemical reaction kinetics, we utilized a finite element analysis–based simulation model. Within this model, key parameters such as cofactor arrangement, enzyme activity, and electrode properties/design (e.g., i 0 , A 0 ), can be explored. We first validated the fidelity of the model by verifying its alignment with empirical measurements ( SI Appendix , Fig. S6 ). Then, to study the effect of mass transport limitation, we simulated the transduced NADH oxidation current resulting from different NADH diffusion distances across three types of electrodes: SWCNT (TMR’s framework), GC, and PD/CP (operated at their intended oxidation voltages) . As shown in , the results indicate that TMR transduces over 100-fold larger current. Next, we studied the reaction kinetics of the electrodes in terms of their specific area and exchange current density. To conduct this study in isolation from mass transport limitations, we reconfigured the model to simulate NADH oxidation (generated by a model dehydrogenase reaction) directly taking place at the electrodes’ surfaces. The corresponding simulation results for a range of hypothetical exchange current densities and specific areas are shown in . They specifically indicate that not only does the TMR substantially outperform alternative GC and PD/CP electrodes, but it also facilitates catalytic reaction at a rate only limited by the enzyme activity (as evident from plateaued current response). Reaching this limit can be achieved across a wide range of enzyme activities, as shown in . TMR Integrates Multifunctional Enzymatic Reactions for Versatile Metabolite Monitoring. Besides its distinguishing support for cofactor-assisted enzymatic/electrochemical reactions, the TMR platform can facilitate multiple enzymatic reactions in tandem, further setting it apart from conventional in vivo enzymatic sensors designed for single reactions. We harnessed these special properties for direct/intermediated metabolite detection and supplemental interference inactivation (further enhancing SNR in our system). To demonstrate direct metabolite detection, we employed dehydrogenase enzymes along with the cofactor NAD + to catalyze metabolic reactions, resulting in NADH as the electroactive end-product . In this setting, metabolite substrates, serving as targets, are quantified through the electrochemical analysis of NADH (self-mediated by SWCNT). We introduced ten different dehydrogenases into the TMR design, each targeting a distinct metabolite: β -hydroxybutyrate (BHB), D -glucose, L -glutamate, glucose 6-phosphate (G6P), ethanol, D -lactate, L -lactate, L -leucine, cholesterol, and glycerol. Calibration plots for each enzymatic TMR were obtained through amperometric measurements conducted within the physiological concentration ranges of the analytes (as detailed in SI Appendix , Table S1 ). shows that all TMR sensors exhibited consistent, monotonic responses to target metabolite concentrations with reproducible sensitivities and minimal interdevice variations. To target metabolites lacking specific oxidoreductases, we utilized existing metabolic pathways and harnessed the TMR’s versatility to build an oxidoreductase linkage. We configured the TMR to facilitate cascaded enzymatic reactions, intermediating the target metabolite into a form catalyzable by a corresponding oxidoreductase enzyme for electrochemical detection . We demonstrated both cofactor-less and cofactor-assisted enzymatic intermediation by integrating appropriate enzymes and, when necessary, cofactors within the TMR architecture. We leveraged the dehydrogenase-based electrochemical sensing interfaces developed for direct detection as the TMR’s base to facilitate the final stage of the cascade. This integrated design implements the entire intermediation and detection stages within a single sensor construct. With cofactor-less intermediation, we demonstrated the sensing of lactose and glucose 1-phosphate (G1P), which lack corresponding oxidoreductases. By integrating β -galactosidase ( β -GAL) and phosphoglucomutase (PGM) enzymes within the original D -glucose-TMR and G6P-TMR sensing interfaces, we transformed lactose and G1P into D -glucose and G6P for subsequent electrochemical detection, respectively . To illustrate the extensibility of our approach to cofactor-assisted intermediation, we drew inspiration from the glycolysis pathway and demonstrated intermediated glucose detection as proof of concept . Using hexokinase and ATP cofactors integrated within the TMR framework, D -glucose is converted into G6P, whose concentration is measurable via the NAD + /G6P-dehydrogenase base . As shown in , all TMR sensors implementing cascaded reactions consistently displayed monotonic responses to varying concentrations of target metabolites, exhibiting reproducible sensitivities and minimal variations. In the case of severe fluctuations in intermediary metabolite levels in vivo affecting the sensor response, the confounding effect can be mitigated by calibrating the primary TMR sensor response against a secondary TMR detecting the intermediary metabolites. Since the primary TMR incorporates the design of the secondary TMR as its subunit, accurate concentration estimation is ensured ( SI Appendix , Fig. S7 ). We specifically employed enzymatic inactivation to neutralize interference from ascorbic acid (AA) by integrating the corresponding enzyme layer (AA oxidase, AAOx). In most enzymatic sensing scenarios, including cofactor-based ones, AA serves as the most dominant noise source, distorting sensor responses through unwanted reactions with the electrodes’ substrates . The TMR’s exceptionally large specific area makes it suitable for immobilizing AAOx with high loading to effectively counter this challenge. We followed the aforementioned procedure for characterizing the SNR to study and benchmark the performance of the AAOx-coupled TMR in minimizing interference from AA. shows that our enzymatic inactivation strategy was extremely effective, as evidenced by the AAOx-coupled TMR’s more than 100-fold larger SNR compared to other traditionally used electrodes ( SI Appendix , Fig. S8 ). To further ensure the TMR’s selectivity, we recorded the representative BHB, D -glucose, and L -glutamate-TMRs’ responses to a panel of progressively introduced molecules, including small molecules, ionic species, and electroactive species at their physiologically relevant concentrations, with AA (dominant interference) tested at a high concentration (100 μM, compared to 50 μM, high end of salivary AA concentration) . As shown in and SI Appendix , Fig. S9 , the TMRs exhibited negligible response against the interference group or the enzymatic inactivation product (here, hydrogen peroxide generated by the AAOx-catalyzed reaction) due to its self-mediating capability for driving cofactor oxidation at 0 V. The latter findings particularly illustrate there is no reaction crosstalk between the two enzymatic layers (i.e. interference inactivation and detection enzymes). To demonstrate multiplexed metabolite monitoring, we fabricated an array of 7 TMRs onto a soft substrate (styrene–ethylene–butylene–styrene block copolymer, SEBS), sharing a single reference electrode. Each TMR targeted a specific metabolite: BHB, D -glucose, L -glutamate, G6P, ethanol, D -lactate, and L -leucine. We tested this array’s response in serum by concurrently recording the amperometric measurements of all 7 channels and intermittently introducing individual analyte targets. As shown in , the sensors stably responded to their corresponding analytes with no detectable crosstalk. The incorporation of an encapsulation layer (here, polyvinyl chloride, PVC), within the TMR design, combined with the robust immobilization of cofactor molecules on the TMR’s framework, ensures the stability and reversibility of the enzymatic TMR’s response. The reversibility of TMR was assessed by repeatedly immersing representative BHB, D -glucose, and L -glutamate-TMRs in solutions with increasing or decreasing target concentrations and continuously recording their responses at each concentration level. In all cases, the TMRs consistently adjusted to the expected response levels, with changes of less than 7.5% for each introduced concentration ( and SI Appendix , Fig. S10 ). The TMRs’ antifouling capability was assessed through continuous measurements in a protein-rich environment (phosphate-buffered saline, PBS, buffer with 20 mg/mL bovine serum albumin) . During 1,000-min studies involving varying target concentrations, the enzymatic TMR responses’ declines were within 2% at each level ( SI Appendix , Fig. S11 ), demonstrating the TMR’s ability to facilitate small molecule (i.e., metabolite) diffusion to the sensing substrate while effectively blocking larger protein molecules (fouling agents). We also conducted a prolonged characterization study, continuously recording the TMR’s response in a PBS buffer. As shown in , the TMR exhibited minimal response deviation, remaining within a few percentages, even after 3 d of continuous operation, indicating negligible leakage of sensing molecules. TMR Tracks Metabolite Dynamics In Vivo for Metabolic Disorders and Gut–Brain Axis. Collectively, the ex vivo characterization results support the high level of adaptability, sensitivity, selectivity, stability, and reversibility of the TMR architecture for in vivo biomonitoring. After validating the TMR’s biocompatibility through cellular viability studies ( SI Appendix , Fig. S12 ), we adapted and deployed the TMR sensors for two metabolic acidosis scenarios: ketoacidosis and D -lactate acidosis. For both scenarios, we initially established the significance of the target metabolites in relevant biomatrices and evaluated the accuracy of the TMR sensors for their analysis. Then, we applied the TMR sensors for real-time and continuous in vivo monitoring, demonstrating their potential for tracking the metabolite dynamics underlying metabolic states and the gut–brain axis. Ketoacidosis, a serious metabolic disorder often associated with diabetes, arises when ketone bodies like BHB accumulate in the bloodstream, causing an acid–base imbalance . Here, we first investigated the utility of BHB sensing in sweat and saliva for noninvasive wearable and mobile health monitoring. This approach is beneficial for individuals with conditions such as diabetes or those on ketogenic diets for epilepsy . However, from a sensing perspective, it is challenging due to over 10-fold secretion-induced dilution of BHB and high background noise from fluctuating interfering molecules such as AA (often influenced by diet), which current enzymatic electrochemical sensors fail to address . Our TMR sensor, with its intrinsically high SNR measurements and low limit of detection ( SI Appendix , Fig. S13 ), can effectively overcome these challenges. We sampled saliva and sweat from two cohorts: epileptic patients on a ketogenic diet and healthy subjects who consumed a ketone supplement. Our results showed strong correlations between sweat and saliva BHB levels vs. blood ( r = 0.84 for saliva–blood, and r = 0.92 for sweat–blood, ), also validating our BHB-TMR’s accuracy in analyzing sweat and saliva (mean bias −2 μM with 95% CI within ±45 μM, ). We confirmed the TMR’s compatibility with low-power consumer electronics for wireless operation ( SI Appendix , Fig. S14 ) and used it to track changes in metabolic states. SI Appendix , Fig. S15 depicts the TMR detecting elevated salivary BHB concentration in a fasting healthy subject, indicating ketosis. Following consumption of a carbohydrate-rich beverage, salivary BHB levels rapidly dropped from ~250 to ~100 μM within an hour, suggesting a shift from ketosis to glycolysis (corroborated by capillary blood glucose analysis) . To validate TMR’s in vivo monitoring capability relevant to diabetic ketoacidosis, we tracked blood BHB and glucose levels in mice. BHB- and glucose-TMRs were integrated into an array, alongside a TMR lacking detection enzymes serving as a negative control. As shown in and SI Appendix , Fig. S16 , following intravenous administration of each metabolite, both BHB and glucose TMRs (affixed on the mouse back for subdermal blood analysis) promptly captured the dynamic changes of the corresponding analytes, while the negative control maintained its baseline response. The results highlight the TMR’s ability for in vivo metabolic data acquisition with minute-level resolution, surpassing the sampling rates of traditional methods by two to three orders of magnitude, especially advantageous in small animals with limited sampling volume thresholds . For D -lactate acidosis, we focused on detecting the bacterial metabolite D -lactate in blood and brain for its applications in disease diagnosis and treatment (e.g., short bowel syndrome and encephalopathy) and for advancing understanding of microbiome–gut–brain axis dynamics. D -lactate is a byproduct of gut bacterial carbohydrate fermentation . Several interacting factors, including dysbiosis of the gut microbiota, malabsorption of intestinal nutrients, and decreased intestinal barrier integrity can enhance flux of D -lactate from the intestinal lumen into systemic circulation . D -lactate typically crosses the blood–brain barrier via monocarboxylate transporters, but can exhibit increased translocation in pathological conditions. Within the brain, D -lactate accumulates at least in part due to its slower metabolism compared to endogenous L -lactate, leading to neurotoxicity and various neurological complications, including confusion, disorientation, and seizures . To study the effect of bacterial fermentation on D -lactate levels in blood vs. brain, we monocolonized mice with Bacteroides thetaiotaomicron , a prominent member of the human gut microbiome that plays a crucial role in digesting complex carbohydrates. Mice were then fed a custom diet containing the host nondigestible carbohydrate levan as the sole carbohydrate source, which B. thetaiotaomicron selectively ferments . SI Appendix , Fig. S17 shows that, compared to germ-free controls, colonization with B. thetaiotaomicron modestly increased serum D -lactate levels without elevating brain D -lactate levels. This suggests that under nonpathological conditions, the gut microbiome promotes D -lactate in the serum without affecting brain levels. We next modeled high carbohydrate feeding and intestinal barrier dysfunction as key risk factors for D -lactate acidosis and comorbid encephalopathy. To do so, we first treated conventionally colonized mice with dextran sodium sulfate (DSS), a common model of experimental colitis, to induce intestinal barrier permeability . Following 7 d of DSS treatment, mice were fasted and orally gavaged with a mixture of host nondigestible carbohydrates (fructooligosaccharide, inulin, cellulose, and gum arabic) to promote rapid bacterial fermentation and D -lactate production. demonstrates that DSS-treated mice exhibited significantly elevated D -lactate levels in the brain compared to vehicle-treated controls, indicating that intestinal injury leads to increased entry of bacterial-derived D -lactate into the brain. There was no significant correlation between serum and brain D -lactate levels within individual animals ( SI Appendix , Fig. S18 ), emphasizing the necessity for methods that can concurrently measure bacterial metabolites in circulation and local environments like the brain to better understand the microbiota–host interactions. The TMR proves to be a fitting solution, given its high accuracy in analyzing D -lactate in both blood and brain matrices, as demonstrated in (mean bias −3 μM with 95% CI within ± 36 μM). We deployed TMR sensors for in vivo monitoring of local and circulating D -lactate in a rat model. We affixed D -lactate-TMRs to the brain and back for subdural and subdermal analysis, respectively. An accompanying L -lactate-TMR sensor analyzing blood served as a negative control. We continuously recorded the TMR sensor responses before and after intravenous D -lactate injection. The control device exhibited minor transient disturbances postinjection ( SI Appendix , Fig. S19 ), attributable to the momentarily increased osmotic load from the lactate buffer injection, consistent with prior reports . As shown in , the D -lactate sensors tracked acute increases in D -lactate following the injection, revealing a slower rate of concentration increase in the brain compared to the blood, suggesting limited transport rates of D -lactate into the brain .
The electrode’s support for cofactor-assisted enzymatic reactions and electrochemical analysis of modified cofactor end-products is fundamental to versatile metabolite intermediation and detection. To put it in perspective, considering cofactor NAD alone: It facilitates roughly ten times more reactions than cofactor-less oxidase reactions, leading to a proportional increase in directly detectable metabolites (~800, Dataset S2 ). However, the involvement of cofactor molecules, which expands reaction versatility, also increases reaction complexity, posing two key challenges. First, the essential cofactor molecules required to drive reactions are scarce in vivo, necessitating integration into the sensor for in vivo operations . This contrasts with oxidase-based reactions, which benefit from naturally abundant oxygen . Second, the larger end-product size in cofactor-based sensing hinders the application of size-exclusion methods commonly used in oxidase sensing for interference mitigation. Coupled with high overpotential requirements, this increases the sensor’s susceptibility to interfering molecules in electrochemical analysis . The TMR electrode’s unique features overcome these challenges from multiple aspects, as we describe below. Cofactors such as NAD + and ATP (adenosine triphosphate) can be directly adsorbed onto the electrode’s SWCNT framework through π–π stacking interactions. Utilizing scanning transmission electron microscopy (S/TEM) in conjunction with energy-dispersive X-ray spectroscopy (EDS), we visualized this integration. and SI Appendix , Fig. S1 depict the morphology of the cofactor-integrated SWCNT for the cases of NAD + and ATP, respectively. The corresponding EDS images in and SI Appendix , Fig. S1 indicate a consistent and concentrated cofactor distribution across the SWCNT framework. The successful immobilization of NAD + can be further verified via cyclic voltammetry (CV), which captures the distinctive redox signatures of this electroactive cofactor . The TMR electrode can electrochemically analyze the electroactive end-product of cofactor-assisted enzymatic reactions with exceptionally high efficiency. We studied its performance in the context of NAD redox reactions, chosen as a model due to their broad applicability to dehydrogenase-based systems within the oxidoreductase library. In this context, the electrochemical oxidation of NADH theoretically involves a two-electron transfer process, represented by the electron transfer number, n e , of 2 . Nonetheless, empirically, due to NADH’s inherent instability resulting in irreversible decomposition and electrode fouling, the effective electron transfer number is reduced . For conventional electrodes such as glassy carbon (GC) and carbon paste incorporating 1,10-phenanthroline-5,6,-dione-based mediators (PD/CP), n e is close to 1 . Our TMR electrode, based on acid-treated SWCNT, demonstrates significantly higher NADH oxidation efficiency with an n e as high as 1.90 ± 0.07 (obtained via rotating disk electrode, RDE, analysis, SI Appendix , Fig. S2 ), while also exhibiting high NADH sensitivity at 0 V (vs. Ag/AgCl) oxidation potential (i.e., self-mediating reactions, ). This enhanced performance can be attributed to the large specific area of the SWCNT substrate ( and SI Appendix , Fig. S3 ), the presence of quinone-based groups on the acid-treated SWCNT, and the catalytic property of the SWCNT’s edge plane . The high oxidation efficiency of the TMR electrode also contributes to its superior antifouling performance. The results of our antifouling study (1-h oxidation at 150 μM NADH) illustrate that the degradation in the TMR electrode’s NADH response is less than one percent, which is over 10-fold smaller than that observed with alternative electrodes all operated at their intended voltages for NADH oxidation ( , Inset ). To further assess its long-term stability, the TMR sensor was subjected to continuous NADH oxidation for 10 h—10 times the duration of the initial test—where it still exhibited minimal signal degradation (1.2%, SI Appendix , Fig. S4 ). The TMR’s self-mediation not only enhances the signal generated by electroactive end-product redox reactions but also minimizes noise from electroactive interference, thus addressing a fundamental challenge in enzymatic sensing. To illustrate this benefit, we characterized the TMR’s sensitivity to NADH and a panel of electroactive molecules commonly found in biofluids . For comparison, we conducted the same procedure using GC and PD/CP electrodes and defined the ratio of the electrodes’ reaction sensitivity to NADH vs. interfering analytes as a measure of SNR . This definition of SNR offers comprehensive coverage across a spectrum of target and interference concentrations, ensuring applicability to diverse sensing scenarios. It is distinct from traditional single-point interference characterization methods, which often use large concentration differences between the target and interfering molecules . demonstrates that for all interference cases, the TMR exhibited significantly higher SNR compared to the alternatives ( SI Appendix , Fig. S5 ). The TMR architecture also enhances both the mass transport of the enzymatic reaction product and the associated electrochemical reaction kinetics, ultimately achieving optimal overall reaction rates. In TMR, cofactor molecules (NAD + ) are directly immobilized onto the porous electrode substrate with a large specific area ( A 0,TMR ~ 10 8 /m). This design reduces the diffusion distance of the enzyme product (NADH) to the substrate electrode, down to a few nanometers. In contrast, conventional cofactor-based enzymatic sensors superficially immobilize cofactors atop the electrode substrate, such as GC and PD/CP, within the enzyme layer of approximately a few micrometers thickness . Consequently, their diffusion length scales are on the order of micrometer scale. Furthermore, the TMR drives NADH oxidation at a substantially higher intrinsic rate than conventional electrodes, as evidenced by its large exchange current density ( i 0,TMR ~ 17.4 ± 0.3 A/m 2 , SI Appendix , Fig. S2 , as compared to i 0,GC ~ 0.06 A/m 2 and i 0,PD/CP ~ 0.35 A/m 2 ) . This together with TMR’s extremely large specific area ( and SI Appendix , Fig. S3 ), allows TMR to drive overall catalytic reactions (within a fixed footprint) at a dramatically higher capacity than alternative electrodes. To study the TMR’s enhanced enzymatic/electrochemical reaction kinetics, we utilized a finite element analysis–based simulation model. Within this model, key parameters such as cofactor arrangement, enzyme activity, and electrode properties/design (e.g., i 0 , A 0 ), can be explored. We first validated the fidelity of the model by verifying its alignment with empirical measurements ( SI Appendix , Fig. S6 ). Then, to study the effect of mass transport limitation, we simulated the transduced NADH oxidation current resulting from different NADH diffusion distances across three types of electrodes: SWCNT (TMR’s framework), GC, and PD/CP (operated at their intended oxidation voltages) . As shown in , the results indicate that TMR transduces over 100-fold larger current. Next, we studied the reaction kinetics of the electrodes in terms of their specific area and exchange current density. To conduct this study in isolation from mass transport limitations, we reconfigured the model to simulate NADH oxidation (generated by a model dehydrogenase reaction) directly taking place at the electrodes’ surfaces. The corresponding simulation results for a range of hypothetical exchange current densities and specific areas are shown in . They specifically indicate that not only does the TMR substantially outperform alternative GC and PD/CP electrodes, but it also facilitates catalytic reaction at a rate only limited by the enzyme activity (as evident from plateaued current response). Reaching this limit can be achieved across a wide range of enzyme activities, as shown in .
Besides its distinguishing support for cofactor-assisted enzymatic/electrochemical reactions, the TMR platform can facilitate multiple enzymatic reactions in tandem, further setting it apart from conventional in vivo enzymatic sensors designed for single reactions. We harnessed these special properties for direct/intermediated metabolite detection and supplemental interference inactivation (further enhancing SNR in our system). To demonstrate direct metabolite detection, we employed dehydrogenase enzymes along with the cofactor NAD + to catalyze metabolic reactions, resulting in NADH as the electroactive end-product . In this setting, metabolite substrates, serving as targets, are quantified through the electrochemical analysis of NADH (self-mediated by SWCNT). We introduced ten different dehydrogenases into the TMR design, each targeting a distinct metabolite: β -hydroxybutyrate (BHB), D -glucose, L -glutamate, glucose 6-phosphate (G6P), ethanol, D -lactate, L -lactate, L -leucine, cholesterol, and glycerol. Calibration plots for each enzymatic TMR were obtained through amperometric measurements conducted within the physiological concentration ranges of the analytes (as detailed in SI Appendix , Table S1 ). shows that all TMR sensors exhibited consistent, monotonic responses to target metabolite concentrations with reproducible sensitivities and minimal interdevice variations. To target metabolites lacking specific oxidoreductases, we utilized existing metabolic pathways and harnessed the TMR’s versatility to build an oxidoreductase linkage. We configured the TMR to facilitate cascaded enzymatic reactions, intermediating the target metabolite into a form catalyzable by a corresponding oxidoreductase enzyme for electrochemical detection . We demonstrated both cofactor-less and cofactor-assisted enzymatic intermediation by integrating appropriate enzymes and, when necessary, cofactors within the TMR architecture. We leveraged the dehydrogenase-based electrochemical sensing interfaces developed for direct detection as the TMR’s base to facilitate the final stage of the cascade. This integrated design implements the entire intermediation and detection stages within a single sensor construct. With cofactor-less intermediation, we demonstrated the sensing of lactose and glucose 1-phosphate (G1P), which lack corresponding oxidoreductases. By integrating β -galactosidase ( β -GAL) and phosphoglucomutase (PGM) enzymes within the original D -glucose-TMR and G6P-TMR sensing interfaces, we transformed lactose and G1P into D -glucose and G6P for subsequent electrochemical detection, respectively . To illustrate the extensibility of our approach to cofactor-assisted intermediation, we drew inspiration from the glycolysis pathway and demonstrated intermediated glucose detection as proof of concept . Using hexokinase and ATP cofactors integrated within the TMR framework, D -glucose is converted into G6P, whose concentration is measurable via the NAD + /G6P-dehydrogenase base . As shown in , all TMR sensors implementing cascaded reactions consistently displayed monotonic responses to varying concentrations of target metabolites, exhibiting reproducible sensitivities and minimal variations. In the case of severe fluctuations in intermediary metabolite levels in vivo affecting the sensor response, the confounding effect can be mitigated by calibrating the primary TMR sensor response against a secondary TMR detecting the intermediary metabolites. Since the primary TMR incorporates the design of the secondary TMR as its subunit, accurate concentration estimation is ensured ( SI Appendix , Fig. S7 ). We specifically employed enzymatic inactivation to neutralize interference from ascorbic acid (AA) by integrating the corresponding enzyme layer (AA oxidase, AAOx). In most enzymatic sensing scenarios, including cofactor-based ones, AA serves as the most dominant noise source, distorting sensor responses through unwanted reactions with the electrodes’ substrates . The TMR’s exceptionally large specific area makes it suitable for immobilizing AAOx with high loading to effectively counter this challenge. We followed the aforementioned procedure for characterizing the SNR to study and benchmark the performance of the AAOx-coupled TMR in minimizing interference from AA. shows that our enzymatic inactivation strategy was extremely effective, as evidenced by the AAOx-coupled TMR’s more than 100-fold larger SNR compared to other traditionally used electrodes ( SI Appendix , Fig. S8 ). To further ensure the TMR’s selectivity, we recorded the representative BHB, D -glucose, and L -glutamate-TMRs’ responses to a panel of progressively introduced molecules, including small molecules, ionic species, and electroactive species at their physiologically relevant concentrations, with AA (dominant interference) tested at a high concentration (100 μM, compared to 50 μM, high end of salivary AA concentration) . As shown in and SI Appendix , Fig. S9 , the TMRs exhibited negligible response against the interference group or the enzymatic inactivation product (here, hydrogen peroxide generated by the AAOx-catalyzed reaction) due to its self-mediating capability for driving cofactor oxidation at 0 V. The latter findings particularly illustrate there is no reaction crosstalk between the two enzymatic layers (i.e. interference inactivation and detection enzymes). To demonstrate multiplexed metabolite monitoring, we fabricated an array of 7 TMRs onto a soft substrate (styrene–ethylene–butylene–styrene block copolymer, SEBS), sharing a single reference electrode. Each TMR targeted a specific metabolite: BHB, D -glucose, L -glutamate, G6P, ethanol, D -lactate, and L -leucine. We tested this array’s response in serum by concurrently recording the amperometric measurements of all 7 channels and intermittently introducing individual analyte targets. As shown in , the sensors stably responded to their corresponding analytes with no detectable crosstalk. The incorporation of an encapsulation layer (here, polyvinyl chloride, PVC), within the TMR design, combined with the robust immobilization of cofactor molecules on the TMR’s framework, ensures the stability and reversibility of the enzymatic TMR’s response. The reversibility of TMR was assessed by repeatedly immersing representative BHB, D -glucose, and L -glutamate-TMRs in solutions with increasing or decreasing target concentrations and continuously recording their responses at each concentration level. In all cases, the TMRs consistently adjusted to the expected response levels, with changes of less than 7.5% for each introduced concentration ( and SI Appendix , Fig. S10 ). The TMRs’ antifouling capability was assessed through continuous measurements in a protein-rich environment (phosphate-buffered saline, PBS, buffer with 20 mg/mL bovine serum albumin) . During 1,000-min studies involving varying target concentrations, the enzymatic TMR responses’ declines were within 2% at each level ( SI Appendix , Fig. S11 ), demonstrating the TMR’s ability to facilitate small molecule (i.e., metabolite) diffusion to the sensing substrate while effectively blocking larger protein molecules (fouling agents). We also conducted a prolonged characterization study, continuously recording the TMR’s response in a PBS buffer. As shown in , the TMR exhibited minimal response deviation, remaining within a few percentages, even after 3 d of continuous operation, indicating negligible leakage of sensing molecules.
Collectively, the ex vivo characterization results support the high level of adaptability, sensitivity, selectivity, stability, and reversibility of the TMR architecture for in vivo biomonitoring. After validating the TMR’s biocompatibility through cellular viability studies ( SI Appendix , Fig. S12 ), we adapted and deployed the TMR sensors for two metabolic acidosis scenarios: ketoacidosis and D -lactate acidosis. For both scenarios, we initially established the significance of the target metabolites in relevant biomatrices and evaluated the accuracy of the TMR sensors for their analysis. Then, we applied the TMR sensors for real-time and continuous in vivo monitoring, demonstrating their potential for tracking the metabolite dynamics underlying metabolic states and the gut–brain axis. Ketoacidosis, a serious metabolic disorder often associated with diabetes, arises when ketone bodies like BHB accumulate in the bloodstream, causing an acid–base imbalance . Here, we first investigated the utility of BHB sensing in sweat and saliva for noninvasive wearable and mobile health monitoring. This approach is beneficial for individuals with conditions such as diabetes or those on ketogenic diets for epilepsy . However, from a sensing perspective, it is challenging due to over 10-fold secretion-induced dilution of BHB and high background noise from fluctuating interfering molecules such as AA (often influenced by diet), which current enzymatic electrochemical sensors fail to address . Our TMR sensor, with its intrinsically high SNR measurements and low limit of detection ( SI Appendix , Fig. S13 ), can effectively overcome these challenges. We sampled saliva and sweat from two cohorts: epileptic patients on a ketogenic diet and healthy subjects who consumed a ketone supplement. Our results showed strong correlations between sweat and saliva BHB levels vs. blood ( r = 0.84 for saliva–blood, and r = 0.92 for sweat–blood, ), also validating our BHB-TMR’s accuracy in analyzing sweat and saliva (mean bias −2 μM with 95% CI within ±45 μM, ). We confirmed the TMR’s compatibility with low-power consumer electronics for wireless operation ( SI Appendix , Fig. S14 ) and used it to track changes in metabolic states. SI Appendix , Fig. S15 depicts the TMR detecting elevated salivary BHB concentration in a fasting healthy subject, indicating ketosis. Following consumption of a carbohydrate-rich beverage, salivary BHB levels rapidly dropped from ~250 to ~100 μM within an hour, suggesting a shift from ketosis to glycolysis (corroborated by capillary blood glucose analysis) . To validate TMR’s in vivo monitoring capability relevant to diabetic ketoacidosis, we tracked blood BHB and glucose levels in mice. BHB- and glucose-TMRs were integrated into an array, alongside a TMR lacking detection enzymes serving as a negative control. As shown in and SI Appendix , Fig. S16 , following intravenous administration of each metabolite, both BHB and glucose TMRs (affixed on the mouse back for subdermal blood analysis) promptly captured the dynamic changes of the corresponding analytes, while the negative control maintained its baseline response. The results highlight the TMR’s ability for in vivo metabolic data acquisition with minute-level resolution, surpassing the sampling rates of traditional methods by two to three orders of magnitude, especially advantageous in small animals with limited sampling volume thresholds . For D -lactate acidosis, we focused on detecting the bacterial metabolite D -lactate in blood and brain for its applications in disease diagnosis and treatment (e.g., short bowel syndrome and encephalopathy) and for advancing understanding of microbiome–gut–brain axis dynamics. D -lactate is a byproduct of gut bacterial carbohydrate fermentation . Several interacting factors, including dysbiosis of the gut microbiota, malabsorption of intestinal nutrients, and decreased intestinal barrier integrity can enhance flux of D -lactate from the intestinal lumen into systemic circulation . D -lactate typically crosses the blood–brain barrier via monocarboxylate transporters, but can exhibit increased translocation in pathological conditions. Within the brain, D -lactate accumulates at least in part due to its slower metabolism compared to endogenous L -lactate, leading to neurotoxicity and various neurological complications, including confusion, disorientation, and seizures . To study the effect of bacterial fermentation on D -lactate levels in blood vs. brain, we monocolonized mice with Bacteroides thetaiotaomicron , a prominent member of the human gut microbiome that plays a crucial role in digesting complex carbohydrates. Mice were then fed a custom diet containing the host nondigestible carbohydrate levan as the sole carbohydrate source, which B. thetaiotaomicron selectively ferments . SI Appendix , Fig. S17 shows that, compared to germ-free controls, colonization with B. thetaiotaomicron modestly increased serum D -lactate levels without elevating brain D -lactate levels. This suggests that under nonpathological conditions, the gut microbiome promotes D -lactate in the serum without affecting brain levels. We next modeled high carbohydrate feeding and intestinal barrier dysfunction as key risk factors for D -lactate acidosis and comorbid encephalopathy. To do so, we first treated conventionally colonized mice with dextran sodium sulfate (DSS), a common model of experimental colitis, to induce intestinal barrier permeability . Following 7 d of DSS treatment, mice were fasted and orally gavaged with a mixture of host nondigestible carbohydrates (fructooligosaccharide, inulin, cellulose, and gum arabic) to promote rapid bacterial fermentation and D -lactate production. demonstrates that DSS-treated mice exhibited significantly elevated D -lactate levels in the brain compared to vehicle-treated controls, indicating that intestinal injury leads to increased entry of bacterial-derived D -lactate into the brain. There was no significant correlation between serum and brain D -lactate levels within individual animals ( SI Appendix , Fig. S18 ), emphasizing the necessity for methods that can concurrently measure bacterial metabolites in circulation and local environments like the brain to better understand the microbiota–host interactions. The TMR proves to be a fitting solution, given its high accuracy in analyzing D -lactate in both blood and brain matrices, as demonstrated in (mean bias −3 μM with 95% CI within ± 36 μM). We deployed TMR sensors for in vivo monitoring of local and circulating D -lactate in a rat model. We affixed D -lactate-TMRs to the brain and back for subdural and subdermal analysis, respectively. An accompanying L -lactate-TMR sensor analyzing blood served as a negative control. We continuously recorded the TMR sensor responses before and after intravenous D -lactate injection. The control device exhibited minor transient disturbances postinjection ( SI Appendix , Fig. S19 ), attributable to the momentarily increased osmotic load from the lactate buffer injection, consistent with prior reports . As shown in , the D -lactate sensors tracked acute increases in D -lactate following the injection, revealing a slower rate of concentration increase in the brain compared to the blood, suggesting limited transport rates of D -lactate into the brain .
Our strategy harnesses naturally proven metabolic pathways that are linkable to oxidoreductase-based electrochemical analysis as a blueprint for bioelectronic design. Implemented through the TMR electrode with exceptional electrochemical properties, this design makes multifunctional use of evolutionarily robust molecular toolkits (enzymes and cofactors) to support underlying reactions. This approach enables reliable monitoring of a plethora of metabolites, with NAD-assisted enzymatic sensing alone capable of directly detecting over 800 metabolites. To support broader in vivo applications, TMR sensors could benefit from enhancing their antifouling properties (e.g., exploring the use of surfactants) , further miniaturization, and integration with soft or microneedle bioelectronic substrates or lateral flow devices for analyzing various biomatrices in diverse clinical settings. Additionally, the TMR’s solution-based fabrication is compatible with industrial manufacturing processes, enabling flexible and streamlined large-scale production. TMR’s versatility in monitoring endogenous and bacterial metabolites in vivo across various biomatrices makes it a powerful metabolomics tool for propelling biomedical research and healthcare. In microbiome research, it can help decipher the temporal dynamics of microbiota–host metabolic communication, recognized as one of the “greatest challenges” in the field . TMR’s adaptation into wearable and implantable formats can advance sparse metabolite-based point-of-care testing to continuous point-of-person monitoring for chronic disease prevention/management, fitness optimization, and infectious disease detection . Moreover, TMR’s focus on metabolic pathways aligns seamlessly with tracking bacterial and tumor metabolism within their microenvironments . This capability can empower the design and monitoring of the efficacy of antibiotics and chemotherapeutics targeting key metabolic processes within pathogens or cancer cells to minimize drug resistance . Thus, future efforts should also include large-scale clinical trials to validate TMR’s clinical utility in these applications, while ensuring adherence to regulatory standards for safe and effective implementation. Ultimately, scaling and deployment of TMR in these contexts will generate massive, multidimensional, and real-time metabolic datasets with high temporal resolutions, facilitating deeper understanding and interaction with biology and advancing personalized medicine.
A detailed description of the materials and methods used in this study can be found in SI Appendix , including information on the fabrication and characterization of our TMR-based sensors, the electrochemical reaction simulation models, biocompatibility tests, biological sample collection and quantification, the design and operation of the wireless printed circuit board (PCB) module, and in vivo animal and human subject studies. All animal experiments were performed in compliance with protocols approved by the University of California, Los Angeles Animal Research Committee (UCLA ARC Protocol Nos. 2015-079, 2019-019, and 2021-011). The conducted human subject experiments were performed in compliance with the protocols that are approved by the Institutional Review Board at the University of California, Los Angeles (IRB#17-000170). All subjects gave written informed consent before participation in the study.
Appendix 01 (PDF) Dataset S01 (XLSX) Dataset S02 (XLSX)
|
Identifying a Biocontrol Bacterium with Disease-Prevention Potential and Employing It as a Powerful Biocontrol Agent Against | 3d0a0dab-39e7-4605-b5c9-6ae25c2b5c6d | 11766301 | Microbiology[mh] | The tomato ( Solanum lycopersicum ) of the Solanaceae family is cultivated worldwide . In recent years, tomato has become the leading crop in the world, providing farmers with an important economic resource. However, tomato production is seriously restricted by soil-borne pathogens such as F. oxysporum , which is known to infect different types of hosts and cause serious damage to crops such as cotton, tomatoes, and bananas . F. oxysporum enters the plant through the root and propagates within the vascular bundle, disrupting the ability of the root tissue to transport nutrients to the vascular bundle, resulting in the host’s eventual death, seriously reducing fruit yield and quality. Moreover, the frequent and excessive use of chemical fungicides produces serious environmental and food safety problems . Therefore, a study on several topics related to the control of F. oxysporum is necessary. Biological control is a safe and environmentally friendly option for preventing and treating tomato plant wilt disease. It offers a new alternative that is beneficial for humans and the environment . Biological control has been studied for over 100 years and is considered a viable alternative method to chemical control . It is one of the best potential strategies for the control of soil-borne diseases . Bacteria as the primary source of biological antibacterial agents have the advantages of rapid reproduction and easy preparation of fermentation solutions and the most noticeable control effect . Li et al. found that the A57 strain of Bacillus subtilis could considerably suppress cotton standing blight and yellow wilt, inhibiting the germination of conidia and causing spore deformation to achieve a control effect . Other studies showed that Bacillus megaterium BM1 and B. velezensis RC218 were both capable of considerably reducing the occurrence of wheat blast disease . Lu et al. reported that a new species of streptomyces , Streptomyces rimosus M527, not only displayed broad-spectrum antifungal activity but also showed the strongest antagonistic activity against the spore germination of F. oxysporum f.sp. cucumerinum . In addition, biological control has not only focused on the control effects of strains. Studies showed that Bacillus secretes antagonistic compounds. Wang et al. identified three fungal lipopeptides (iturin, surfactin, and bacillomycin D) secreted by Bacillus amylolyticus W19 . In recent years, chemical residues on fruits that are difficult to clean and harmful to consumers and the massive use of chemical agents to control fungal diseases has caused a decline in the structure of soil aggregate and the destruction of microbial communities . Although there are many reports on the biocontrol of B. velezensis against plant fungal disease, there are very few B. velezensis agents now used in agricultural practices. Therefore, it is necessary to screen out more effective B. velezensis biocontrol strains and understand the influences on plants and pathogens. In this work, we isolated and identified a biocontrol bacterial strain, Y-4, and used it as a research object. The research was mainly carried out in the following ways: (a) an analysis of a potential antibacterial mechanism through whole genome sequencing and metabolic gene clusters, (b) determining the antagonism activity of strain Y-4 on F. oxysporum through a culture dish, (c) evaluating the control efficiency of the Y-4 strain on wild tomatoes indoors, and (d) determining the induction of antioxidant enzymes in tomato leaves by strain Y-4. This study aimed to provide the resources and methods for the control of F. oxysporum , using strain Y-4 as a biological agent. 2.1. Isolation of Biocontrol Bacterium Y-4 A total of six biocontrol strains that could inhibit the reproduction of F. oxysporum were selected from the rhizosphere soil. Thus, we took strain Y-4 as the main research object . 2.2. Identification of Biocontrol Bacterium Y-4 Y-4 colonies were creamy white, oily, paste-like, smooth at the edges, and slightly folded at the center and smooth and opaque on the surface ( A). After microscopic observation and after Gram staining, strain Y-4 was Gram-positive with a purple body and rod-shaped bacteria ( A). 2.2.1. Molecular Identification of Y-4 The 16S rDNA sequence of strain Y-4 was submitted to NCBI’s GenBank database for homology comparison with reported sequences. The strain with more similar homology to strain Y-4 is B. velezensis Y17W strain (MT573877) ( B). To further determine the taxonomic information of strain Y-4, we compared each sequence of strain Y-4 with the NT database. The results are shown in . According to the identity and score comparison results, the complete genomes of B. velezensis strain BA-26 chromosome and B. velezensis strain CGMCC 11640 chromosome are more similar to that of the Y-4 genome . Combined with the comparison results of the 16S rRNA and NT databases, strain Y-4 was determined to be a strain of B. velezensis . 2.2.2. Physiological and Biochemical Identification of Y-4 The results showed that strain Y-4 was an aerobic and alkali-producing type of bacteria. The catalase test was positive. Methyl red was negative . Strain Y-4 can form biofilms that protect plants from infections ( C). Furthermore, it can secrete protease, amylase, cellulase, and siderophores ( and D). 2.3. Genome-Wide Information of the Strain From the assembly results, we learned that the subread length of strain Y-4 was 3,984,866 bp, G + C content was 46.45%, and the proportion of the A, T, G, and C base content was 26.75%, 26.80%, 23.25%, and 23.20%, respectively ( and ). We submitted the sequence data to the NCBI database and obtained the accession number CP139053. 2.4. Genome Structure Annotation Information of the Strain In total, 4022 coding genes were predicted, with a 3,555,231 bp length for all of the projected coding genes. The predicted number of tRNA was 87, with a total length of 6694 bp. There were 3 types of rRNA, 27 were predicted, and the total length was 41,230 bp . 2.5. Gene Function Annotation of the Strain 2.5.1. Annotations on Basic Functions of Strain To study Y-4’s genome function, its sequence was annotated in five databases: NR, SwissProt, COG/KOG, KEGG, and GO. There were 3937 genes annotated in the NR database, 3500 in SwissProt, 3115 in COG/KOG, 2226 in KEGG, and 3055 in GO ( A). The COG database is a system annotated by NCBI based on the distance of the gene linear homology. Through sequence similarity alignment, a protein sequence is assigned to a COG cluster. The Y-4 biocontrol bacterium had 3115 (77.84%) annotated genes that were classified into 24 categories ( B). Among all categories, the number of genes involved in amino acid transport and metabolism was the largest, involving 312 genes ( B). A term is the fundamental unit of GO. In the GO database, 3055 genes (76.34%) of the Y-4 genes were annotated ( C). Cellular anatomical entity had the highest number of genes (1899) in the secondary classification. Then, cellular and metabolic process annotated 1788 and 1440 genes. Catalytic and binding activities annotated 1667 and 1461 genes, respectively ( C). In the KEGG database, 2226 (55.62%) of the Y-4 genes were annotated. The results revealed that the metabolism portion had the most annotated genes, with 1102 genes, followed by the environmental information processing section. In the secondary categorization, 237 genes were annotated to carbohydrate metabolism and 204 genes to amino acid metabolism. The number of genes in these two categories is relatively higher than in the others ( D). 2.5.2. Annotation on Special Functions of Strain In the CAZy database, strain Y-4 annotated to 85 genes, with glycoside hydrolases having the most genes annotated (39 genes), followed by glycosyl transferases (18 genes) and carbohydrate esterases (17 genes) ( E). Polysaccharide lyases had the fewest genes annotated, with only three ( E). In the PHI mutant phenotype statistics, we observed the most prominent number of reduced virulence genes in the Y-4 sample with a gene count of 904 ( F). This is substantially greater than the phenotypes of other mutations. It should be noted that reduced virulence is regarded as a key mutation trait for reducing the pathogenicity of pathogenic bacteria ( F). 2.6. Gene Cluster Analysis of Secondary Metabolites of Strain According to the anticipated results, the strain Y-4 genome had 13 metabolite synthesis gene clusters. Surfactin (lipopeptide), butirosin A/butirosin B (aminoglycoside), macrolactin H (polyketone), bacillaene (polyketone), fengycin (lipopeptide), difficidin (polyketone), bacillibactin (siderophore), and bacilysin (dipeptide) were identified as fungistatic substances . Six compounds (macrolactin H, bacillaene, fengycin, difficidin, bacillibactin, and bacilysin) were 100% similar to known gene clusters, and one compound (surfactin) had 91% similarity. Moreover, the four gene clusters responsible for secondary metabolite synthesis showed no resemblance to the known clusters, and they biosynthesized two terpenes, one T3PKS and one NRPS, respectively . 2.7. Antifungal Activities of Biocontrol Bacterium Y-4 2.7.1. Determination of Broad-Spectrum Bacterial Inhibition of Biocontrol Bacterium Y-4 In this experiment, the Y-4 bacterial suspension showed different degrees of bactericidal effects against six pathogenic bacteria ( B. cinerea , F. oxysporum f.sp. cucumerinum , Colletotrichum orbiculare Arx , Sclerotinia sclerotiorum (Lib.) de Bary , Verticillium fusarium and Fusarium equiseti ). These results showed that Y-4 has broad-spectrum bacterial inhibition ( and A). 2.7.2. Determination of the Inhibition Effect of F. oxysporum Through the culture dish confrontation results, strain Y-4 had the strongest inhibitory effect on F. oxysporum , which was significantly different from CONTROL. In addition, strain wz-37, which was used as a positive control, inhibited the F. oxysporum to a lesser extent than strain Y-4 ( B and ). 2.7.3. Trypan Blue Dyeing Observation of Y-4 The blue area of the tomato leaves treated with the bacterial suspension was much smaller compared to the control group, among which the stained area of leaves treated with the strain Y-4 bacterial suspension was the smallest, indicating that the leaves had the largest number of viable cells. In contrast, the leaves in the CONTROL were completely stained and the leaf cells were completely inactive. After treatment with strain wz-37, the leaf cells were noticeably stained. However, the prevention and control effect of strain Y-4 was found to be higher than that of strain wz-37 ( C,D). 2.8. Determination of Indoor Control Efficiency of Biocontrol Bacterium Y-4 In greenhouse prevention studies, the disease index of the plants treated with strain Y-4 was the lowest and the control efficiency was the best, which were 19.56 and 71.88%, respectively ( A,B). All the plants with the addition of the biological control bacteria had a lower disease index than the CONTROL group. The control efficiency of wz-37 was considerably lower than that of Y-4 ( B). 2.9. Induction of Antioxidant Enzymes in Tomato Leaves by Biocontrol Bacterium Y-4 POD enzyme activity in tomato leaves treated with bacterial suspension of strains Y-4 and wz-37, both reaching a maximum on day 3, with 721.00 min/g for Y-4 and 442.33 min/g for wz-37. The highest POD activity was observed in the strain Y-4 treatment ( C). Therefore, spraying the leaves with the Y-4 biocontrol suspension could considerably increase the POD enzyme activity and further increase the plant’s resistance to pathogenic bacteria. The trend of SOD and POD activities in tomato leaves was similar. Strain wz-37 treatments reached the highest value on the second day, which was 116.89 U/g. The difference was that the treatment of strain Y-4 reached its highest value on the fifth day, which was 136.58 U/g. Moreover, strain Y-4 had the highest SOD activity among all groups ( D). In general, Y-4 bacterial suspensions can boost POD and SOD activities in leaves. A total of six biocontrol strains that could inhibit the reproduction of F. oxysporum were selected from the rhizosphere soil. Thus, we took strain Y-4 as the main research object . Y-4 colonies were creamy white, oily, paste-like, smooth at the edges, and slightly folded at the center and smooth and opaque on the surface ( A). After microscopic observation and after Gram staining, strain Y-4 was Gram-positive with a purple body and rod-shaped bacteria ( A). 2.2.1. Molecular Identification of Y-4 The 16S rDNA sequence of strain Y-4 was submitted to NCBI’s GenBank database for homology comparison with reported sequences. The strain with more similar homology to strain Y-4 is B. velezensis Y17W strain (MT573877) ( B). To further determine the taxonomic information of strain Y-4, we compared each sequence of strain Y-4 with the NT database. The results are shown in . According to the identity and score comparison results, the complete genomes of B. velezensis strain BA-26 chromosome and B. velezensis strain CGMCC 11640 chromosome are more similar to that of the Y-4 genome . Combined with the comparison results of the 16S rRNA and NT databases, strain Y-4 was determined to be a strain of B. velezensis . 2.2.2. Physiological and Biochemical Identification of Y-4 The results showed that strain Y-4 was an aerobic and alkali-producing type of bacteria. The catalase test was positive. Methyl red was negative . Strain Y-4 can form biofilms that protect plants from infections ( C). Furthermore, it can secrete protease, amylase, cellulase, and siderophores ( and D). The 16S rDNA sequence of strain Y-4 was submitted to NCBI’s GenBank database for homology comparison with reported sequences. The strain with more similar homology to strain Y-4 is B. velezensis Y17W strain (MT573877) ( B). To further determine the taxonomic information of strain Y-4, we compared each sequence of strain Y-4 with the NT database. The results are shown in . According to the identity and score comparison results, the complete genomes of B. velezensis strain BA-26 chromosome and B. velezensis strain CGMCC 11640 chromosome are more similar to that of the Y-4 genome . Combined with the comparison results of the 16S rRNA and NT databases, strain Y-4 was determined to be a strain of B. velezensis . The results showed that strain Y-4 was an aerobic and alkali-producing type of bacteria. The catalase test was positive. Methyl red was negative . Strain Y-4 can form biofilms that protect plants from infections ( C). Furthermore, it can secrete protease, amylase, cellulase, and siderophores ( and D). From the assembly results, we learned that the subread length of strain Y-4 was 3,984,866 bp, G + C content was 46.45%, and the proportion of the A, T, G, and C base content was 26.75%, 26.80%, 23.25%, and 23.20%, respectively ( and ). We submitted the sequence data to the NCBI database and obtained the accession number CP139053. In total, 4022 coding genes were predicted, with a 3,555,231 bp length for all of the projected coding genes. The predicted number of tRNA was 87, with a total length of 6694 bp. There were 3 types of rRNA, 27 were predicted, and the total length was 41,230 bp . 2.5.1. Annotations on Basic Functions of Strain To study Y-4’s genome function, its sequence was annotated in five databases: NR, SwissProt, COG/KOG, KEGG, and GO. There were 3937 genes annotated in the NR database, 3500 in SwissProt, 3115 in COG/KOG, 2226 in KEGG, and 3055 in GO ( A). The COG database is a system annotated by NCBI based on the distance of the gene linear homology. Through sequence similarity alignment, a protein sequence is assigned to a COG cluster. The Y-4 biocontrol bacterium had 3115 (77.84%) annotated genes that were classified into 24 categories ( B). Among all categories, the number of genes involved in amino acid transport and metabolism was the largest, involving 312 genes ( B). A term is the fundamental unit of GO. In the GO database, 3055 genes (76.34%) of the Y-4 genes were annotated ( C). Cellular anatomical entity had the highest number of genes (1899) in the secondary classification. Then, cellular and metabolic process annotated 1788 and 1440 genes. Catalytic and binding activities annotated 1667 and 1461 genes, respectively ( C). In the KEGG database, 2226 (55.62%) of the Y-4 genes were annotated. The results revealed that the metabolism portion had the most annotated genes, with 1102 genes, followed by the environmental information processing section. In the secondary categorization, 237 genes were annotated to carbohydrate metabolism and 204 genes to amino acid metabolism. The number of genes in these two categories is relatively higher than in the others ( D). 2.5.2. Annotation on Special Functions of Strain In the CAZy database, strain Y-4 annotated to 85 genes, with glycoside hydrolases having the most genes annotated (39 genes), followed by glycosyl transferases (18 genes) and carbohydrate esterases (17 genes) ( E). Polysaccharide lyases had the fewest genes annotated, with only three ( E). In the PHI mutant phenotype statistics, we observed the most prominent number of reduced virulence genes in the Y-4 sample with a gene count of 904 ( F). This is substantially greater than the phenotypes of other mutations. It should be noted that reduced virulence is regarded as a key mutation trait for reducing the pathogenicity of pathogenic bacteria ( F). To study Y-4’s genome function, its sequence was annotated in five databases: NR, SwissProt, COG/KOG, KEGG, and GO. There were 3937 genes annotated in the NR database, 3500 in SwissProt, 3115 in COG/KOG, 2226 in KEGG, and 3055 in GO ( A). The COG database is a system annotated by NCBI based on the distance of the gene linear homology. Through sequence similarity alignment, a protein sequence is assigned to a COG cluster. The Y-4 biocontrol bacterium had 3115 (77.84%) annotated genes that were classified into 24 categories ( B). Among all categories, the number of genes involved in amino acid transport and metabolism was the largest, involving 312 genes ( B). A term is the fundamental unit of GO. In the GO database, 3055 genes (76.34%) of the Y-4 genes were annotated ( C). Cellular anatomical entity had the highest number of genes (1899) in the secondary classification. Then, cellular and metabolic process annotated 1788 and 1440 genes. Catalytic and binding activities annotated 1667 and 1461 genes, respectively ( C). In the KEGG database, 2226 (55.62%) of the Y-4 genes were annotated. The results revealed that the metabolism portion had the most annotated genes, with 1102 genes, followed by the environmental information processing section. In the secondary categorization, 237 genes were annotated to carbohydrate metabolism and 204 genes to amino acid metabolism. The number of genes in these two categories is relatively higher than in the others ( D). In the CAZy database, strain Y-4 annotated to 85 genes, with glycoside hydrolases having the most genes annotated (39 genes), followed by glycosyl transferases (18 genes) and carbohydrate esterases (17 genes) ( E). Polysaccharide lyases had the fewest genes annotated, with only three ( E). In the PHI mutant phenotype statistics, we observed the most prominent number of reduced virulence genes in the Y-4 sample with a gene count of 904 ( F). This is substantially greater than the phenotypes of other mutations. It should be noted that reduced virulence is regarded as a key mutation trait for reducing the pathogenicity of pathogenic bacteria ( F). According to the anticipated results, the strain Y-4 genome had 13 metabolite synthesis gene clusters. Surfactin (lipopeptide), butirosin A/butirosin B (aminoglycoside), macrolactin H (polyketone), bacillaene (polyketone), fengycin (lipopeptide), difficidin (polyketone), bacillibactin (siderophore), and bacilysin (dipeptide) were identified as fungistatic substances . Six compounds (macrolactin H, bacillaene, fengycin, difficidin, bacillibactin, and bacilysin) were 100% similar to known gene clusters, and one compound (surfactin) had 91% similarity. Moreover, the four gene clusters responsible for secondary metabolite synthesis showed no resemblance to the known clusters, and they biosynthesized two terpenes, one T3PKS and one NRPS, respectively . 2.7.1. Determination of Broad-Spectrum Bacterial Inhibition of Biocontrol Bacterium Y-4 In this experiment, the Y-4 bacterial suspension showed different degrees of bactericidal effects against six pathogenic bacteria ( B. cinerea , F. oxysporum f.sp. cucumerinum , Colletotrichum orbiculare Arx , Sclerotinia sclerotiorum (Lib.) de Bary , Verticillium fusarium and Fusarium equiseti ). These results showed that Y-4 has broad-spectrum bacterial inhibition ( and A). 2.7.2. Determination of the Inhibition Effect of F. oxysporum Through the culture dish confrontation results, strain Y-4 had the strongest inhibitory effect on F. oxysporum , which was significantly different from CONTROL. In addition, strain wz-37, which was used as a positive control, inhibited the F. oxysporum to a lesser extent than strain Y-4 ( B and ). 2.7.3. Trypan Blue Dyeing Observation of Y-4 The blue area of the tomato leaves treated with the bacterial suspension was much smaller compared to the control group, among which the stained area of leaves treated with the strain Y-4 bacterial suspension was the smallest, indicating that the leaves had the largest number of viable cells. In contrast, the leaves in the CONTROL were completely stained and the leaf cells were completely inactive. After treatment with strain wz-37, the leaf cells were noticeably stained. However, the prevention and control effect of strain Y-4 was found to be higher than that of strain wz-37 ( C,D). In this experiment, the Y-4 bacterial suspension showed different degrees of bactericidal effects against six pathogenic bacteria ( B. cinerea , F. oxysporum f.sp. cucumerinum , Colletotrichum orbiculare Arx , Sclerotinia sclerotiorum (Lib.) de Bary , Verticillium fusarium and Fusarium equiseti ). These results showed that Y-4 has broad-spectrum bacterial inhibition ( and A). F. oxysporum Through the culture dish confrontation results, strain Y-4 had the strongest inhibitory effect on F. oxysporum , which was significantly different from CONTROL. In addition, strain wz-37, which was used as a positive control, inhibited the F. oxysporum to a lesser extent than strain Y-4 ( B and ). The blue area of the tomato leaves treated with the bacterial suspension was much smaller compared to the control group, among which the stained area of leaves treated with the strain Y-4 bacterial suspension was the smallest, indicating that the leaves had the largest number of viable cells. In contrast, the leaves in the CONTROL were completely stained and the leaf cells were completely inactive. After treatment with strain wz-37, the leaf cells were noticeably stained. However, the prevention and control effect of strain Y-4 was found to be higher than that of strain wz-37 ( C,D). In greenhouse prevention studies, the disease index of the plants treated with strain Y-4 was the lowest and the control efficiency was the best, which were 19.56 and 71.88%, respectively ( A,B). All the plants with the addition of the biological control bacteria had a lower disease index than the CONTROL group. The control efficiency of wz-37 was considerably lower than that of Y-4 ( B). POD enzyme activity in tomato leaves treated with bacterial suspension of strains Y-4 and wz-37, both reaching a maximum on day 3, with 721.00 min/g for Y-4 and 442.33 min/g for wz-37. The highest POD activity was observed in the strain Y-4 treatment ( C). Therefore, spraying the leaves with the Y-4 biocontrol suspension could considerably increase the POD enzyme activity and further increase the plant’s resistance to pathogenic bacteria. The trend of SOD and POD activities in tomato leaves was similar. Strain wz-37 treatments reached the highest value on the second day, which was 116.89 U/g. The difference was that the treatment of strain Y-4 reached its highest value on the fifth day, which was 136.58 U/g. Moreover, strain Y-4 had the highest SOD activity among all groups ( D). In general, Y-4 bacterial suspensions can boost POD and SOD activities in leaves. Among vegetable crops, the tomato stands out as one of the most widely consumed and cultivated worldwide. Tomato fruit production is significant globally, with an annual production of 182 million tons according to the FAO (Food and Agriculture Organization of the United Nations). Tomatoes are easily infected by diseases during growth, resulting in a decline in fruit quality. F. oxysporum is a key pathogen that causes Fusarium wilt in tomato plants . To reduce the threat of this pathogen, an efficient and ecofriendly control agent for F. oxysporum is required. Furthermore, we were discovered that Y-4 can be used to prevent a range of fungal illnesses and had potent broad-spectrum antibacterial action ( A). Bacillus is a widely employed biological resource for crop protection due to its antagonistic and growth-promoting properties. Bacillus is a commonly used biological resource for protecting crops . B. velezensis is a branch of Bacillus and is a relatively lately discovered type. The prospects for development are strong, and it has many ways it can be used . It was previously found that B. velezensis can prevent and control various kinds of plant-disease-causing pathogens, such as Verticillium dahliae Kleb, Alternaria brassicae , Pyricularia grisea (Cooke) Sacc, and Lettuce Root Rot ( Pythium ) . In this work, we isolated and identified a strain of B. velezensis Y-4 from soil that showed strong antifungal activity. In this test, strain Y-4 could effectively inhibit F. oxysporum infections in plants, thereby reducing the disease index of tomato plants. Through indoor prevention experiments, we found that strain Y-4 was successful in preventing disease caused by F. oxysporum , with it controlling up to 71.88%, considerably higher than that of the CONTROL, and decreasing the disease index ( B). Previous studies showed that the control effect of B. velezensis D18 bacterial suspension on banana blight was 67.9%, which could effectively control the occurrence of fungal diseases . In this study, the control efficiency of the Y-4 bacterial suspension was better than that of the D18 bacterial suspension. Wang et al. isolated a bacterial strain (SF18-3) and identified it as B. velezensis . Its fermentation liquor effectively controlled Xanthomonas campestris pv.vesicatoria (Doidge) Dye (Bacterial Spot Disease of Pepper), with a control efficiency of 70.17% . POD and SOD are two defense enzymes that exist widely in plants. When plants are biologically stressed, they frequently enhance the activity of defensive enzymes to resist fungal infection. In our study, POD and SOD activities increased at first and then decreased gradually ( C,D). POD activity was consistent with the research results of Wang et al. . Our results showed that, after spraying foliar with a biocontrol agent, the strains first propagated and occupied a favorable living space to inhibit the infection of F. oxysporum . When F. oxysporum invades plants, it will cause a large amount of ROS to accumulate. ROS will destroy biofilms and cause membrane peroxidation. The SOD enzyme plays an important role in clearing ROS and alleviating the damage caused by ROS to plants. More kinds of bioactive metabolites are produced by B. velezensis than by other bacterial strains . We analyzed 13 gene clusters in strain Y-4, of which 9 gene clusters can synthesize known antibiotics, including locillomycin/locillomycin B/locillomycin C, surfactin, butirosin A/butirosin B, macrolactin H, bacillaene, fengycin, and others . Fengycin, bacillibactin, and surfactin are classic lipopeptide antibacterial substances. Thurlow et al. reported that a strain of B. velezensis AP193 can produce the polyketide compound difficidin . B. velezensis FZB4 can produce related surfactin and fengycin substances, which achieve antibacterial effects by changing the cell wall of the mycelium or conidia of specific fungi . Fengycin has been reported to be the main antibacterial substance produced by B. velezensis . It can significantly inhibit the growth of Ralstonia solanacearum and F. oxysporum and inhibit the growth of Fusarium solani by interfering with the cell membrane . Bacilysin is a non-ribosomal formed dipeptide antibiotic composed of L-alanine and L-anticapsin, produced by Bacillus . Metabolites produced by microorganisms can regulate the exchange of substances and competition and interaction between organisms ; they are also used as a starter to induce resistance in plants. Therefore, B. velezensis has broad application prospects in biopesticides. According to earlier research, B. velezensis has functional genes that prevent harmful bacteria. For example, a strain of B. velezensis isolated from rice rhizosphere can induce resistance to disease in Arabidopsis thaliana and an overproduction of the PAD4 gene in plants . In the CAZy database, strain Y-4 involves 39 genes annotated in glycoside hydrolases ( E). The majority of glycoside hydrolases are linked to cell wall metabolism, defense mechanisms, secondary metabolism, glycolipid metabolism, and signal transduction . In the PHI database, we found that up to 904 genes in strain Y-4 were annotated into reduced virulence ( F). Therefore, we speculated that the antagonistic effect of strain Y-4 was due to the existence of a disease-resistant gene. The analysis results of the PHI data showed us the direction for further in-depth research. 4.1. Bacteria, Pathogen, and Culture Conditions B. velezensis wz-37 was previously isolated in the laboratory from rhizosphere soil samples of tomatoes, cucumbers, corn, and wheat. The wz-37 was deposited in the China General Microbiological Culture Collection Centre (CGMCC No. 15766) (Beijing, China). The bacterial isolates were grown on LB agar media at 28 °C in the dark (the same as below unless stated otherwise) for 3 days before being used. F. oxysporum , B. cinerea , F. oxysporum f.sp. cucumerinum , Colletotrichum orbiculare Arx , Verticillium fusarium , Fusarium equiseti , and Sclerotinia sclerotiorum (Lib.) de Bary were preserved in the Horticultural Biotechnology Laboratory of Northeast Agricultural University, China (Harbin, China). The pathogen was grown on potato dextrose agar (PDA) at 28 °C in the dark (the same as below unless stated otherwise) for 7 d before being used. Bacterial suspension: On a clean bench, the purified cultures Y-4 and wz-37 (Quantity: 1%) were inoculated into liquid LB agar media and placed in a shaker at 200 rpm, and the bacterial suspension was obtained by shaking the culture for 48 h. We used sterile water to adjust the bacterial suspension concentration to 1 × 10 9 CFU/mL. F. oxysporum suspension: F. oxysporum was incubated in culture dishes for 7 days, and the mycelium was allowed to grow all over the Petri dish, with germinating spores attached to the surface. The mycelium was rinsed with sterile water to disperse the spores in the solution and filtered, and the spore solution was adjusted using a spectrophotometer. We used sterile water to adjust the F. oxysporum suspension concentration to 1 × 10 6 CFU/mL. 4.2. Tomato Seeds The tomato variety used in the experiment was “DRK0568” (also known as the Provence tomato). The seedlings were cultured in the Horticultural Biotechnology Laboratory of Northeast Agricultural University, China (Harbin, China). 4.3. Isolation of Biocontrol Bacterium Soil samples were collected from the rhizosphere soil of healthy Schefflera heptaphylla plants in Xishuangbanna Autonomous Prefecture, Yunnan Province, China. The surface soil was removed, and soil samples were collected at a depth of 5–10 cm. Sampling was performed using the multipoint sampling method, and samples were kept in self-sealing bags and numbered. Soil samples were used to isolate the biocontrol bacterium. The isolation method described for biocontrol strains followed the procedure of Li et al. . After the colony grew out on the culture dish, it was randomly selected for purification. Using F. oxysporum as an indicator pathogen, the bacteria with a good antagonistic effect were screened using the culture dish antagonistic method. 4.4. Identification of Biocontrol Bacterium The biocontrol bacterium was cultured on an LB agar media and observed for morphology (the underlining method is shown in ). 4.4.1. Molecular Identification The DNA extraction and identification of the strain via 16S rRNA were conducted by Beijing LiuheHuada Gene Technology Co., China (Beijing, China). The DNA extraction method was according to the instructions of the Bacterial Genome Kit (Tiangen Biochemical Technology Co., Ltd., Beijing, China). The 16S rRNA sequences of strain Y-4 were submitted to the GenBank database of the NCBI and subjected to BLAST comparison and homology with the reported sequences. The phylogenetic tree of Y-4 evaluated in 1000 replicates was constructed using the neighbor-joining method in the MEGA 10.2.6 software. The whole genome sequencing analysis of strain Y-4 was operated by the Wuhan Frasergen company, China (Wuhan, China). For the bacterial genome assembly part, pb_assembly_hifi_microbial (from smrtlink11.0.0) software was used for assembly, and the second-generation data pair reads-based assembly correction was used to obtain the final assembly results using pilon-1.24 software . Each sequence was compared with the NT database using the BLASTn tool of the ncbi-blast-2.11.0+ software with the comparison options “-task megablast -outfmt 5 -evalue 1e-5 -max_target_seqs 3”. The remaining settings were set to default values. The strain identification was determined by combining the results of the NT database comparison and 16S rRNA gene sequencing. 4.4.2. Physiological and Biochemical Identification Aerobic and anaerobic, catalase, glucose oxidative fermentation, and methyl red tests all refer to Berger’s Manual for the identification of bacteria . Determination of the biofilm-forming ability followed the method of Sun et al. . We referred to Tagele et al. for the determination of protease and Slama et al. for the determination of amylase . In the determination of cellulase, the cellulase detection culture dish was configured. A 6 mm diameter sterile filter paper was placed at the center of the test culture dish and 2 μL of Y-4 bacteria suspension cultured for 48 h was dripped onto the filter paper. The detection culture dish was placed in a constant-temperature incubator at 28 °C to culture for 3–7 days. If there was a transparent halo around the filter paper, it indicated that the strain could produce cellulase. The determination of siderophores followed the procedure of Bernhard Schwyn and J.B. Neilands (1987) . The presence of an orange halo around the colony indicated siderophore production. All experiments in this part were repeated three times. 4.5. Annotation of the Structure and Function of Bacterial Genomes 4.5.1. Genome Structure Annotation We used Glimmer (v3.02) to find the coding genes of microbial DNA (particularly bacteria, archaea, and viruses), tRNAscan-SE (2.0.9) to predict the tRNA of the microbial genome, and RNAmmer (v1.2) to predict rRNA. 4.5.2. Annotation on the Database of Basic Gene Functions Regarding the basic function annotation, we compared the predicted gene protein sequences to the Cluster of Orthologous Groups of Proteins (COG), KEGG, and GO databases. We used diamond (v2.0.9.147) software to compare the protein sequence of the predicted gene to the COG2020 database using the BLASTp command. The E value 1e-5, and the hit with the highest score is selected as the final annotation result. For KEGG, we used the BLASTp command of the diamond (v2.0.9.147) software to align the protein sequence of the predicted gene into the entire library of KOBAS-v3.0 and then used KOBAS-v3.0 software to analyze the comparison results to map the gene ID to KO. Finally, the hierarchical relationship table in the KEGG BRITE library (last updated: 11 October 2021) was used to annotate each level. For GO, we mapped GO term annotations through ID mapping (20210616) using SwissProt annotation results and then annotated at various levels using go-basic.obo (v2021-09-01). 4.5.3. Annotation on the Database of Special Gene Functions To annotate CAZy , we employed dbCAN2 (an automatic web annotation tool for carbohydrate-related enzymes, annotation software HMMER-v3.3.2, parameter E Value <= 1e-5, coverage >= 0.35). In addition, we utilized diamond (v2.0.9.147) software’s BLASTp command to compare the predicted gene’s protein sequence to the PHI database. 4.6. Annotation of the Secondary Metabolite Gene Cluster In this study, the gene clusters produced by the secondary metabolites of the strain were analyzed using anti-SMASH (v5.1.1). 4.7. Antifungal Activities of Biocontrol Bacterium 4.7.1. Determination of Broad-Spectrum Antifungal Effect of Biocontrol Bacterium The abovementioned pathogens were made into 6 mm fungal chunks (we punched holes vertically on the surface of the fungal culture medium and used a sterile inoculation needle to pick it out to obtain a fungal chunk). The fungal chunk was inoculated on the right side of the sterile PDA agar media center, and the sterile Oxford cup was placed on the left. Using a pipette, 80–100 μL of Y-4 and wz-37 bacterial suspensions were injected into the Oxford cups. On the right side of the control group was an F. oxysporum chunk, and 80–100 μL of sterile water was added to the Oxford cup on the left side. We used wz-37 as a positive control group. The culture dish was sealed and incubated in a constant-temperature oven at 28 °C for 7 days. The width of the inhibition zone was measured. The test was repeated three times. 4.7.2. Determination of the Inhibition Effect of F. oxysporum First, we made F. oxysporum chunks (the production method of F. oxysporum chunks is the same as in ). The culture dishes were filled with PDA agar media and dried. We placed the F. oxysporum chunks on the right side of the Petri dish and placed the sterile Oxford cup on the left side. Using a pipette, 80–100 μL of Y-4 and wz-37 bacterial suspensions were injected into Oxford cups, and CONTROL added an equal amount of sterile water to them. The culture dish was sealed and incubated in a constant-temperature oven at 28 °C for 7 days. The width of the inhibition zone was measured. The test was repeated three times. 4.7.3. Trypan Blue Dyeing Observation The pretreatment for the leaves was as follows: The leaves were immersed in 75% ethanol for 30 s. After rinsing, the leaves were immersed in 3% sodium hypochlorite for 3 min and then rinsed three times with sterile water. The petioles were wrapped in sterile, moist, skimmed cotton wool. We applied sterile water (CONTROL), Y-4 bacterial suspension, and wz-37 bacterial suspension (1 × 10 9 CFU/mL) to the leaves. A suspension of F. oxysporum spores (1 × 10 6 CFU/mL) was evenly sprayed on dried leaves. A staining test was performed after 3 days. The dyeing process followed the method described by Yin et al. . The stained leaves were observed for cell color under an electron microscope. Deep cell color indicated leaf cell death, and light color indicated leaf cells were active. 4.8. Determination of Indoor Control Efficiency of Biocontrol Bacterium Biocontrol suspension and F. oxysporum suspension (preparation method followed ) were placed into two spray bottles, respectively. First, we sprayed biocontrol suspension evenly on the surface of the plant. After 24 h, we sprayed F. oxysporum suspension. After 15 days of treatment, plant disease indices and prevention effects were recorded. The test was repeated five times. The disease index and disease control effect were calculated using the following formulas . The treatment was as follows: (1) CONTROL+ F. oxysporum : we sprayed the surface of tomato plants with sterile water first, followed by the F. oxysporum suspension. (2) Y-4 + F. oxysporum : we sprayed the surface of tomato plants with Y-4 suspension first, followed by the F. oxysporum suspension. (3) wz-37 + F. oxysporum : we sprayed the surface of tomato plants with wz-37 suspension first, followed by the F. oxysporum suspension. Disease index = [∑(The number of diseased plants in this grade × Disease grade)/(Total number of plants investigated × the highest disease grade)] × 100. Control efficiency = [(Disease index of control − Disease index of treated group)/Disease index of control] × 100. 4.9. Induction of Antioxidant Enzymes in Tomato Leaves by Biocontrol Bacterium Leaf samples were collected after 7 days of treatment of the plants, as in . The collection period lasted 7 days. For the sampling standard, leaves with the same growth position and similar leaf size were selected and liquid nitrogen was used for quick freezing, followed by storage in −80 °C refrigerators. Peroxidase (POD) and superoxide (SOD) kits (Suzhou Grace Biotechnology Co., Ltd., Suzhou, China) were used to determine the activity of defense enzymes. 4.10. Statistical Analysis Excel: numerical values were the mean ± standard deviation (SD) of triplicates (Version 16.77.1). Significance analysis was performed using the one-way ANOVA test ( p ≤ 0.05); multiple group comparisons were performed using Tukey after ANOVA in SPSS (Version 26.0). The error bars indicate the SD of the data. The column and line chart were drawn using GraphPad Prism 9 (Version 9.0.0). B. velezensis wz-37 was previously isolated in the laboratory from rhizosphere soil samples of tomatoes, cucumbers, corn, and wheat. The wz-37 was deposited in the China General Microbiological Culture Collection Centre (CGMCC No. 15766) (Beijing, China). The bacterial isolates were grown on LB agar media at 28 °C in the dark (the same as below unless stated otherwise) for 3 days before being used. F. oxysporum , B. cinerea , F. oxysporum f.sp. cucumerinum , Colletotrichum orbiculare Arx , Verticillium fusarium , Fusarium equiseti , and Sclerotinia sclerotiorum (Lib.) de Bary were preserved in the Horticultural Biotechnology Laboratory of Northeast Agricultural University, China (Harbin, China). The pathogen was grown on potato dextrose agar (PDA) at 28 °C in the dark (the same as below unless stated otherwise) for 7 d before being used. Bacterial suspension: On a clean bench, the purified cultures Y-4 and wz-37 (Quantity: 1%) were inoculated into liquid LB agar media and placed in a shaker at 200 rpm, and the bacterial suspension was obtained by shaking the culture for 48 h. We used sterile water to adjust the bacterial suspension concentration to 1 × 10 9 CFU/mL. F. oxysporum suspension: F. oxysporum was incubated in culture dishes for 7 days, and the mycelium was allowed to grow all over the Petri dish, with germinating spores attached to the surface. The mycelium was rinsed with sterile water to disperse the spores in the solution and filtered, and the spore solution was adjusted using a spectrophotometer. We used sterile water to adjust the F. oxysporum suspension concentration to 1 × 10 6 CFU/mL. The tomato variety used in the experiment was “DRK0568” (also known as the Provence tomato). The seedlings were cultured in the Horticultural Biotechnology Laboratory of Northeast Agricultural University, China (Harbin, China). Soil samples were collected from the rhizosphere soil of healthy Schefflera heptaphylla plants in Xishuangbanna Autonomous Prefecture, Yunnan Province, China. The surface soil was removed, and soil samples were collected at a depth of 5–10 cm. Sampling was performed using the multipoint sampling method, and samples were kept in self-sealing bags and numbered. Soil samples were used to isolate the biocontrol bacterium. The isolation method described for biocontrol strains followed the procedure of Li et al. . After the colony grew out on the culture dish, it was randomly selected for purification. Using F. oxysporum as an indicator pathogen, the bacteria with a good antagonistic effect were screened using the culture dish antagonistic method. The biocontrol bacterium was cultured on an LB agar media and observed for morphology (the underlining method is shown in ). 4.4.1. Molecular Identification The DNA extraction and identification of the strain via 16S rRNA were conducted by Beijing LiuheHuada Gene Technology Co., China (Beijing, China). The DNA extraction method was according to the instructions of the Bacterial Genome Kit (Tiangen Biochemical Technology Co., Ltd., Beijing, China). The 16S rRNA sequences of strain Y-4 were submitted to the GenBank database of the NCBI and subjected to BLAST comparison and homology with the reported sequences. The phylogenetic tree of Y-4 evaluated in 1000 replicates was constructed using the neighbor-joining method in the MEGA 10.2.6 software. The whole genome sequencing analysis of strain Y-4 was operated by the Wuhan Frasergen company, China (Wuhan, China). For the bacterial genome assembly part, pb_assembly_hifi_microbial (from smrtlink11.0.0) software was used for assembly, and the second-generation data pair reads-based assembly correction was used to obtain the final assembly results using pilon-1.24 software . Each sequence was compared with the NT database using the BLASTn tool of the ncbi-blast-2.11.0+ software with the comparison options “-task megablast -outfmt 5 -evalue 1e-5 -max_target_seqs 3”. The remaining settings were set to default values. The strain identification was determined by combining the results of the NT database comparison and 16S rRNA gene sequencing. 4.4.2. Physiological and Biochemical Identification Aerobic and anaerobic, catalase, glucose oxidative fermentation, and methyl red tests all refer to Berger’s Manual for the identification of bacteria . Determination of the biofilm-forming ability followed the method of Sun et al. . We referred to Tagele et al. for the determination of protease and Slama et al. for the determination of amylase . In the determination of cellulase, the cellulase detection culture dish was configured. A 6 mm diameter sterile filter paper was placed at the center of the test culture dish and 2 μL of Y-4 bacteria suspension cultured for 48 h was dripped onto the filter paper. The detection culture dish was placed in a constant-temperature incubator at 28 °C to culture for 3–7 days. If there was a transparent halo around the filter paper, it indicated that the strain could produce cellulase. The determination of siderophores followed the procedure of Bernhard Schwyn and J.B. Neilands (1987) . The presence of an orange halo around the colony indicated siderophore production. All experiments in this part were repeated three times. The DNA extraction and identification of the strain via 16S rRNA were conducted by Beijing LiuheHuada Gene Technology Co., China (Beijing, China). The DNA extraction method was according to the instructions of the Bacterial Genome Kit (Tiangen Biochemical Technology Co., Ltd., Beijing, China). The 16S rRNA sequences of strain Y-4 were submitted to the GenBank database of the NCBI and subjected to BLAST comparison and homology with the reported sequences. The phylogenetic tree of Y-4 evaluated in 1000 replicates was constructed using the neighbor-joining method in the MEGA 10.2.6 software. The whole genome sequencing analysis of strain Y-4 was operated by the Wuhan Frasergen company, China (Wuhan, China). For the bacterial genome assembly part, pb_assembly_hifi_microbial (from smrtlink11.0.0) software was used for assembly, and the second-generation data pair reads-based assembly correction was used to obtain the final assembly results using pilon-1.24 software . Each sequence was compared with the NT database using the BLASTn tool of the ncbi-blast-2.11.0+ software with the comparison options “-task megablast -outfmt 5 -evalue 1e-5 -max_target_seqs 3”. The remaining settings were set to default values. The strain identification was determined by combining the results of the NT database comparison and 16S rRNA gene sequencing. Aerobic and anaerobic, catalase, glucose oxidative fermentation, and methyl red tests all refer to Berger’s Manual for the identification of bacteria . Determination of the biofilm-forming ability followed the method of Sun et al. . We referred to Tagele et al. for the determination of protease and Slama et al. for the determination of amylase . In the determination of cellulase, the cellulase detection culture dish was configured. A 6 mm diameter sterile filter paper was placed at the center of the test culture dish and 2 μL of Y-4 bacteria suspension cultured for 48 h was dripped onto the filter paper. The detection culture dish was placed in a constant-temperature incubator at 28 °C to culture for 3–7 days. If there was a transparent halo around the filter paper, it indicated that the strain could produce cellulase. The determination of siderophores followed the procedure of Bernhard Schwyn and J.B. Neilands (1987) . The presence of an orange halo around the colony indicated siderophore production. All experiments in this part were repeated three times. 4.5.1. Genome Structure Annotation We used Glimmer (v3.02) to find the coding genes of microbial DNA (particularly bacteria, archaea, and viruses), tRNAscan-SE (2.0.9) to predict the tRNA of the microbial genome, and RNAmmer (v1.2) to predict rRNA. 4.5.2. Annotation on the Database of Basic Gene Functions Regarding the basic function annotation, we compared the predicted gene protein sequences to the Cluster of Orthologous Groups of Proteins (COG), KEGG, and GO databases. We used diamond (v2.0.9.147) software to compare the protein sequence of the predicted gene to the COG2020 database using the BLASTp command. The E value 1e-5, and the hit with the highest score is selected as the final annotation result. For KEGG, we used the BLASTp command of the diamond (v2.0.9.147) software to align the protein sequence of the predicted gene into the entire library of KOBAS-v3.0 and then used KOBAS-v3.0 software to analyze the comparison results to map the gene ID to KO. Finally, the hierarchical relationship table in the KEGG BRITE library (last updated: 11 October 2021) was used to annotate each level. For GO, we mapped GO term annotations through ID mapping (20210616) using SwissProt annotation results and then annotated at various levels using go-basic.obo (v2021-09-01). 4.5.3. Annotation on the Database of Special Gene Functions To annotate CAZy , we employed dbCAN2 (an automatic web annotation tool for carbohydrate-related enzymes, annotation software HMMER-v3.3.2, parameter E Value <= 1e-5, coverage >= 0.35). In addition, we utilized diamond (v2.0.9.147) software’s BLASTp command to compare the predicted gene’s protein sequence to the PHI database. We used Glimmer (v3.02) to find the coding genes of microbial DNA (particularly bacteria, archaea, and viruses), tRNAscan-SE (2.0.9) to predict the tRNA of the microbial genome, and RNAmmer (v1.2) to predict rRNA. Regarding the basic function annotation, we compared the predicted gene protein sequences to the Cluster of Orthologous Groups of Proteins (COG), KEGG, and GO databases. We used diamond (v2.0.9.147) software to compare the protein sequence of the predicted gene to the COG2020 database using the BLASTp command. The E value 1e-5, and the hit with the highest score is selected as the final annotation result. For KEGG, we used the BLASTp command of the diamond (v2.0.9.147) software to align the protein sequence of the predicted gene into the entire library of KOBAS-v3.0 and then used KOBAS-v3.0 software to analyze the comparison results to map the gene ID to KO. Finally, the hierarchical relationship table in the KEGG BRITE library (last updated: 11 October 2021) was used to annotate each level. For GO, we mapped GO term annotations through ID mapping (20210616) using SwissProt annotation results and then annotated at various levels using go-basic.obo (v2021-09-01). To annotate CAZy , we employed dbCAN2 (an automatic web annotation tool for carbohydrate-related enzymes, annotation software HMMER-v3.3.2, parameter E Value <= 1e-5, coverage >= 0.35). In addition, we utilized diamond (v2.0.9.147) software’s BLASTp command to compare the predicted gene’s protein sequence to the PHI database. In this study, the gene clusters produced by the secondary metabolites of the strain were analyzed using anti-SMASH (v5.1.1). 4.7.1. Determination of Broad-Spectrum Antifungal Effect of Biocontrol Bacterium The abovementioned pathogens were made into 6 mm fungal chunks (we punched holes vertically on the surface of the fungal culture medium and used a sterile inoculation needle to pick it out to obtain a fungal chunk). The fungal chunk was inoculated on the right side of the sterile PDA agar media center, and the sterile Oxford cup was placed on the left. Using a pipette, 80–100 μL of Y-4 and wz-37 bacterial suspensions were injected into the Oxford cups. On the right side of the control group was an F. oxysporum chunk, and 80–100 μL of sterile water was added to the Oxford cup on the left side. We used wz-37 as a positive control group. The culture dish was sealed and incubated in a constant-temperature oven at 28 °C for 7 days. The width of the inhibition zone was measured. The test was repeated three times. 4.7.2. Determination of the Inhibition Effect of F. oxysporum First, we made F. oxysporum chunks (the production method of F. oxysporum chunks is the same as in ). The culture dishes were filled with PDA agar media and dried. We placed the F. oxysporum chunks on the right side of the Petri dish and placed the sterile Oxford cup on the left side. Using a pipette, 80–100 μL of Y-4 and wz-37 bacterial suspensions were injected into Oxford cups, and CONTROL added an equal amount of sterile water to them. The culture dish was sealed and incubated in a constant-temperature oven at 28 °C for 7 days. The width of the inhibition zone was measured. The test was repeated three times. 4.7.3. Trypan Blue Dyeing Observation The pretreatment for the leaves was as follows: The leaves were immersed in 75% ethanol for 30 s. After rinsing, the leaves were immersed in 3% sodium hypochlorite for 3 min and then rinsed three times with sterile water. The petioles were wrapped in sterile, moist, skimmed cotton wool. We applied sterile water (CONTROL), Y-4 bacterial suspension, and wz-37 bacterial suspension (1 × 10 9 CFU/mL) to the leaves. A suspension of F. oxysporum spores (1 × 10 6 CFU/mL) was evenly sprayed on dried leaves. A staining test was performed after 3 days. The dyeing process followed the method described by Yin et al. . The stained leaves were observed for cell color under an electron microscope. Deep cell color indicated leaf cell death, and light color indicated leaf cells were active. The abovementioned pathogens were made into 6 mm fungal chunks (we punched holes vertically on the surface of the fungal culture medium and used a sterile inoculation needle to pick it out to obtain a fungal chunk). The fungal chunk was inoculated on the right side of the sterile PDA agar media center, and the sterile Oxford cup was placed on the left. Using a pipette, 80–100 μL of Y-4 and wz-37 bacterial suspensions were injected into the Oxford cups. On the right side of the control group was an F. oxysporum chunk, and 80–100 μL of sterile water was added to the Oxford cup on the left side. We used wz-37 as a positive control group. The culture dish was sealed and incubated in a constant-temperature oven at 28 °C for 7 days. The width of the inhibition zone was measured. The test was repeated three times. F. oxysporum First, we made F. oxysporum chunks (the production method of F. oxysporum chunks is the same as in ). The culture dishes were filled with PDA agar media and dried. We placed the F. oxysporum chunks on the right side of the Petri dish and placed the sterile Oxford cup on the left side. Using a pipette, 80–100 μL of Y-4 and wz-37 bacterial suspensions were injected into Oxford cups, and CONTROL added an equal amount of sterile water to them. The culture dish was sealed and incubated in a constant-temperature oven at 28 °C for 7 days. The width of the inhibition zone was measured. The test was repeated three times. The pretreatment for the leaves was as follows: The leaves were immersed in 75% ethanol for 30 s. After rinsing, the leaves were immersed in 3% sodium hypochlorite for 3 min and then rinsed three times with sterile water. The petioles were wrapped in sterile, moist, skimmed cotton wool. We applied sterile water (CONTROL), Y-4 bacterial suspension, and wz-37 bacterial suspension (1 × 10 9 CFU/mL) to the leaves. A suspension of F. oxysporum spores (1 × 10 6 CFU/mL) was evenly sprayed on dried leaves. A staining test was performed after 3 days. The dyeing process followed the method described by Yin et al. . The stained leaves were observed for cell color under an electron microscope. Deep cell color indicated leaf cell death, and light color indicated leaf cells were active. Biocontrol suspension and F. oxysporum suspension (preparation method followed ) were placed into two spray bottles, respectively. First, we sprayed biocontrol suspension evenly on the surface of the plant. After 24 h, we sprayed F. oxysporum suspension. After 15 days of treatment, plant disease indices and prevention effects were recorded. The test was repeated five times. The disease index and disease control effect were calculated using the following formulas . The treatment was as follows: (1) CONTROL+ F. oxysporum : we sprayed the surface of tomato plants with sterile water first, followed by the F. oxysporum suspension. (2) Y-4 + F. oxysporum : we sprayed the surface of tomato plants with Y-4 suspension first, followed by the F. oxysporum suspension. (3) wz-37 + F. oxysporum : we sprayed the surface of tomato plants with wz-37 suspension first, followed by the F. oxysporum suspension. Disease index = [∑(The number of diseased plants in this grade × Disease grade)/(Total number of plants investigated × the highest disease grade)] × 100. Control efficiency = [(Disease index of control − Disease index of treated group)/Disease index of control] × 100. Leaf samples were collected after 7 days of treatment of the plants, as in . The collection period lasted 7 days. For the sampling standard, leaves with the same growth position and similar leaf size were selected and liquid nitrogen was used for quick freezing, followed by storage in −80 °C refrigerators. Peroxidase (POD) and superoxide (SOD) kits (Suzhou Grace Biotechnology Co., Ltd., Suzhou, China) were used to determine the activity of defense enzymes. Excel: numerical values were the mean ± standard deviation (SD) of triplicates (Version 16.77.1). Significance analysis was performed using the one-way ANOVA test ( p ≤ 0.05); multiple group comparisons were performed using Tukey after ANOVA in SPSS (Version 26.0). The error bars indicate the SD of the data. The column and line chart were drawn using GraphPad Prism 9 (Version 9.0.0). In conclusion, strain Y-4 as a biocontrol bacterium has been widely used in plant disease management owing to its high efficiency, safety, and environmental friendliness. Strain Y-4 is an aerobic bacterium, which can produce biofilm and many extracellular enzymes. Strain Y-4 plays a crucial role in reducing F. oxysporum -induced blight, safeguarding tomato plants against harmful bacteria and ensuring high-quality and high-yield tomato fruits. In summary, strain Y-4 is a newly discovered and highly effective agent for pathogen control. |
Relative improvement in language vs. motor functions with reperfusion therapies for large vessel occlusion | 4da1155c-44f4-45b3-b8eb-d5d30190dcfd | 11850585 | Surgical Procedures, Operative[mh] | Treatment of acute ischemic stroke is focused on improving perfusion, both to salvage tissue and to improve function. Endovascular thrombectomy (EVT) is a minimally invasive procedure that includes a variety of interventions to remove the clot that is causing ischemia, using catheters inserted into the arteries. Intravenous thrombolysis aims to dissolve the clot causing ischemia by giving medication (e.g. tissue plasminogen activator, or tPA) intravenously. Both EVT and thrombolysis are effective treatments for acute ischemic stroke secondary to large vessel occlusion (LVO). However, they also carry some risks. When clinicians, patients, and their caregivers weigh the potential risks versus benefits of reperfusion therapy, they should consider what functions are likely to recover if blood flow can be restored. Little information is currently available to allow clinicians to know what functions are likely to improve with reperfusion. Because deep and motor areas of the brain, including caudate, putamen, insular ribbon, middle frontal gyrus, frontal lobe subcortical white matter, precentral gyrus, and frontal lobe paracentral lobule are particularly vulnerable to hypoperfusion and infarct relatively early in acute stroke , we hypothesized that reperfusion therapies are more likely to improve language function and neglect (which depend on cortical areas), more than motor function. That is, successful restoration of blood flow might be more likely to salvage areas that are not already ischemic, including areas critical for language – and spatial attention – , such as left inferior frontal gyrus, temporal cortex, and inferior parietal cortex. Although language and spatial attention clearly depend on complex cortical-subcortical networks, our previous studies have shown that hypoperfusion of cortical regions (or thalamus) is necessary to cause frank aphasia and neglect in acute stroke, and reperfusion of these cortical regions results in recovery , – . The goal of our study is to evaluate the percent improvement (mean change in score/maximum score) for different items of the National Institutes of Health Score Scale (NIHSS) with and without EVT, and/or thrombolysis in series of patients who had LVO, pretreatment CT angiogram (CTA) to confirm LVO, and NIHSS for evaluation of acute ischemic stroke (a convenience sample from three hospitals). Participants In this IRB approved retrospective study of our prospective collected database, patients with AIS caused by an LVO (defined as distal internal artery, M1, or proximal M2 middle cerebral artery segments) on CTA from 2017 to 2022 from three centers within our larger hospital enterprise. Demographics and interventions received are described in the Results section. Due to the retrospective nature of the study, The Johns Hopkins Institutional Review Board waived the need of obtaining informed consent. Behavioral scoring On the NIHSS, language is evaluated with a 3-point scale (0 = no aphasia, 1 = mild-moderate aphasia, and 2 = severe aphasia). There are additional items that require language: 1b. Level of consciousness questions (0–2 points) and 1c. Level of consciousness commands (0–2 points), such that a Total Language score ranges from 0 to 6). Motor function includes 4 points strength of each arm, 4 points strength of each leg, and 2 points for facial strength (1 point for unilateral facial weakness). Since our participants all had unilateral MCA or ICA occlusions, we assumed their stroke-induced motor dysfunction would be unilateral, for a total range = 0–9). Neglect (11. Extinction and inattention) is assessed with a single item of 0–2 points. To evaluate percent change in each function, for each patient we divided the change in score from pre- to post-treatment, by the maximum possible score for that function, and multiplied by 100. For example, if a patient improved from severe aphasia to mild-moderate aphasia, the percent change would be ½ x 100 = 50% for the language item; if the same person improved from complete right hemiplegia of the face, arm, and leg, the percent change in motor function would be 9/9 × 100, or 100%. All participants were scored for each of the functions of interest. All participants were administered the NIHSS at admission and discharge from the stroke service (a requirement of our stroke center). Statistical analysis We first evaluated percent change (change in score/maximum score) in language, total language (language item + orientation questions and commands), motor (strength in arms and legs and face) across treatment groups (thrombolysis only, EVT only, EVT plus thrombolysis, and no reperfusion therapy) using ANOVA. For all further analyses, we combined EVT only and EVT plus thrombolysis, because there were only small differences between these two groups in percent change in the items of interest (see Table for exact numbers for each group). We then used paired t-tests to evaluate differences in percent change in total motor function versus the language item (and total language score) for participants who were aphasic (1 or more points on the language item) and/or had right sided weakness of the arm, leg, or face (1 or more points). We also used paired t-tests to evaluate differences in percent change in total motor function versus neglect/extinction for participants who were had neglect/extinction (1 or more points on this item) and/or had weakness of the left arm, leg, or face (1 or more points). Finally, we carried out Fisher’s exact tests to evaluate the association between improvement (dichotomous value as 0 and > 0 points change) and each treatment. P value of = < 0.05 was considered significant. In this IRB approved retrospective study of our prospective collected database, patients with AIS caused by an LVO (defined as distal internal artery, M1, or proximal M2 middle cerebral artery segments) on CTA from 2017 to 2022 from three centers within our larger hospital enterprise. Demographics and interventions received are described in the Results section. Due to the retrospective nature of the study, The Johns Hopkins Institutional Review Board waived the need of obtaining informed consent. On the NIHSS, language is evaluated with a 3-point scale (0 = no aphasia, 1 = mild-moderate aphasia, and 2 = severe aphasia). There are additional items that require language: 1b. Level of consciousness questions (0–2 points) and 1c. Level of consciousness commands (0–2 points), such that a Total Language score ranges from 0 to 6). Motor function includes 4 points strength of each arm, 4 points strength of each leg, and 2 points for facial strength (1 point for unilateral facial weakness). Since our participants all had unilateral MCA or ICA occlusions, we assumed their stroke-induced motor dysfunction would be unilateral, for a total range = 0–9). Neglect (11. Extinction and inattention) is assessed with a single item of 0–2 points. To evaluate percent change in each function, for each patient we divided the change in score from pre- to post-treatment, by the maximum possible score for that function, and multiplied by 100. For example, if a patient improved from severe aphasia to mild-moderate aphasia, the percent change would be ½ x 100 = 50% for the language item; if the same person improved from complete right hemiplegia of the face, arm, and leg, the percent change in motor function would be 9/9 × 100, or 100%. All participants were scored for each of the functions of interest. All participants were administered the NIHSS at admission and discharge from the stroke service (a requirement of our stroke center). We first evaluated percent change (change in score/maximum score) in language, total language (language item + orientation questions and commands), motor (strength in arms and legs and face) across treatment groups (thrombolysis only, EVT only, EVT plus thrombolysis, and no reperfusion therapy) using ANOVA. For all further analyses, we combined EVT only and EVT plus thrombolysis, because there were only small differences between these two groups in percent change in the items of interest (see Table for exact numbers for each group). We then used paired t-tests to evaluate differences in percent change in total motor function versus the language item (and total language score) for participants who were aphasic (1 or more points on the language item) and/or had right sided weakness of the arm, leg, or face (1 or more points). We also used paired t-tests to evaluate differences in percent change in total motor function versus neglect/extinction for participants who were had neglect/extinction (1 or more points on this item) and/or had weakness of the left arm, leg, or face (1 or more points). Finally, we carried out Fisher’s exact tests to evaluate the association between improvement (dichotomous value as 0 and > 0 points change) and each treatment. P value of = < 0.05 was considered significant. A total of 290 patients with LVO were included in the analyses. Mean age was 61.8 (SD 14.0; range 18–97); 139 (47.9%) were female. MRI confirmed infarct was in the left hemisphere in 150 (51.7%) and right hemisphere in 140 (48.3%). Of the 290 patients, 37 (12.8%) received thrombolysis only; 21 (7.2%) underwent EVT and thrombolysis; 18 (6.2%) underwent EVT only, and 214 (73.8%) received neither treatment. In this study, the thrombolysis used was intravenous tPA for all patients. Patients who received neither treatment had one or more contra-indications (most commonly delayed arrival). The mean NIHSS score was 4.7 (SD 4.4) in the thrombolysis group, 11.8 (SD 7.1) in the EVT plus thrombolysis group, 10.8 (SD 7.0) in the EVT only group, and 2.7 (SD 4.4) in the group who received neither intervention. For the entire population ( n = 290), there were significant differences between treatment groups for all outcome measures (Table ). For all outcome measures (percent change in language, total language, motor, and neglect) there were significant effects of treatment group ( p < 0.0001 for all), with the greatest change in the EVT + thrombolysis only group, then EVT only group, followed by thrombolysis only, followed by no intervention. For remaining analyses, we combine EVT only and EVT + thrombolysis, since the differences were small and non-significant (by t-test) for all outcomes. However, the results for EVT only and for EVT + thrombolysis groups are given in the supplement (Table S). For both groups, the trends were the same as for the combined EVT (with or without thrombolysis), but differences in percent gains were not statistically significant, given the small numbers in each group. Likewise, hereafter, we also report results for language item alone, rather than total language, as there was generally less change in orientation and simple commands, but no significant difference between the percent improved in the two outcomes with thrombolysis ( p = 0.80) or EVT ( p = 0.70) (see Table ). The most important comparison in improvement is among those patients who had deficits at baseline. Therefore, we compared percent change in each function for the subset who had deficits at baseline (Table ). For patients with aphasia and/or right sided weakness ( n = 94), the percent change in language was greater than the percent change in weakness (29.8 vs. 12.4; t = 5.3; df93; p < 0.0001; see Table for 95% confidence intervals) for the same patients. For this subset of 94 patients who had deficits at baseline, the greater improvement in language than motor function was observed for all treatment groups (Table ). For those who received thrombolysis alone ( n = 16) the mean difference was 35.4% (SD 35.4) versus 6.6% (SD 14.3) (t = 3.1; df15; p = 0.008). For those who received EVT (with or without thrombolysis) ( n = 23),the difference for percent change in language vs. weakness was 46.4% (SD 37.3) versus 28.1% (SD19.0) (t = 2.3; df22; p = 0.03). For patients who received neither EVT nor IV thrombolysis (the largest group, n = 55), the difference was 21.2% (SD 29.7) versus 7.5% (SD 15.7) (t = 3.9; df54; p = 0.0003). Relatively few patients had neglect at baseline, and there was no significant difference in percentage improvement in neglect versus weakness, in those with neglect and/or weakness pretreatment, with either intervention. However, among the few patients with neglect and weakness pretreatment, there was greater percent improvement in neglect compared to motor function (66.7 ± 28.9 versus − 1.7% ±3.0; t = 4.3; df2; p = 0.0498). Only 26 patients had neglect/extinction pre-treatment; 6/8 (75%) who received thrombolysis showed some improvement in neglect; 13/14 (92.9%) who received EVT showed some improvement with treatment; and 7/10 (70%) showed some improvement with neither intervention (ns by Fisher’s exact). The association between change in each outcome measure (as a dichotomous value) and treatment group was significant by Fisher’s exact both language and strength. Figure shows the percentage of patients who showed any improvement on language, weakness, or neglect for those who had deficits at baseline. Interestingly, of those with aphasia and/or right sided weakness at baseline, 62.5% improved in language with thrombolysis, 78.3% improved in language with EVT, and fewer than half (42.9%) improved with neither intervention. For all functions, improvements were greatest with EVT, then thrombolysis, then no treatment. The difference between treatment groups was significant by Fishers Exact for language ( p = 0.01) and motor function ( p < 0.0001). A greater percentage of patients showed some improvement in strength than language without either intervention, while a greater percentage of patients showed some improvement in language than strength with thrombolysis (Fig. ). However, there were no significant differences in percentage of patients who made any improvement in language or any improvement in strength, of those who were aphasia and/or weak pre-treatment, for any of the intervention groups (by Fisher’s exact). Consistent with our hypothesis, the predominantly cortical function of language improved more (measured as percent change in score) than strength in the same patients, with EVT and/or thrombolysis. Unsurprisingly, EVT had an enormous effect on language, motor, and spatial attention functions; and thrombolysis had a significant, but smaller effect. However, among those with no intervention aimed to restore blood flow, a higher percentage of patients with deficits at baseline improved in strength than in language (or spatial attention). Therefore, it is unlikely that results can be accounted for by a more sensitive assessment item for language than for motor strength. Rather, results may reflect that reperfusion achieved with EVT or thrombolysis (or both), affects cortical regions that affect language more than strength. While reperfusion of the motor strip would also improve motor function (as reflected in strong effects of treatment on motor function), early infarct of subcortical tissues may limit the percent improvement in strength. Our results are important for weighing the risks and benefits of interventions and for counseling regarding prognosis. Patients and families should be advised that deficits such as aphasia might improve more than hemiplegia with intervention, although patients are likely to show at least some improvement in strength with or without treatment (albeit more with treatment). This information may also have implications for selecting outcome measures for interventions to restore blood flow in acute stroke. Currently, the most common outcome measure is the modified Rankin Scale (mRS). However, the mRS is not especially sensitive to deficits that may improve most, such as aphasia. It is more sensitive to motor functions that impede walking. Many previous authors have reviewed the outcome measures used in acute stroke trials (e.g., 13, 14). One study recommended adding “extended/instrumental activities and advanced mobility as components of the primary outcome measure” . Although the mRS was the common outcome assessment (64.3%); others used more sensitive measures such as the Barthel index (40.5%), but many failed to provide full details on outcome assessment methodology . Some trials have addressed the limited sensitivity of the mRS by using the utility-weighted mRS (ranges from 0 [death] to 10 [no symptoms or disability]) . There are also ongoing collaborations among stroke trialists to develop more sensitive outcome measures especially for mild stroke, in order to evaluate the effectiveness of interventions for patients with mild deficits (e.g., mRS 0 or 1) . It is hoped that future acute stroke trials will use a single, validated outcome measure that is sensitive to improvement (or deterioration) in all functions important to stroke patients. There are important limitations to this study. We used items on the NIHSS as the only assessment of language, spatial attention, and strength. While the scoring is reliable, they are very limited measures of all of these functions. It is possible that fine motor control or other motor function might improve more than proximal strength measured by holding up each arm for 10 s and each leg for 5 s. Neglect/extinction was uncommon as measured with the NIHSS, and so we could not reliably compare percent improvement in this domain to other domains. A previous study showed that change in a more objective measurement of neglect (with simple line cancellation) actually correlated better with change in volume hypoperfusion than did change in the total NIHSS score in patients with right hemisphere stroke . Furthermore, while we analyzed percentage change in NIHSS score as a continuous numerical value, the absolute change is a limited interval in scale. Since the NIHSS score questions for language and neglect are less granular (0–3 and 0–2 respectively) than for motor function (0–9 for each side of the body), it is possible that the NIHSS is more sensitive for motor function than for change in language or neglect (e.g., changing from moderate to mild aphasia is no change in score). However, the NIHSS also might over-estimate change in language. That is, a patient who improves from global aphasia to severe aphasia (or severe aphasia to mild-to-moderate aphasia) will have an improvement of 50% using our methodology. It may be more “difficult” to show an improvement of 50% in motor function. We did try to partially address this concern by evaluating change in “total language”, using points for correct answers to level of consciousness questions as well as the language item (for a range from 0 to 6), but the percentage change in language was not significantly different. The difference in scales for motor and language/neglect remains a limitation that we could not resolve. This study was also an observational study, retrospectively analyzing data from a prospectively collected sample of patients. Intervention was not randomized, but was generally determined by following American Heart Association guidelines for treatment of acute ischemic stroke. The aim of this study was not, however, to compare interventions, but to compare improvements across functions in individuals in each intervention group. We also did not report additional demographic details other than age and sex (e.g., education, medications, co-morbidities), because the patients in each comparison (e.g., percent improvement in language versus percent improvement in motor function) were the identical patients. That is, all patients who had language and/or motor deficits were included in the primary analyses, and each patient would have been scored for percent change in language, percent change in total language, and percent change in motor function. It is unclear what variables (other than reperfusion) would differentially affect language versus motor functions in the same individuals. Finally, our hypothesis that motor function will not improve as much as language/neglect was based on the assumption that deeper areas of the brain suffer irreversible ischemia faster than cortical areas. This assumption was based on previous studies of evolution of infarct in acute stroke that used MRI or Positron Emission Topography , . However, it is a limitation of this study that we did not have a second image in these participants to show that deeper areas were more likely to evolve to infarct than cortical areas. Future studies to confirm our hypothesis should include analysis of both an initial image and follow-up image of the infarct, at the same times as behavioral testing. Future studies should also evaluate the influence of age and other variables that might affect outcome. Nevertheless, results provide novel and important information about the likelihood and estimated degree of improvement in gross measures of aphasia, neglect, and weakness in acute ischemic stroke, which may be useful in clinical decision-making. Moreover, results can help clinician provide patients and families a realistic idea of what is likely to improve the most with treatment. In patients with both aphasia and weakness pretreatment, language is likely to improve to a greater degree than weakness with thrombolysis or EVT. Below is the link to the electronic supplementary material. Supplementary Material 1 |
Cervical cancer screening by cytology and the burden of epithelial abnormalities in low resource settings: a tertiary-center 42-year study | fc78cf94-a467-41fb-81ee-6c3e476ff58f | 11253459 | Pathology[mh] | Cervical cancer, a matter of public health concern, is the fourth most prevalent form of cancer in women worldwide and ranks as the 14th in Egypt . The natural history of cervical cancer has a multistep carcinogenesis model, wherein the critical stages are Human Papilloma Virus (HPV) infection, progression to a precancerous state, and subsequent development of invasive cancer . The incidence of cervical cancer, a largely preventable cancer, has decreased over the past few decades in most countries, although the extent of this reduction varies. HPV vaccination and screening for precursor lesions, followed by appropriate follow-up and treatment contribute to this reduction . This laid the foundation for the World Health Organization (WHO) global initiative to eliminate cervical cancer. One cornerstone of this initiative aims at achieving a target of 70% screening coverage of women aged 30–49 years with a high-performance test . Screening forms the initial mandatory step of the two approaches towards the prevention of cervical cancer, whether the screen-and-treat approach or the screen, triage and treat approach . Screening is available in many countries through either population-based (organized) or non-population-based (unorganized) programs, as well as through opportunistic screening. Participation rates and coverage vary significantly across countries and settings. The key factors influencing participation are socioeconomic status, health insurance coverage, public awareness, and level of education. Additionally, some women may face barriers to accessing these services due to a lack of power, authority, or control, which can be major obstacles to their participation . According to the 2015 Egypt Health Issues Survey, the knowledge and practice of Pap smears for cervical cancer screening among women are quite limited. Only 7 percent of women between the ages of 15 and 59 had ever heard of a pap smear, and a very small percentage (0.3 percent) had undergone the procedure. The proportion of women who were aware of pap smears exceeded 10 percent only among women residing in urban governorates (14 percent), women working for cash (13 percent), and women in the highest wealth quintile (11 percent). In all subgroups, one percent or less of women reported ever having had a Pap smear . Given the limited knowledge and low practice of cervical screening among Egyptian women, coupled with the absence of a national population-based screening program, there is a lack of comprehensive understanding regarding the trends of epithelial abnormalities in cervical cytology within the largest population in Africa and the Middle East. Previous research has not thoroughly investigated the outcomes of cervical cancer screening initiatives among Egyptian women or how they evolve over time. Therefore, our objective was to examine the temporal trends in cervical cytology practice and the observed epithelial abnormalities. This retrospective chart review used routinely collected health data from the Early Cancer Detection Unit (ECDU) following Institutional Review Board approval from the Department of Obstetrics and Gynecology, Ain Shams University, Cairo, Egypt. Study population and data source Study population included adult women who attended Ain Shams University Hospital of Obstetrics and Gynecology, whether in the ECDU unit, Gynecologic outpatient clinic, or inpatient and were willing to have a cervical cytology for the first time. The database did not include women who were pregnant, menstruating, known to have invasive or preinvasive cervical cancer, or a prior cervical surgery. The ECDU began opportunistic screening and treatment for cervical cancer in March 1981. The ECDU provides the service at a subsidized cost that is much cheaper than women would find at other healthcare facilities in Egypt. Participants included all adult women who did not have any prior screening in the past 10 years and who underwent conventional cervical screening in the ECDU from March 1981 to December 2022. No formal referral process exists between healthcare facilities in Egypt. However, the resources available at the ECDU for the detection and treatment of cervical cancer make screening at the ECDU an essential step in the continuum of cancer prevention and control within the country. As a result, many women are referred to the ECDU from public or private healthcare facilities across Egypt to obtain a pap smear to obtain follow-up evaluations after a suspicious screening using other methods, such as visual inspection aided by acetic acid or Lugol’s iodine, or to begin the process of cancer treatment. The ECDU does not employ a systematic method of recruiting women to undergo cervical cancer screening; therefore, the women were either self-referred or referred by a healthcare provider for screening. Data identifying the source of referral for screening as the individual or a healthcare provider were not captured in this dataset. According to ECDU institutional practice, women who have low-grade infections at the time of a Pap smear are screened again within 6 months of the prior visit, indicating that women often have multiple screening visits. The ECDU started to use an electronic database in 1986, in addition to paper records, to capture relevant demographics (i.e., age, address, marital status, number of pregnancies), risk factors for cervical cancer (i.e., previous sexually transmitted infections, parity, oral contraceptive use), and clinical outcomes (outcomes of Pap smear) of women who were screened at the clinic. In the first Pap smear, a unique identifying number was used to update each woman’s record. The database has been maintained since its inception, making it a reliable source of longitudinal data for examining trends in cervical cancer screening in a tertiary university healthcare facility in Egypt. All paper records, prior to the establishment of the electronic records system, were transcribed into a database. Cervical cytology (Pap smear) was obtained using an Ayer spatula and spread over a marked glass slide, which was placed in 95% ethyl alcohol and sent to the cytopathology laboratory for conventional cytology examination. The data were recorded using a structured form. The results of the examined smears were reported according to the Bethesda System for Reporting Cervical Cytology. The terminology “CIN1”, “2”, and “3” was used in the ECDU until 2004. Following the adoption of the Bethesda system, the ECDU transformed all the data to match the new system terminology. Data analyses The ECDU provided a deidentified dataset spanning a 42-year period. We used this dataset to describe the clinical and demographic characteristics of screened women, to examine longitudinal trends in screening practice and to report trends in epithelial abnormalities. Epithelial abnormalities captured at screening were categorized into normal results, low-grade abnormalities (Atypical Squamous Cells of Undetermined Significance or AS-CUS, and Low-grade Squamous Intraepithelial Lesion or LSIL classifications), and high-grade abnormalities (Atypical Squamous Cells, High-grade Lesion or ASC-H, Atypical Glandular Cells or AGUS, High-grade Squamous Intraepithelial Lesion or HSIL, and Invasive Cancer). The Bethesda system of classification of cervical smears has evolved over the time period in this study. We have re-categorized results to fit the updated classification where possible. In an effort to maintain consistency of our reported results, we have transitioned our old cytology results to the comparable categories in the Bethesda system. This change allowed us to present all our statistics in a consistent and standardized format, making it easier to track trends and identify insights. The reports of moderate and severe dysplasia were grouped in the group of HSIL. Mild dysplasia and HPV infection were grouped as LSIL. We summarized participants’ characteristics, including age at first screening, reporting means and standard deviations. We summarized Pap smear results by the Bethesda classification system as categorical variables, reporting proportions for each variable. The primary outcome of interest was the number of pap smears completed by year, including the cytological results of those screenings. The data were analyzed using the Statistical Package for the Social Science (SPSS), Version 20 (IBM Corp., Armonk, USA). Study population included adult women who attended Ain Shams University Hospital of Obstetrics and Gynecology, whether in the ECDU unit, Gynecologic outpatient clinic, or inpatient and were willing to have a cervical cytology for the first time. The database did not include women who were pregnant, menstruating, known to have invasive or preinvasive cervical cancer, or a prior cervical surgery. The ECDU began opportunistic screening and treatment for cervical cancer in March 1981. The ECDU provides the service at a subsidized cost that is much cheaper than women would find at other healthcare facilities in Egypt. Participants included all adult women who did not have any prior screening in the past 10 years and who underwent conventional cervical screening in the ECDU from March 1981 to December 2022. No formal referral process exists between healthcare facilities in Egypt. However, the resources available at the ECDU for the detection and treatment of cervical cancer make screening at the ECDU an essential step in the continuum of cancer prevention and control within the country. As a result, many women are referred to the ECDU from public or private healthcare facilities across Egypt to obtain a pap smear to obtain follow-up evaluations after a suspicious screening using other methods, such as visual inspection aided by acetic acid or Lugol’s iodine, or to begin the process of cancer treatment. The ECDU does not employ a systematic method of recruiting women to undergo cervical cancer screening; therefore, the women were either self-referred or referred by a healthcare provider for screening. Data identifying the source of referral for screening as the individual or a healthcare provider were not captured in this dataset. According to ECDU institutional practice, women who have low-grade infections at the time of a Pap smear are screened again within 6 months of the prior visit, indicating that women often have multiple screening visits. The ECDU started to use an electronic database in 1986, in addition to paper records, to capture relevant demographics (i.e., age, address, marital status, number of pregnancies), risk factors for cervical cancer (i.e., previous sexually transmitted infections, parity, oral contraceptive use), and clinical outcomes (outcomes of Pap smear) of women who were screened at the clinic. In the first Pap smear, a unique identifying number was used to update each woman’s record. The database has been maintained since its inception, making it a reliable source of longitudinal data for examining trends in cervical cancer screening in a tertiary university healthcare facility in Egypt. All paper records, prior to the establishment of the electronic records system, were transcribed into a database. Cervical cytology (Pap smear) was obtained using an Ayer spatula and spread over a marked glass slide, which was placed in 95% ethyl alcohol and sent to the cytopathology laboratory for conventional cytology examination. The data were recorded using a structured form. The results of the examined smears were reported according to the Bethesda System for Reporting Cervical Cytology. The terminology “CIN1”, “2”, and “3” was used in the ECDU until 2004. Following the adoption of the Bethesda system, the ECDU transformed all the data to match the new system terminology. The ECDU provided a deidentified dataset spanning a 42-year period. We used this dataset to describe the clinical and demographic characteristics of screened women, to examine longitudinal trends in screening practice and to report trends in epithelial abnormalities. Epithelial abnormalities captured at screening were categorized into normal results, low-grade abnormalities (Atypical Squamous Cells of Undetermined Significance or AS-CUS, and Low-grade Squamous Intraepithelial Lesion or LSIL classifications), and high-grade abnormalities (Atypical Squamous Cells, High-grade Lesion or ASC-H, Atypical Glandular Cells or AGUS, High-grade Squamous Intraepithelial Lesion or HSIL, and Invasive Cancer). The Bethesda system of classification of cervical smears has evolved over the time period in this study. We have re-categorized results to fit the updated classification where possible. In an effort to maintain consistency of our reported results, we have transitioned our old cytology results to the comparable categories in the Bethesda system. This change allowed us to present all our statistics in a consistent and standardized format, making it easier to track trends and identify insights. The reports of moderate and severe dysplasia were grouped in the group of HSIL. Mild dysplasia and HPV infection were grouped as LSIL. We summarized participants’ characteristics, including age at first screening, reporting means and standard deviations. We summarized Pap smear results by the Bethesda classification system as categorical variables, reporting proportions for each variable. The primary outcome of interest was the number of pap smears completed by year, including the cytological results of those screenings. The data were analyzed using the Statistical Package for the Social Science (SPSS), Version 20 (IBM Corp., Armonk, USA). Over a 42-year period from March 1981 to December 2022, 100155 Pap smears were collected from women attending the ECDU at the Department of Obstetrics and Gynecology, Ain Shams University, Cairo, Egypt. Following data cleaning, 95120 smears were obtained for data analysis. The mean age of the women screened was 38.48 years (10.45). The mean age of first intercourse was 20.05 years (1.39). Most women were married and were examined routinely Table . Among women who opted for screening due to clinical symptoms, vaginal discharge was the most common symptom to warrant a smear. Most smears (57072/95120 [60.00%]) were obtained following referral from a healthcare provider to the ECDU. The temporal trend of screening is depicted in Fig. . The median [IQR] number of women who underwent cervical cytology screening was 1982 [1674, 2928]. Epithelial abnormalities were absent in 94.56% of women (89946/95120; with or without inflammatory cells 71971 and 17975, respectively). Pap smears revealed epithelial abnormalities in 5.44% of the participants (5174/95120). Epithelial cell abnormalities included LSIL (low-grade squamous intraepithelial lesion) in 4144 women (4.36%) (condyloma in 3113 (3.27%) and CIN I in 1031 (1.08%) women), atypical squamous cell (ASC) in 378 (0.40) (ASCUS in 351 (0.37%) and ASC-H in 27 (0.03%) women), HSIL (high-grade squamous intraepithelial lesion) in 226 (0.24%) (HSIL-CIN II in 202 and HSIL-CIN III in 24 women), and squamous cell carcinoma in 70 (0.07%) women. There were 184 women (0.19%) with atypical glandular lesions not otherwise specified and 165 women (0.17%) with adenocarcinoma (malignant endocervical cells in 155 and malignant endometrial cells in 10 women), Table . During the period from 1981 to 2004, the percentage of condylomas decreased from 4 to 1%, and that of CIN I decreased from 9% to 0.52%. The detection of LSIL decreased from 59/800 (7.38%) in 1981 to 103/2298 (4.48%) in 2022, while that of HSIL decreased from 10/800 (1.25%) to 3/2298 (0.13%). Preinvasive lesions (ASC, LSIL, and HSIL) declined from 87/800 (10.88%) in 1981 to (123/2298 (5.35%) in 2022. The temporal trends are depicted in Fig. and Fig. . Patients with abnormal cells had a significantly older mean age than those with normal cells (40.5 versus 38.4 years, P < 0.001). Patients with ASC were significantly older than those with LSIL and younger than those with HSIL, and there were more patients with glandular cells (favor neoplastic) ( P < 0.001). The mean age of patients with LSILs was significantly younger than that of patients with other abnormal cells ( P < 0.001). Patients with LSILs and AGUS were significantly younger than patients with malignant squamous cell carcinoma, HSILs or adenocarcinoma ( P < 0.05). Epithelial abnormalities increased significantly with each decade of age, Table . Epithelial abnormalities were significantly more common in women who underwent routine check-ups than in symptomatic women (OR [95% CI] 1.2 [1.11, 1.30], P < 0.001). Women with epithelial abnormalities were significantly younger at first intercourse ( P < 0.001). The yearly number of screened women was positively associated with the observed LSIL (correlation coefficient [95% CI] = 0.84 [0.72, 0.91]) (Fig. ), not significantly associated with the HSIL (correlation coefficient [95% CI] = 0.26 [-0.05, 0.52]) (Fig. ), and negatively associated with the observed suspected invasion (correlation coefficient [95% CI] = -0.55 [-0.73, -0.29]) (Fig. ). Key results By analyzing routinely collected healthcare data from 95,120 satisfactory cervical smear records in the ECDU, this study reported trends in screening uptake and in the observed epithelial abnormalities. A notable increase in the number of annually screened women occurred from 1981 to 1992, after which the number of screened women declined. Women who engaged in sexual intercourse at a young age and who made the conscious decision to undergo routine cytological screening exhibited a greater probability of presenting with epithelial abnormalities. Furthermore, these abnormalities were more prevalent among women who underwent the screening process at a relatively advanced age than among their counterparts. The most common epithelial abnormality was LSILs. The yearly number of cervical screenings by cytology was positively associated with LSIL and negatively associated with invasive lesions. The discrepancy between the recommended cervical cancer screening guidelines and actual clinical practice was identified in a recent systematic review. Only six of 11 countries across North America, Europe, and the Asia–Pacific region have implemented comprehensive population-based screening programs . This observation highlights a potential gap between policy recommendations and real-world implementation of cervical cancer prevention strategies. Current consensus does not recommend the initiation of cervical cancer screening before the age of 21 years in immunocompetent females due to the very low rate of cervical cancer among women aged 20 to 24 years (0.8 per 100,000) . In Egypt, there is a social tendency for early marriage, and the results of this study revealed that epithelial abnormalities are associated with earlier sexual activity. This might have implications for practice. This study revealed a 5.44% prevalence of epithelial abnormalities in cervical smears. This finding aligns with reports from other Arab countries, such as Saudi Arabia (4.27%) , Jordan (3.8%) , the United Arab Emirates (3.3%) , and Kuwait (4.4%) . This similarity might be related to shared cultural and religious factors. This prevalence in our study is much lower than that reported in sub-Saharan Africa, as reported in studies from southwestern Nigeria (34.6%) and Northwest Ethiopia (14.1%) . Coexisting HIV infection in these regions is a potential explanation for the higher rates. In the current study, epithelial abnormalities were significantly more common in women who underwent routine check-ups than in symptomatic women. Other reports have not found such a significant difference . This finding can provide support to the cervical cancer control efforts by providing evidence on the benefit of routine screening rather than only when symptoms arise. In an age-stratified analysis of the distribution of cytological abnormalities, we found that women older than 60 had the highest prevalence of epithelial abnormalities, possibly due to the limited organized screening programs in Egypt in younger women with accumulation of epithelial abnormalities over time. Our results indicate that women with ASC were significantly older than those with LSIL and younger than those with HSIL and glandular cells. Patients with LSILs and AGUS were significantly younger than patients with malignant squamous cell carcinoma, HSILs or adenocarcinoma. This was consistent with the findings of previously published results showing that the incidence of ASC, LSIL, and HSIL peaked in the 30–39-year age group, while the incidence of AGUS peaked among individuals aged 40–49 years. The incidence of malignant lesions further increased after the age of 50 years. The mean ages at diagnosis for patients with LSILs and HSILs were 34.7 and 37.7 years, respectively, while patients with malignant lesions presented with a mean age at diagnosis of 51.8 years . This might have implications for screening practice . Over a 42-year period, the number of screened women has decreased. This would explain the decrease in the number of SIL abnormalities from 6.9% to 4.3% for LSILs and from 1.2% to 0.13% for HSILs. This is because screening services in ECDUs currently operate on an opportunistic basis rather than through a structured, population-based approach. The effectiveness of structured, population-based preventive strategies in reducing the incidence rates of preinvasive and invasive lesions of the cervix cannot be overstated . However, in the vast majority of low- and middle-income countries (LMICs), including Egypt, researchers have identified a diverse array of obstacles to the process of screening . LMICs need prompt and immediate execution of unambiguous regulations, which should be fortified by the ability of the healthcare system to put these regulations into action. We need widespread advocacy within the community and the dissemination of information, alongside the strengthening of policies that promote the well-being of women and ensure gender equality. Strengths and limitations Overall, a four-decade study of cervical screening practice in a low resource setting, offers a wealth of information that can significantly contribute to the understanding and improvement of cervical cancer prevention and control. The extended timeframe allows for the collection of comprehensive data providing insights into trends and changes in cervical screening practices. Understanding how cervical screening practices have evolved over four decades can highlight the impact of policy changes and public health initiatives on screening rates. Highly competent pathologists assessed and confirmed cytologic findings, reducing the chances of incorrect classification of screening results. The results obtained in this study, when interpreted in the context of clinical practice, reveals the actual practice in the largest facility in a country of lower middle income and shows the urgent need to adopt a structured screening program and to incporporate state of the art tools for the diagnosis and management . The extensive dataset collected can serve as a valuable resource for future research, allowing for more detailed analyses and the exploration of new research questions. Limitations of this work includes missing data especially in certain demographic characteristics of women. This might reflect sensitive data in a conservative community or the lack of rigor or inconsistencies when collecting routine data. All demographic data were self-reported by women, which is prone to recall bias. The process of screening for cervical cancer is characterized by opportunistic practices rather than organized efforts. This implies that the subset of women who undergo screening differs from those who do not, as the former group has successfully surmounted various obstacles, such as financial constraints, social factors, cultural influences, and geographic limitations, to avail themselves of screening services. There is also a lack of information relating to the source of referral (e.g., self-referred or provider referred) for screening. The change in terminology used over time may have some implications for the assigned category of abnormality observed. The study lacks the findings on repeated smears for abnormal cytology. HPV testing became a state of the art in the screening and management . However, the current study did not examine how changes in the healthcare system, such as the introduction of new technologies (e.g., HPV testing), have affected cervical screening practices because of the lack of affordable and accessible and HPV testing nationwide. This four-decade longitudinal study could have highlighted the role of preventive measures, such as the HPV vaccine, in reducing the incidence of cervical cancer. Unfortunately, the vaccination coverage is unknown and there is no national vaccination program. By analyzing routinely collected healthcare data from 95,120 satisfactory cervical smear records in the ECDU, this study reported trends in screening uptake and in the observed epithelial abnormalities. A notable increase in the number of annually screened women occurred from 1981 to 1992, after which the number of screened women declined. Women who engaged in sexual intercourse at a young age and who made the conscious decision to undergo routine cytological screening exhibited a greater probability of presenting with epithelial abnormalities. Furthermore, these abnormalities were more prevalent among women who underwent the screening process at a relatively advanced age than among their counterparts. The most common epithelial abnormality was LSILs. The yearly number of cervical screenings by cytology was positively associated with LSIL and negatively associated with invasive lesions. The discrepancy between the recommended cervical cancer screening guidelines and actual clinical practice was identified in a recent systematic review. Only six of 11 countries across North America, Europe, and the Asia–Pacific region have implemented comprehensive population-based screening programs . This observation highlights a potential gap between policy recommendations and real-world implementation of cervical cancer prevention strategies. Current consensus does not recommend the initiation of cervical cancer screening before the age of 21 years in immunocompetent females due to the very low rate of cervical cancer among women aged 20 to 24 years (0.8 per 100,000) . In Egypt, there is a social tendency for early marriage, and the results of this study revealed that epithelial abnormalities are associated with earlier sexual activity. This might have implications for practice. This study revealed a 5.44% prevalence of epithelial abnormalities in cervical smears. This finding aligns with reports from other Arab countries, such as Saudi Arabia (4.27%) , Jordan (3.8%) , the United Arab Emirates (3.3%) , and Kuwait (4.4%) . This similarity might be related to shared cultural and religious factors. This prevalence in our study is much lower than that reported in sub-Saharan Africa, as reported in studies from southwestern Nigeria (34.6%) and Northwest Ethiopia (14.1%) . Coexisting HIV infection in these regions is a potential explanation for the higher rates. In the current study, epithelial abnormalities were significantly more common in women who underwent routine check-ups than in symptomatic women. Other reports have not found such a significant difference . This finding can provide support to the cervical cancer control efforts by providing evidence on the benefit of routine screening rather than only when symptoms arise. In an age-stratified analysis of the distribution of cytological abnormalities, we found that women older than 60 had the highest prevalence of epithelial abnormalities, possibly due to the limited organized screening programs in Egypt in younger women with accumulation of epithelial abnormalities over time. Our results indicate that women with ASC were significantly older than those with LSIL and younger than those with HSIL and glandular cells. Patients with LSILs and AGUS were significantly younger than patients with malignant squamous cell carcinoma, HSILs or adenocarcinoma. This was consistent with the findings of previously published results showing that the incidence of ASC, LSIL, and HSIL peaked in the 30–39-year age group, while the incidence of AGUS peaked among individuals aged 40–49 years. The incidence of malignant lesions further increased after the age of 50 years. The mean ages at diagnosis for patients with LSILs and HSILs were 34.7 and 37.7 years, respectively, while patients with malignant lesions presented with a mean age at diagnosis of 51.8 years . This might have implications for screening practice . Over a 42-year period, the number of screened women has decreased. This would explain the decrease in the number of SIL abnormalities from 6.9% to 4.3% for LSILs and from 1.2% to 0.13% for HSILs. This is because screening services in ECDUs currently operate on an opportunistic basis rather than through a structured, population-based approach. The effectiveness of structured, population-based preventive strategies in reducing the incidence rates of preinvasive and invasive lesions of the cervix cannot be overstated . However, in the vast majority of low- and middle-income countries (LMICs), including Egypt, researchers have identified a diverse array of obstacles to the process of screening . LMICs need prompt and immediate execution of unambiguous regulations, which should be fortified by the ability of the healthcare system to put these regulations into action. We need widespread advocacy within the community and the dissemination of information, alongside the strengthening of policies that promote the well-being of women and ensure gender equality. Overall, a four-decade study of cervical screening practice in a low resource setting, offers a wealth of information that can significantly contribute to the understanding and improvement of cervical cancer prevention and control. The extended timeframe allows for the collection of comprehensive data providing insights into trends and changes in cervical screening practices. Understanding how cervical screening practices have evolved over four decades can highlight the impact of policy changes and public health initiatives on screening rates. Highly competent pathologists assessed and confirmed cytologic findings, reducing the chances of incorrect classification of screening results. The results obtained in this study, when interpreted in the context of clinical practice, reveals the actual practice in the largest facility in a country of lower middle income and shows the urgent need to adopt a structured screening program and to incporporate state of the art tools for the diagnosis and management . The extensive dataset collected can serve as a valuable resource for future research, allowing for more detailed analyses and the exploration of new research questions. Limitations of this work includes missing data especially in certain demographic characteristics of women. This might reflect sensitive data in a conservative community or the lack of rigor or inconsistencies when collecting routine data. All demographic data were self-reported by women, which is prone to recall bias. The process of screening for cervical cancer is characterized by opportunistic practices rather than organized efforts. This implies that the subset of women who undergo screening differs from those who do not, as the former group has successfully surmounted various obstacles, such as financial constraints, social factors, cultural influences, and geographic limitations, to avail themselves of screening services. There is also a lack of information relating to the source of referral (e.g., self-referred or provider referred) for screening. The change in terminology used over time may have some implications for the assigned category of abnormality observed. The study lacks the findings on repeated smears for abnormal cytology. HPV testing became a state of the art in the screening and management . However, the current study did not examine how changes in the healthcare system, such as the introduction of new technologies (e.g., HPV testing), have affected cervical screening practices because of the lack of affordable and accessible and HPV testing nationwide. This four-decade longitudinal study could have highlighted the role of preventive measures, such as the HPV vaccine, in reducing the incidence of cervical cancer. Unfortunately, the vaccination coverage is unknown and there is no national vaccination program. The findings can inform public health policy and help in the allocation of resources to improve screening programs and implement effective preventive measures. The findings clearly show the need for a population-based national program if we are keen to meet the World Health Organization’s 2030 targets for cervical cancer elimination. |
Addressing bias and knowledge gaps regarding race and ethnicity in neonatology manuscript review | 2d9ec3df-971e-493c-b67b-129c2cba9583 | 9616711 | Pediatrics[mh] | In response to the murders of Breonna Taylor and George Floyd there has been a shift in public and academic attention to address racism, racial disparities, and health equity. Health equity is achieved when every person has the ability to attain their health potential. One of many major barriers to health equity include racism, or discrimination on the basis of one’s racial group. Racism can be individualized, internalized, and systemic and all forms contribute to racial disparities in health outcomes. Systemic racism is a form of racism that is embedded in laws, policies, and institutions, including academic medicine, that has resulted in a disparate distribution of goods, services, and opportunities for racial groups . Despite an abundance of calls for papers addressing health equity in major journals, the extent that reviewers and editors are adequately trained to critically evaluate the use of social constructs, like race, in research studies is highly variable. A major contributor to this knowledge deficit is the historical false belief in race as a biological construct by the scientific community and a paucity of published articles naming racism as a major driver of racial disparities . Omission of rigorous research standards for evaluating race and racism has contributed to harmful rhetoric such as the biologic fallacy of race . In addition to knowledge gaps by reviewers regarding the evaluation of social variables like race, explicit or implicit bias can occur in the manuscript review process , which may be more epitomized during peer review of articles focused on health equity that use social variables in their approach. For example, microaggressions are a form of discrimination defined as “slights” that communicate a negative attitude toward marginalized groups. Microaggressions disproportionately impact marginalized groups and are commonplace in the workforce’s daily lives; peer-review is no exception . In order to combat bias in reviews, scholars have suggested diversification of editorial boards, as well as intentional recruitment, education, and compensation of diverse pools of peer reviewers . Others have called for explicit standards for evaluation of race and ethnicity . In light of these concerns, more recently, some journals have established new author guidelines for addressing race and racism . However, standardized criteria have not been agreed upon or adopted for many academic journals. As health equity researchers in academic neonatology, we offer . Our own personal positive and negative experiences within the last two years that highlight knowledge gaps and bias in the peer-review process ; A brief summary of publicly available data from major, neonatal-focused journals regarding existing processes to evaluate health equity research and address bias in the review process; and Our recommendations for neonatology-focused journals regarding these aforementioned issues.
Knowledge gaps during peer-review “After submitting a study for peer review that tracked hospital practices by race/ethnicity and language, a reviewer argued there was no rationale as to why such disparities in hospital practices could exist and questioned why we chose to examine this. Denial of disparities in care quality by this reviewer suggested a significant knowledge gap of long-standing literature. I alerted my concern to the editor who omitted this review and sent it out to a different reviewer.” “When exploring the experience of traditionally marginalized communities in a qualitative study, a reviewer suggested that to increase the validity of the study, we should compare the experience to the majority’s experience. Centering whiteness and white normativity was problematic in a study designed to center at the margins . Following this comment, our team opted to include prose in the discussion about findings of previous studies focused on white populations. It was eventually published.” “As a peer reviewer I suggested capitalizing the ‘b’ in Black when identifying race and to not use ‘Blacks’ when referring to Black persons. The authors responded that they preferred not to edit for readability, despite the possibility of offensive interpretation and accepted terminology. 12 Upon re-review, the editor agreed and sent me a positive reply acknowledging the ‘teaching moment’ for the authors.” Bias in the peer-review process “I revealed my identity in a commentary and received an inappropriate comment during the review process. I did not know where to anonymously report my experience of discrimination to avoid worrying about my future relationship with the journal.”
“After submitting a study for peer review that tracked hospital practices by race/ethnicity and language, a reviewer argued there was no rationale as to why such disparities in hospital practices could exist and questioned why we chose to examine this. Denial of disparities in care quality by this reviewer suggested a significant knowledge gap of long-standing literature. I alerted my concern to the editor who omitted this review and sent it out to a different reviewer.” “When exploring the experience of traditionally marginalized communities in a qualitative study, a reviewer suggested that to increase the validity of the study, we should compare the experience to the majority’s experience. Centering whiteness and white normativity was problematic in a study designed to center at the margins . Following this comment, our team opted to include prose in the discussion about findings of previous studies focused on white populations. It was eventually published.” “As a peer reviewer I suggested capitalizing the ‘b’ in Black when identifying race and to not use ‘Blacks’ when referring to Black persons. The authors responded that they preferred not to edit for readability, despite the possibility of offensive interpretation and accepted terminology. 12 Upon re-review, the editor agreed and sent me a positive reply acknowledging the ‘teaching moment’ for the authors.”
“I revealed my identity in a commentary and received an inappropriate comment during the review process. I did not know where to anonymously report my experience of discrimination to avoid worrying about my future relationship with the journal.”
To better understand the extent that journals serving the academic neonatology audience have guidance regarding evaluation of social variables, like race and ethnicity, in articles and processes to address bias in peer-review, we examined the websites of 12 major academic journals that publish in neonatology. Journals were chosen by combining our searches of academic journals with high frequency of neonatal-perinatal material based on a PubMed query (currently utilized by neopapers, an automated literature Twitter account that has been created to publish recent articles with content related to neonatology and an active account in the #neoTwitter community ), and authors’ familiarity. Journal characteristics were created by the authors to evaluate previous commitment to health equity topics, transparency of diversity, equity, and inclusion issues, intention to diversify editorial staff, and existence of an anonymous system of reporting discrimination in peer review. No formal recommendations or regimented criteria exist to evaluate journal processes for inclusion of health equity content or bias in review, thus our evaluation metrics were developed through iterative discussion by authors and guidance from previous literature . We summarize findings in Table . We found that more than 75% of journals have published at least one original research, commentary, and perspective piece on health equity since journal conception, suggesting recognition of addressing social variables in neonatology journals. Regarding processes which may improve bias in peer-review more broadly, no journal had readily available data on racial, ethnic, or gender diversity of reviewers, editors, editorial board, but four (33%) had a statement of current efforts to diversify reviewers, editors, and/or editorial boards and only one (8%) journal provided information for how to apply to be an editorial board member on their website. Only one (8%) journal had a statement separate from Committee of Publication Ethics (COPE) guidelines for how to address bias in peer-review. COPE is an organization dedicated to providing resources and leadership on publication ethics which has recently published guidance on addressing bias in peer review. Although contact information was nearly always available for both editor-in-chief and members of the editorial board as a potential pathway to report discrimination (83% and 100% respectively), we could find no evidence of journals with a transparent system of anonymous peer review feedback to report racism, bias, or discrimination on their website.
Our anecdotal experience and review of publicly available data from journal websites suggest that there is room for improvement to address knowledge gaps in peer-review of neonatology articles focused on health equity, which often utilize social variables like race and ethnicity in their methodology and therefore may increase potential for bias in the peer-review process. With heightened national attention on the role of race, racism, and other social factors on health outcomes, we anticipate that research in this area will continue to grow. Therefore, journal guidelines for authors and reviewers are needed to educate the neonatal research community and set standards on use of race and racism in research. While our experiences focus primarily on the social construct of race and ethnicity, we believe that our experiences and our recommendations may impact those doing research in other domains that also utilize social variables such as income, primary language, and immigration status. Researchers also must be protected from discrimination and bias in the peer-review process. Few journals have made transparent efforts to diversify staff or develop mechanisms for providing anonymous feedback in the setting of perceived racism and discrimination in the review process. In our review, many journals have statements demonstrating commitment to adhere to COPE guidelines, which recently organized a Diversity Equity and Inclusion (DEI) committee that has provided resources and a commitment to addressing ethics and DEI for journals . A few journals we evaluated have also signed the joint commitment for action on inclusion and diversity in publishing, launched in June 2020 by the Royal Society of Chemistry with ongoing efforts to set minimum standards for inclusion and diversity for scholarly publishing. Planned efforts include, but are not limited to, setting minimum targets to achieve diverse representation of authors, reviewers, and editorial boards, developing language standards, reviewing and revising editor and editorial board member selection processes, and publicly reporting their progress . We are encouraged by the intention and progress made by several journals and publishing bodies, and hope to see fully transparent standards for DEI in the peer-review process across all neonatology publishing journals. We consider the diversification of reviewer, editorial boards, and editors to be of particular importance for the health equity publication process in neonatology journals. Not only does the inclusion of perspectives of lived experience and participation in scholarly health equity activities advance the quality of work in our field, it also begins to address historical exclusion of minoritized individuals from scientific discourse . In our field, there continues to be underrepresentation of minoritized trainees and physicians scientists . Harm during the peer review process can negatively impact the pursuit of antiracism and health equity work and disproportionately impacts minoritized researchers . Building infrastructure for transparency and accountability is necessary for ongoing publication of high quality health equity research . We hope that our recommendations on how to improve the peer-review process in neonatology journals can help improve the trust of neonatal researchers and mitigate systemic inequities in publication in research focused on health equity. Our review was limited to information readily available on journal websites. This may not fully encompass efforts made by academic journals to support health equity research and address bias in the peer-review process. The journal processes evaluated were designed by authors and thus are not previously validated and may not sufficiently evaluate the performance of journals. Our perspective piece does not compare the performance of neonatology journals, which tend to be clinically focused, to social science or public health focused journals. Performance in neonatology journals may be different from journals dedicated specifically to health equity. Regardless we see importance in addressing bias and knowledge gaps within our field while understanding challenges may be different or similar to other fields.
We offer the following recommendations to improve the peer-review process: A standard of proficiency of reviewers in evaluation of social variables and constructs, including race and racism. While there are a few resources available that address this topic , it is unclear what standard exists or should be followed among neonatology journals. At minimum, we recommend statements that confer that race is a social construct without biological basis and explicitly stating racism as a primary etiology of racial disparities. Transparency of current demographics of authors, reviewers, editorial boards, and editors. Although lack of diversity in academia is a widely known problem and we suspect it is no different in neonatology, the demographics of the participants in the peer-review process was not explicitly stated in the journals reviewed. Transparency offers a route towards accountability. Diversification of reviewers, editorial boards, and editors with transparent, publicly announced target dates and goals. Transparency, evaluation, and equal opportunity of editorial board selection process to facilitate diversification. Recruitment and appropriate compensation for subject matter experts for time. We are not aware of any current resources or guidelines that define subject matter expertise prior to review. In the case of health equity research, lived experience should be recognized as a form of subject matter expertise. Similarly, we are not aware of resources to guide overall reviewer recruitment nor compensation for reviewers by journals. If compensation for peer review is not provided by journals, institutions should consider ways in which to support faculty and trainees who participate in the peer-review process through financial incentives, promotion, or other forms of meaningful recognition. Standardized and robust training on an antiracism and the measurement and evaluation of social constructs in academic medicine and biomedical research that begins early and continues throughout professional careers. Anonymous reporting mechanism for authors to report racism and/or other types of bias in the peer-review process
|
ESMO Open: from Cancer Horizons to Science for Optimal Cancer Care—a tale stretching over 8 years | 6ff194c8-ef79-460c-9a61-0392aea9d405 | 10764984 | Internal Medicine[mh] | |
The Disruptions of Sphingolipid and Sterol Metabolism in the Short Fiber of | a5b5aa7b-39e9-4c8d-8c27-84da89ea7562 | 11818067 | Biochemistry[mh] | Cotton is an important natural fiber crop in the world. Cotton fiber is the major economic product of cotton and a predominant material in the textile industry. Fiber is a single cell derived from the epidermis of the outer integument of the cotton ovule with extreme elongation and significant thickening on the SCW (secondary cell wall) .The growth and development of fiber can be divided into four distinct and overlapping stages: initiation, elongation (primary wall formation), SCW deposition, and maturation. On the day of anthesis, a large number of globular protrusions were observed on the ovule surface, followed by polar elongation. Around 10 DPA (days post-anthesis), the fiber elongation rate reaches the peak, which is the rapid elongation stage of fibers . Fiber elongation lasts until 16–20 DPA. After about 15 DPA, fiber elongation gradually stops followed by the synthesis of the secondary cell wall, a stage which is also called the transition period from primary wall synthesis to secondary wall synthesis or the initiation of secondary wall synthesis . After 20 DPA, the elongation of the fiber cells stopped completely, and the cells entered into the stage of SCW deposition . The elongation and SCW synthesis stages are two important stages of fiber development, which determine fiber quality (length, strength, and fineness). Since the fiber is a single cell with an extremely elongated (final length up to 30–40 mm) and strikingly thickened secondary wall (cellulose content >90%), and the two actions last for a long period, it is regarded as an ideal material for studying cell elongation, cellulose synthesis, and SCW deposition . However, the regulatory mechanism of cotton fiber cell elongation and SCW formation still needs further exploration. Cotton fiber mutants are a powerful resource to study the regulatory mechanism of fiber cell development owing to the morphological and biochemical variances in their fiber cells. Ligon lintless ( li-1 ) is a simply inherited, monogenic dominant mutant characterized by very short and thick fibers (about 6 mm in length) . This presents an excellent model system for studying the underlying molecular and cellular processes involved in cotton fiber elongation . The fiber initiation and early elongation of li-1 was similar to that of its wild-type TM-1 (Texas marker-1). At 7 DPA, fiber elongation was inhibited and stopped completely at 14 DPA. During 7–14 DPA, the fiber elongation rate of the li-1 mutant was much lower than that of the wild type in the same period . Lipids are essential components of all plant cells, not only as the main component of cell membranes, but also as an important energy source and quality indicator . They are involved in the regulation of various life processes, such as transport, signaling, energy conversion, cell development and differentiation, and apoptosis . In plants, fatty acid signaling plays a crucial role in defense and development. These studies are of great significance to basic biology and agriculture. In both rapid elongation and SCW deposition, lipid synthesis is required for cotton fiber cell development . Previous studies reported that the transcript of genes involved in lipid synthesis such as fatty acid desaturase, acyl carrier protein, glycerol-3-phosphate acyltransferase, acyltransferase, lipid transfer proteins, and elongase are significantly enriched in fibers . Furthermore, a few studies have been conducted by metabolomics. Glycerides were detected in fibers and showed that polar lipid phosphatidylglycerol including PC (phosphatidylserine), PE (phosphatidylcholine), PI (phosphatidylinositol), PA (phosphatidic), monogalactosyldiacylglycerol (MGDG), and digalactosyldiacylglycerol (DGDG) had the highest content in elongation fibers (7–10 DPA) . Six glycerophospholipids (GPLs) were detected in the wild-type fast elongation fibers and ovules, as well as the lintless – fuzzless mutant ovules by targeted lipidomics. Phosphatidylinositol (PI) (containing the linolenic acid group) was significantly accumulated in the elongating fibers, indicating that PI plays a role in the elongation of fibers . The content of saturated VLCFA (very-long-chain fatty acids) in elongating fibers is significantly higher than that in ovules and lintless – fuzzless mutant ovules. The exogenous application of ACE, a fatty acid synthesis inhibitor, inhibited fiber elongation, while VLCFA promoted fiber elongation by induced ethylene synthesis . By analyzing the differences in metabonomics between the elongation stage and the secondary wall synthesis stage of fibers, the result showed that the lipid metabolism was active in the fiber elongation stage . These studies indicated that lipid metabolism plays important roles in the elongation and secondary wall synthesis of fibers. Sphingolipids are complex lipids that consist of three main components, long-chain fatty acids (LCFAs) or the very-long-chain fatty acids (VLCFAs), the long-chain base (LCB) of sphingosine, and the polar head group . Recently, a few documents revealed sphingolipid was essential for fiber growth and development. The exogenous application of FB1 (Fumonisin B1), a sphingolipid synthesis inhibitor, strongly inhibited fiber elongation and altered the activity of lipid raft activity in fiber cells . One kind of phytoceramide molecules containing hydroxylated and saturated VLCFA is important for fiber cell elongation . The contents of all GluCer (glycosylceramides) and GIPC (glycosyl inositol phospho ceramides) molecular species were decreased in 0-DPA ovules of Xuzhou142 lintless – fuzzless mutants and Xinxiangxiaoji lintless – fuzzless mutants when compared with the wild-type Xuzhou142 . Overexpressing GhCS1 , a ceramide synthase gene, inhibited fiber cell initiation and elongation . Since VLCFA is a composition of sphingolipid molecules, downregulating GhKCRL1 , a gene involved in the VLCFA biosynthesis pathway blocked sphingolipid synthesis and suppressed fiber cell elongation . Regulating GhLCBK1, a sphingosine kinase in cotton, could regulate fiber elongation and SCW deposition through affecting sphingosine-1-phophate and auxin synthesis . These reports indicated that sphingolipids play important roles in fiber cell development. However, the sphingolipid profile in the short fibers of the li-1 mutant is unknown. High-throughput lipid mass spectrometry is used to track metabolic changes and rapidly analyze the changes in individual lipid molecules in the wild type and mutant as well as various developmental stages. The untargeted lipidomics model can realize the systematic analysis of various types of lipids in the sample without bias. In order to clarify the lipid differences between the rapid elongation stage (10 DPA) and secondary cell wall synthesis stage (20 DPA) of fibers, as well as the lipid differences between the short fibers of the li-1 mutant and its wild-type normal fibers at the rapid elongation stage (10 DPA), and to reveal the function of lipids in fiber development, firstly, we analyzed the lipid changes in the 10-DPA and 20-DPA fiber cells of wild-type (TM-1), as well as the lipid differences in 10-DPA fibers between the TM-1 and li-1 mutant, through untargeted lipidomics assay. On this basis, further analysis was conducted on the differences in sphingolipids and sterols in 10-DPA fibers between the TM-1 and li-1 mutant through targeted lipidomics assay. The results indicated that the disruptions of sphingolipid and sterol metabolism may be an important cause for the hindered elongation of li-1 short fibers. This study provides a new clue for further analyzing the regulatory mechanisms of fiber growth and development.
2.1. Untargeted Lipidomics Analysis in Fiber Cells To identify the lipid differences in cotton fiber cells at the rapid elongation stage and secondary wall synthesis stage, as well as the lipid differences in fibers of the li-1 mutant, the OPLS-DA (Orthogonal Partial Least Squares Discriminant Analysis) model was used. Both R2Y and Q2 are the evaluation parameters of the model. As shown in the score plot ( A,B), there were six scores in each group and two groups were clearly separated. In the model, R 2 Y = 0.994 and Q 2 = 0.980 in the group of the 10-DPA fibers and 20-DPA fibers of the wild type, and R 2 Y = 0.949 and Q 2 = 0.870 in the group of the 10-DPA fibers of the li-1 mutant and 10-DPA fibers of the wild type, indicating that the quality of the OPLS-DA model was excellent to screen the key lipids between two samples. In total, 7 lipid classes (glycospholipids, sphingolipids, glycolipids, sterol esters, prenol lipids, fatty acyls, and saccharolipids) were identified in fiber cells including 33 lipid sub-classes and 793 lipid molecular species ( C). Sphingolipids comprised seven lipid sub-classes and 197 lipid molecule species (129 ceramides, 3 phosphoceramides, 6 sphingosine, 1 phytosphingosine, 1 sphingomyelin, 54 glucosylceramides, and 3 disaccharide ceramides). Glycerophospholipids included 11 lipid sub-classes and 313 lipid molecule species; the number of molecular species in each sub-class was 10 CL, 1 LPA, 16 LPC, 15 LPE, 1 LPI, 80 PA, 66 PC, 48 PE, 23 PG, 41 PI, and 12 PS. Glycerolipids had three lipid sub-classes: MG, DG, and TG; the number of molecular species was 1, 73, and 128, respectively. Three lipid sub-classes of fatty acyls were detected such as FA, OAHFA, and WE; the number of molecular species was 4, 1, and 3, respectively. Five lipid sub-classes of saccharolipids, DGDG, DGMG, MGDG, SQDG, and MGMG, were detected; the number of molecular species was 23, 1, 15, 7, and 2, respectively. Sterol esters included three lipid sub-classes: AGlcSiE, SiE, and StE; the number of molecular species was 17, 2, and 1, respectively. One lipid class Co (comprised 5 lipid molecular species) was detected in fiber cells ( C). In the 10-DPA fiber cells, the lipid sub-classes PA, PC, PE, PI, Cer, CerG1, DG, TG, and AGlcSiE possessed a higher lipid intensity while the lipid sub-classes LPI, Mg, DGMG, MGMG, and StE had a lower lipid intensity ( and ). The results indicated that glycerophspholipid, glycerolipid, and sphingolipid are predominant lipids and contained more kinds of lipid molecule species in fiber cells. 2.2. The Lipid Difference Between the Stage of Rapid Elongation and SCW Deposition of Fiber Cells The rapid elongation period of fiber cells is around 10 DPA. After 15 DPA, the SCW of fiber begins to synthesize and a large amount of cellulose is deposited, while cell elongation stops gradually. Around 20 DPA, fiber cells ceased elongation completely. In order to reveal the lipid difference in various developmental stages of fiber, we analyzed the lipid difference between the 10-DPA fibers and 20-DPA fibers from TM-1. The results are shown in ( A). Compared with the 10-DPA fibers, 13 lipid sub-classes increased and 20 lipid sub-classes decreased in the 20-DPA fibers. By statistical analysis, 23 lipid classes changed significantly ( p < 0.05). Among them, seven lipid sub-classes increased such as MGMG, SiE, MG, and so on. Meanwhile, 16 lipid sub-classes decreased and the top 5 lipid sub-classes with the most decrease were So, TG, LPC, PG, and CerG2. The results indicated that the lipid intensity of various lipid classes significantly changed between the rapid elongation and SCW synthesis stages, and the lipids were majorly enriched in the rapid elongation fiber cells. The detailed lipid profile was further analyzed . Compared with the 10-DPA fibers, 82 lipid molecule species significantly changed in the 20-DPA fibers (VIP > 1 and p < 0.05). Among them, the lipid intensity of 26 lipid molecule species increased and 56 lipid molecule species decreased ( B and ). The top five lipid molecule species with the largest increase in lipid intensity were PA (15:0/18:3), TG (38:2), CerG1 (d32:3), CER (d18:0/23:0), and CER (d18:0/25:0), and their increase folds were 22.54, 13.92, 10.63, 6.87, and 6.72, respectively. The top five lipid molecule species with the largest decrease folds were TG (20:1/18:2/18:3), TG (19:1/18:3/18:4), TG (18:3/18:3), TG (19:1/18:2/18:3), and TG (16:1/18:3/18:3), which decreased by 34.48-, 32.26-, 22.73-, 16.67-, and 15.15-fold, respectively ( B). The lipid molecule species with a significant change in lipid intensity mainly belonged to glycerophspholipid (57%), glycerolipid (27%), and sphingolipid (12%), and were mainly involved in the lipid sub-classes of PA (22%), TG (21%), PC (15%), and PE (11%) ( C,D). The results indicated that most lipid molecule species of GP, GL, and SP were strikingly enriched in the rapid elongation stage, and a few lipid molecule species were enriched at the stage of SCW deposition, which might play some roles in SCW formation. 2.3. The Lipid Change in Fiber Cells of li-1 Mutant The li-1 mutant is a super-short fiber mutant with less than 6 mm length fibers, which is an ideal model for studying fiber development . To explore the role of lipids in cotton fiber elongation, we identified the lipid changes between the li-1 mutant and its wild type (TM-1) in the 10-DPA fibers. As shown in ( A), compared with the 10-DPA fibers of TM-1, 23 lipid sub-classes increased and 10 lipid sub-classes decreased in the li-1 mutant, and 25 of 33 lipid sub-classes changed significantly ( p < 0.05). Among them, 20 lipid sub-classes increased and the top 5 lipid sub-classes (fold change) with the largest increase were SiE (64.37), StE (3.99), TG (2.46), OAHFA (2.06), and CL (1.94), while only 5 lipid sub-classes decreased, and the top 5 with the largest decrease were SM (3.13), FA (1.85), CerG2 (1.67), LPE (1.64), and LPC (1.37). The results indicated that the lipid intensity of most lipid classes increased significantly in the 10-DPA fiber cells of the li-1 mutant compared with the 10-DPA fiber cells of the wild type. Compared with the 10-DPA fibers of TM-1, the lipid intensity of 68 lipid molecule species changed significantly in the 10-DPA fibers of the mutant (VIP > 1 and p < 0.05). Among them, 60 lipid molecule species increased and 8 decreased ( B and ). The top five lipid molecule species with the largest increase were SiE (18:3), TG (18:1/18:1/18:1), TG (16:0/18:1/18:1), TG (16:0/16:0/18:1), and TG (18:0/18:1/18:3), and their increase folds were 59.56, 19.94, 19.10, 13.64, and 8.99, respectively. The top five lipid molecule species with the largest decrease in folds were PLE (18:3), PLC (16:0), PLC (18:3), CerG1 (d18:2/18:0), and CerG2 (d34:4), and their decreased folds were 3.04-, 2.64-, 2.57-, 2.42-, and 1.73-fold, respectively ( B). The lipid molecule species with a significant change in lipid intensity mainly belonged to glycerolipid (43%) and glycerophspholipid (35%), and mainly involved the lipid sub-classes of TG (38%) and PA (15%) ( C,D). The results indicated that most lipid molecule species of TG and GP were strikingly enriched in the 10-DPA fibers of the li-1 mutant. Interestingly, the lipid molecule species of SiE (18:3) increased by 59.56-fold, indicating it may be a key factor contributing to short fibers. 2.4. The Changes in Sphingolipids in li-1 Mutant Fiber Cells The results of the untargeted lipidomics showed significant changes in sphingolipids in the mutant fiber cells compared with the wild-type fibers. To further illuminate the detailed changes in sphingolipid composition and content, we detected the sphingolipid profile of the 10-DPA fiber cells from the li-1 mutant and TM-1 by means of UHPLC–MS/MS. The results showed that six sub-classes of sphingolipids and 69 molecular species of sphingolipids were detected ( A), including sphingosines (Sph), sphingosine-1-phosphate (S1P), ceramides (Cer), phytoceramides (PhytoCer), glucosylceramides (GluCer), and glycosyl inositol phospho ceramides (GIPC); the number of molecular species was 4, 2, 14, 31, 13, and 5 for each class, respectively. Compared with the wild type, the content of the Sph (PhytoSph and Sph) and GIPC classes decreased. Meanwhile, PhytoCer-OHFA, GluCer, and Phyto-GluCer increased ( B). Further analysis revealed that the content of all molecular species of sphingosine and GIPC decreased ( C,F and ). On the contrary, the content of all molecular species of GluCer (Phyto-GluCer and GluCer) and PhytoCer OHFA increased ( G,H and ). These results indicated that the sphingolipid profile was significantly altered in the li-1 fibers. It is suggested that GIPC synthesis was impaired, which resulted in all GIPCs decreasing and all GluCer and Cers increasing. 2.5. The Changes in Phytosterols in li-1 Mutant Fiber Cells In the untargeted lipidomics analysis, the content of sitosteryl ester (SiE) was 59.56 times higher in the li-1 mutant fibers than in the TM-1 fibers, indicating a striking change in steryl esters occurred in the mutant fiber cells. To further detail the changes in phytosterols in the mutant, we detected the phytosterol profile in both the li-1 and TM-1 10-DPA fibers. Five sterols (cholesterol, sitosterol, campesterol, stigmasterol, and stigmastanol) and two steryl esters were detected in the 10-DPA fiber cells ( A). Compared with the wild type, the total sterol and all sub-classes of sterols were increased in the mutant fibers, among which, the cholesterol, sitosterol, and stigmastanol were increased by 85%, 36%, and 31%, respectively ( A). Meanwhile, compared to the wild type, the mutant fiber cells showed a striking increase in campesteryl ester, sitosteryl ester, and total steryl ester, with 21.8-, 48.7-, and 45.5-fold increases, respectively ( B). Given the proportion between sitosterol, campesterol, and stigmasterol is important for sterol function in plant development, we analyzed the ratio of campesterol to sitosterol (C/S) and stigmasterol to sitosterol (St/Si). The result showed that the ratios of C/S and St/Si were declined in the fibers of the li-1 mutant ( C), which might result from the content of sitosterol increasing greatly in the li-1 fiber cells. These results indicated that the contents of sterol and steryl ester were altered dramatically in the li-1 mutant fiber cells. 2.6. The Expression of Genes Involved in Lipid Metabolism Was Altered in li-1 Fiber Cells To clarify the metabolic pathways and key genes that undergo significant changes in the mutant short fiber cells, we analyzed the transcriptome of the 10-DPA fiber cells of the li-1 mutant and its wild-type TM-1. The results showed significant changes in the expression of 8180 genes in the li-1 mutant 10-DPA fibers, with 5600 genes upregulated and 3580 genes downregulated . The results of the GO annotations analysis indicated that the differentially expressed genes were enriched in the biological process (catalytic activity and binding), cellular component (membrane part and membrane), and molecular function (cellular process and metabolic process) . The results of the KEGG enrichment analysis revealed that the differentially expressed genes were enriched in lipid metabolism pathways, including fatty acid association and degradation, glycerolipid and glycerophospholipid metabolism, Glycosphingolipid biosynthesis, and Brassinosteroid biosynthesis, 7 out of the top 20 metabolic pathways . These results illuminated that the lipid metabolism was disrupted in the li -1 mutant fiber cells compared with the TM-1. 2.7. The Expression Levels of Key Genes in Lipid Metabolism Were Elevated in li-1 Fiber Cells To confirm the expression change in genes involved in the lipid metabolism, we detected the expression levels of selected genes in the li-1 and TM-1 fibers by RT-qPCR. Gh_D12G0217 (LAG1 longevity assurance homolog 2), Gh_A07G0513 (LAG1 longevity assurance homolog 3), and Gh_D10G0211 (Lactosylceramide 4-alpha-galactosyltransferase) were involved in sphingolipid biosynthesis and were downregulated in the li-1 mutant . Gh_A08G1600 and Gh_A06G0144 encoded phospholipase D beta 1 and phospholipase A 2A, respectively. They play a role in phospholipid metabolism and were upregulated in the li-1 fiber cells. Gh_A02G0884 (GDSL-domain protein) and Gh_D06G2376 (3-ketoacyl-CoA synthase 19) decreased in the li-1 mutant. Gh_D03G1074 and Gh_A05G3810 encoded two HXXXD-type acyl-transferase family proteins, which were increased in the li-1 mutant. Gh_A01G1605 (alcohol dehydrogenase 1) might function in fatty acid degradation and was strikingly increased in the li-1 mutant . The result indicated the expression levels of genes involved in lipid biosynthesis and degradation were changed greatly, suggesting that the lipid metabolism was disrupted in the li-1 mutant fiber cells. 2.8. The Number of Oil Bodies Was Increased in li-1 Leaf and Fiber Cells As mentioned earlier, the content of triglycerides (TG) and steryl esters is far higher in the li-1 fibers than in the TM-1 fibers. Given that the hydrophobic core of the oil body (lipid droplet) is mainly composed of TG and steryl esters , we further detected the distribution and quantity of oil bodies in the mutant leaves and fiber cells by Nile red staining. The results showed that only a few small oil bodies were present in the TM-1 leaves and fiber cells while a great number of oil bodies were present in the li-1 leaves and fiber cells . This result further confirmed that the lipid metabolism was disrupted in the li-1 mutant.
To identify the lipid differences in cotton fiber cells at the rapid elongation stage and secondary wall synthesis stage, as well as the lipid differences in fibers of the li-1 mutant, the OPLS-DA (Orthogonal Partial Least Squares Discriminant Analysis) model was used. Both R2Y and Q2 are the evaluation parameters of the model. As shown in the score plot ( A,B), there were six scores in each group and two groups were clearly separated. In the model, R 2 Y = 0.994 and Q 2 = 0.980 in the group of the 10-DPA fibers and 20-DPA fibers of the wild type, and R 2 Y = 0.949 and Q 2 = 0.870 in the group of the 10-DPA fibers of the li-1 mutant and 10-DPA fibers of the wild type, indicating that the quality of the OPLS-DA model was excellent to screen the key lipids between two samples. In total, 7 lipid classes (glycospholipids, sphingolipids, glycolipids, sterol esters, prenol lipids, fatty acyls, and saccharolipids) were identified in fiber cells including 33 lipid sub-classes and 793 lipid molecular species ( C). Sphingolipids comprised seven lipid sub-classes and 197 lipid molecule species (129 ceramides, 3 phosphoceramides, 6 sphingosine, 1 phytosphingosine, 1 sphingomyelin, 54 glucosylceramides, and 3 disaccharide ceramides). Glycerophospholipids included 11 lipid sub-classes and 313 lipid molecule species; the number of molecular species in each sub-class was 10 CL, 1 LPA, 16 LPC, 15 LPE, 1 LPI, 80 PA, 66 PC, 48 PE, 23 PG, 41 PI, and 12 PS. Glycerolipids had three lipid sub-classes: MG, DG, and TG; the number of molecular species was 1, 73, and 128, respectively. Three lipid sub-classes of fatty acyls were detected such as FA, OAHFA, and WE; the number of molecular species was 4, 1, and 3, respectively. Five lipid sub-classes of saccharolipids, DGDG, DGMG, MGDG, SQDG, and MGMG, were detected; the number of molecular species was 23, 1, 15, 7, and 2, respectively. Sterol esters included three lipid sub-classes: AGlcSiE, SiE, and StE; the number of molecular species was 17, 2, and 1, respectively. One lipid class Co (comprised 5 lipid molecular species) was detected in fiber cells ( C). In the 10-DPA fiber cells, the lipid sub-classes PA, PC, PE, PI, Cer, CerG1, DG, TG, and AGlcSiE possessed a higher lipid intensity while the lipid sub-classes LPI, Mg, DGMG, MGMG, and StE had a lower lipid intensity ( and ). The results indicated that glycerophspholipid, glycerolipid, and sphingolipid are predominant lipids and contained more kinds of lipid molecule species in fiber cells.
The rapid elongation period of fiber cells is around 10 DPA. After 15 DPA, the SCW of fiber begins to synthesize and a large amount of cellulose is deposited, while cell elongation stops gradually. Around 20 DPA, fiber cells ceased elongation completely. In order to reveal the lipid difference in various developmental stages of fiber, we analyzed the lipid difference between the 10-DPA fibers and 20-DPA fibers from TM-1. The results are shown in ( A). Compared with the 10-DPA fibers, 13 lipid sub-classes increased and 20 lipid sub-classes decreased in the 20-DPA fibers. By statistical analysis, 23 lipid classes changed significantly ( p < 0.05). Among them, seven lipid sub-classes increased such as MGMG, SiE, MG, and so on. Meanwhile, 16 lipid sub-classes decreased and the top 5 lipid sub-classes with the most decrease were So, TG, LPC, PG, and CerG2. The results indicated that the lipid intensity of various lipid classes significantly changed between the rapid elongation and SCW synthesis stages, and the lipids were majorly enriched in the rapid elongation fiber cells. The detailed lipid profile was further analyzed . Compared with the 10-DPA fibers, 82 lipid molecule species significantly changed in the 20-DPA fibers (VIP > 1 and p < 0.05). Among them, the lipid intensity of 26 lipid molecule species increased and 56 lipid molecule species decreased ( B and ). The top five lipid molecule species with the largest increase in lipid intensity were PA (15:0/18:3), TG (38:2), CerG1 (d32:3), CER (d18:0/23:0), and CER (d18:0/25:0), and their increase folds were 22.54, 13.92, 10.63, 6.87, and 6.72, respectively. The top five lipid molecule species with the largest decrease folds were TG (20:1/18:2/18:3), TG (19:1/18:3/18:4), TG (18:3/18:3), TG (19:1/18:2/18:3), and TG (16:1/18:3/18:3), which decreased by 34.48-, 32.26-, 22.73-, 16.67-, and 15.15-fold, respectively ( B). The lipid molecule species with a significant change in lipid intensity mainly belonged to glycerophspholipid (57%), glycerolipid (27%), and sphingolipid (12%), and were mainly involved in the lipid sub-classes of PA (22%), TG (21%), PC (15%), and PE (11%) ( C,D). The results indicated that most lipid molecule species of GP, GL, and SP were strikingly enriched in the rapid elongation stage, and a few lipid molecule species were enriched at the stage of SCW deposition, which might play some roles in SCW formation.
The li-1 mutant is a super-short fiber mutant with less than 6 mm length fibers, which is an ideal model for studying fiber development . To explore the role of lipids in cotton fiber elongation, we identified the lipid changes between the li-1 mutant and its wild type (TM-1) in the 10-DPA fibers. As shown in ( A), compared with the 10-DPA fibers of TM-1, 23 lipid sub-classes increased and 10 lipid sub-classes decreased in the li-1 mutant, and 25 of 33 lipid sub-classes changed significantly ( p < 0.05). Among them, 20 lipid sub-classes increased and the top 5 lipid sub-classes (fold change) with the largest increase were SiE (64.37), StE (3.99), TG (2.46), OAHFA (2.06), and CL (1.94), while only 5 lipid sub-classes decreased, and the top 5 with the largest decrease were SM (3.13), FA (1.85), CerG2 (1.67), LPE (1.64), and LPC (1.37). The results indicated that the lipid intensity of most lipid classes increased significantly in the 10-DPA fiber cells of the li-1 mutant compared with the 10-DPA fiber cells of the wild type. Compared with the 10-DPA fibers of TM-1, the lipid intensity of 68 lipid molecule species changed significantly in the 10-DPA fibers of the mutant (VIP > 1 and p < 0.05). Among them, 60 lipid molecule species increased and 8 decreased ( B and ). The top five lipid molecule species with the largest increase were SiE (18:3), TG (18:1/18:1/18:1), TG (16:0/18:1/18:1), TG (16:0/16:0/18:1), and TG (18:0/18:1/18:3), and their increase folds were 59.56, 19.94, 19.10, 13.64, and 8.99, respectively. The top five lipid molecule species with the largest decrease in folds were PLE (18:3), PLC (16:0), PLC (18:3), CerG1 (d18:2/18:0), and CerG2 (d34:4), and their decreased folds were 3.04-, 2.64-, 2.57-, 2.42-, and 1.73-fold, respectively ( B). The lipid molecule species with a significant change in lipid intensity mainly belonged to glycerolipid (43%) and glycerophspholipid (35%), and mainly involved the lipid sub-classes of TG (38%) and PA (15%) ( C,D). The results indicated that most lipid molecule species of TG and GP were strikingly enriched in the 10-DPA fibers of the li-1 mutant. Interestingly, the lipid molecule species of SiE (18:3) increased by 59.56-fold, indicating it may be a key factor contributing to short fibers.
The results of the untargeted lipidomics showed significant changes in sphingolipids in the mutant fiber cells compared with the wild-type fibers. To further illuminate the detailed changes in sphingolipid composition and content, we detected the sphingolipid profile of the 10-DPA fiber cells from the li-1 mutant and TM-1 by means of UHPLC–MS/MS. The results showed that six sub-classes of sphingolipids and 69 molecular species of sphingolipids were detected ( A), including sphingosines (Sph), sphingosine-1-phosphate (S1P), ceramides (Cer), phytoceramides (PhytoCer), glucosylceramides (GluCer), and glycosyl inositol phospho ceramides (GIPC); the number of molecular species was 4, 2, 14, 31, 13, and 5 for each class, respectively. Compared with the wild type, the content of the Sph (PhytoSph and Sph) and GIPC classes decreased. Meanwhile, PhytoCer-OHFA, GluCer, and Phyto-GluCer increased ( B). Further analysis revealed that the content of all molecular species of sphingosine and GIPC decreased ( C,F and ). On the contrary, the content of all molecular species of GluCer (Phyto-GluCer and GluCer) and PhytoCer OHFA increased ( G,H and ). These results indicated that the sphingolipid profile was significantly altered in the li-1 fibers. It is suggested that GIPC synthesis was impaired, which resulted in all GIPCs decreasing and all GluCer and Cers increasing.
In the untargeted lipidomics analysis, the content of sitosteryl ester (SiE) was 59.56 times higher in the li-1 mutant fibers than in the TM-1 fibers, indicating a striking change in steryl esters occurred in the mutant fiber cells. To further detail the changes in phytosterols in the mutant, we detected the phytosterol profile in both the li-1 and TM-1 10-DPA fibers. Five sterols (cholesterol, sitosterol, campesterol, stigmasterol, and stigmastanol) and two steryl esters were detected in the 10-DPA fiber cells ( A). Compared with the wild type, the total sterol and all sub-classes of sterols were increased in the mutant fibers, among which, the cholesterol, sitosterol, and stigmastanol were increased by 85%, 36%, and 31%, respectively ( A). Meanwhile, compared to the wild type, the mutant fiber cells showed a striking increase in campesteryl ester, sitosteryl ester, and total steryl ester, with 21.8-, 48.7-, and 45.5-fold increases, respectively ( B). Given the proportion between sitosterol, campesterol, and stigmasterol is important for sterol function in plant development, we analyzed the ratio of campesterol to sitosterol (C/S) and stigmasterol to sitosterol (St/Si). The result showed that the ratios of C/S and St/Si were declined in the fibers of the li-1 mutant ( C), which might result from the content of sitosterol increasing greatly in the li-1 fiber cells. These results indicated that the contents of sterol and steryl ester were altered dramatically in the li-1 mutant fiber cells.
To clarify the metabolic pathways and key genes that undergo significant changes in the mutant short fiber cells, we analyzed the transcriptome of the 10-DPA fiber cells of the li-1 mutant and its wild-type TM-1. The results showed significant changes in the expression of 8180 genes in the li-1 mutant 10-DPA fibers, with 5600 genes upregulated and 3580 genes downregulated . The results of the GO annotations analysis indicated that the differentially expressed genes were enriched in the biological process (catalytic activity and binding), cellular component (membrane part and membrane), and molecular function (cellular process and metabolic process) . The results of the KEGG enrichment analysis revealed that the differentially expressed genes were enriched in lipid metabolism pathways, including fatty acid association and degradation, glycerolipid and glycerophospholipid metabolism, Glycosphingolipid biosynthesis, and Brassinosteroid biosynthesis, 7 out of the top 20 metabolic pathways . These results illuminated that the lipid metabolism was disrupted in the li -1 mutant fiber cells compared with the TM-1.
To confirm the expression change in genes involved in the lipid metabolism, we detected the expression levels of selected genes in the li-1 and TM-1 fibers by RT-qPCR. Gh_D12G0217 (LAG1 longevity assurance homolog 2), Gh_A07G0513 (LAG1 longevity assurance homolog 3), and Gh_D10G0211 (Lactosylceramide 4-alpha-galactosyltransferase) were involved in sphingolipid biosynthesis and were downregulated in the li-1 mutant . Gh_A08G1600 and Gh_A06G0144 encoded phospholipase D beta 1 and phospholipase A 2A, respectively. They play a role in phospholipid metabolism and were upregulated in the li-1 fiber cells. Gh_A02G0884 (GDSL-domain protein) and Gh_D06G2376 (3-ketoacyl-CoA synthase 19) decreased in the li-1 mutant. Gh_D03G1074 and Gh_A05G3810 encoded two HXXXD-type acyl-transferase family proteins, which were increased in the li-1 mutant. Gh_A01G1605 (alcohol dehydrogenase 1) might function in fatty acid degradation and was strikingly increased in the li-1 mutant . The result indicated the expression levels of genes involved in lipid biosynthesis and degradation were changed greatly, suggesting that the lipid metabolism was disrupted in the li-1 mutant fiber cells.
As mentioned earlier, the content of triglycerides (TG) and steryl esters is far higher in the li-1 fibers than in the TM-1 fibers. Given that the hydrophobic core of the oil body (lipid droplet) is mainly composed of TG and steryl esters , we further detected the distribution and quantity of oil bodies in the mutant leaves and fiber cells by Nile red staining. The results showed that only a few small oil bodies were present in the TM-1 leaves and fiber cells while a great number of oil bodies were present in the li-1 leaves and fiber cells . This result further confirmed that the lipid metabolism was disrupted in the li-1 mutant.
3.1. The Role of Lipids in Fiber Elongation and SCW Deposition The growth and development of fiber can be divided into four distinct and overlapping stages: initiation, elongation (primary wall formation), SCW deposition, and maturation. Following initiation (0~2 DPA), the fiber cells start to elongate. During the fast elongation period (8~12 DPA), the cell size and membrane area expand rapidly, and various metabolic activities need to be increased accordingly. This process requires a lot of lipids . At the stage of SCW deposition (15~40 DPA), cell expansion stopped and cellulose synthesis, transportation, and deposition proceeded steadily. The cellulose synthase complex is located in the plasma membrane. The lipid composition and membrane features at this stage are conducive to the synthesis and organization of cellulose . In this study, we analyzed the difference in lipid groups between the fast elongation stage (10 DPA) and the SCW formation stage of fibers (20 DPA). Most of the lipid components (16/23 lipid sub-classes and 56/82 lipid molecular species) were higher in the elongating fibers, and only a few lipid components (7/23 lipid sub-classes and 26/82 lipid molecular species) were higher in the SCW formation stage. Among them, glycerophospholipids (GP), sphingolipids (SP), and glycerolipids (GL) were enriched in the elongated fibers, which was similar to previous studies. It was reported that the content of polar lipids such as PC, PE, PI, PA, and PG was the highest in the elongating fibers (7–10 DPA) . These results suggested that the accumulation of these lipids may be necessary for fiber elongation. The lipids of the 10-DPA fiber and ovule of the wild type, and the 10-DPA ovule of the fuzzless – lintless mutant were detected by targeted lipidomics. The result showed that phosphatidylinositol (PI) was enriched in fiber cells and PI (34:3) was the highest lipid molecule species . Consistently, in our study, the lipid intensity of PI and molecular species PI (18:3/18:3) and PI (16:0/18:3) molecules was significantly higher in the 10-DPA fiber cells than in the 20-DPA fiber cells of TM-1. The PI intensity of the 10-DPA fibers from the li-1 mutant was also significantly higher than that of the 10-DPA fibers from the wild-type. However, the increased PI molecules were PI (16:0/18:1), PI (16:0/18:2), and PI (18:2/18:2), which indicated that the PI molecules enriched in the mutant 10-DPA fibers were different from those enriched in the wild-type 10-DPA fibers. Phospholipase D (PLD) can hydrolyze phospholipid to phospholipid acid (PA). GhPLDα1 was highly expressed in the 20-DPA fibers, which may lead to a decrease in the phospholipid content and an increase in PA in 20-DPA fibers . Consistently, most PA was enriched in the 20-DPA fibers. On the other hand, phospholipid acid can promote the production of hydrogen peroxide and induce the synthesis of SCW . These results indicate that phospholipids play some roles in the regulation of fiber elongation and SCW synthesis. The intensity of saccharlipid (SL), sterol ester (ST), and wax ester (WE) was enriched at the stage of SCW deposition, which indicated that the enrichment of these lipid components may promote cellulose synthesis and SCW formation. Phytosterols play important roles in the development of fiber cells . Sterols are comprised of free sterols and conjugated sterols such as sitosterol ester (SiE), campesterol ester, stigmasterol (StE) ester, sterol glycosides (SGS), and acetyl sterol glycosides (ASGS). Conjugated sterols play a role in the dynamic balance of sterols and in the synthesis of WEs . Three lipid classes of steryl esters, SiE, AGlcSiE, and StE, were detected in the study, among which, the AGlcSiE intensity was the highest in the rapid elongating fibers, and the intensity of SiE significantly increased in the stage of SCW synthesis. It was suggested that SiE might be associated with cellulose synthesis and SCW formation. Sterol glycosides and acetylsterol glycosides are enriched in elongating fibers. They may play roles in maintaining the balance between sterols and sphingolipids in fiber cell elongation . Sterols and sphingolipids are two key components of lipid rafts, which are functional regions of membranes . Sphingolipids are necessary for plant growth and development, in response to biotic stress or abiotic stress . Sphingolipids also played a role in cotton fiber elongation. FB1 (Fumonisin B1), an inhibitor of sphingolipid synthesis, significantly inhibited fiber elongation and altered the activity of lipid raft activity in fiber cells . In this study, CerG1, CerG2, CerG1 (d18:2/18:0), CerG1 (d18:2/18:1), CerG1 (d36:2), and CerG2 (d34:4) are enriched in the elongating fiber cells. Moreover, these molecules are closely associated with the AGlcSiE lipid molecule , which indicated that these molecules play a role in the regulation of fiber cell elongation. On the other hand, VLCFA is a component of sphingolipid molecules and plays an important role in fiber cell elongation. Application of ACE inhibited fiber elongation while VLCFA promoted cell elongation by promoting ethylene synthesis . These results indicated that there is a close relationship between sphingolipids, sterol, and membrane lipid rafts during the process of fiber elongation. The future study of these relationships may be an important aspect to reveal the regulatory mechanism in the growth and development of fibers. 3.2. The Lipid Metabolism Disruption in the li-1 Mutant Fibers Ligon lintless-1 ( li-1 ) is a dominant mutant of Gossypium hirsutum , which has the phenotype of damaged vegetative growth and short and thick fibers. Although the gene identification and gene expression profile of the fibers of the li-1 mutant have been studied for many years, the regulatory mechanism of the fiber growth deficiency is still unclear . Compared with the wild type, unexpectedly, most lipid classes and lipid molecule species have a higher rather than lower intensity in mutant fibers (20/25 lipid species, 60/68 lipid molecular species), and only 5 lipid species and 8 lipid molecule species are lower in mutant fibers. In the fibers of the li-1 mutant, the lipid classes and lipid molecule species of TG and SL were enriched. Consistently, GL and SL were enriched in the stage of SCW synthesis in the wild-type fibers. The high intensity of GL and SL in the li-1 fiber may promote SCW synthesis and inhibit fiber elongation. Consequently, the li-1 fiber is shorter and thicker than the wild-type fiber. Sphingolipids and GPs were the major lipids in li-1 cells. LPC and PLE were enriched in the 10-DPA fibers during fiber development in the wild type but decreased in the li-1 fiber cells. These two GPs may play a key role in fiber elongation. FB1 is an inhibitor of ceramide synthase that is the center of sphingolipid metabolism and biology . The exogenous application of FB1 strongly inhibited fiber cell growth and its fiber phenotype is similar to that of the li-1 fiber . FB1 treatment resulted in a decrease in the total GIPC and all GIPC molecular species . Consistent with the FB1 treatment, the total GIPC and all molecular species were reduced in the li-1 mutant fibers ( F). Furthermore, the expression levels of two genes encoding ceramide synthases Gh_D12G0217 and Gh_A07G0513 were significantly reduced . These results revealed that GIPC synthesis was inhibited in the li-1 fibers and GIPC might play an important role in fiber cell elongation. There are 10 homologous genes encoding ceramide synthase in the upland cotton genome. In our previous studies, overexpression of GhCS1 (Gh_D07G0583) promoted the synthesis of ceramide molecules containing dihydroxy LCB and VLCFA and inhibited the initiation and elongation of fiber cells . Consistently, the most ceramide molecules containing dihydroxy LCB and VLCFA were significantly increased in the li-1 mutant fibers ( D). Taken together, given sphingolipids are regarded as major regulators of lipid metabolism , sphingolipid balance might be an important factor for the normal growth of fiber cells. During fiber development, SiE was enriched in the 20-DPA fibers while acetylglycosyl sterol ester and StE were enriched in the 10-DPA fibers. However, the lipid intensity of three sterol esters in the mutant fiber was higher than that in the wild-type fiber, and the intensity of the SiE molecule (SiE (18:3)) was 59.56 times that in the wild-type 10-DPA fibers. This change was further confirmed by the targeted lipidomics of phytosterols. Compared with TM-1, the contents of campesteryl ester, sitosteryl ester, and total steryl ester were strikingly increased in the li-1 fibers by 21.8-, 48.7-, and 45.5-fold, respectively ( B). In plant cells, too high or too low levels of sterols will inhibit plant growth, and sterol esters play a role in regulating the level of free sterols . The abnormal enrichment of sterol esters in the li-1 fibers may be a reason for their elongation suppression. Lipid droplets (LDs), also known as oil bodies or lipid bodies, are the “youngest” cellular organelle and are composed of a neutral lipid core surrounded by a phospholipid monolayer draped with hundreds of different proteins . Since the hydrophobic core of LDs is mainly composed of TG and steryl esters , we further investigated the LDs in the leaf and fiber cells. The number and size of oil bodies in the li-1 mutant obviously differ from those of TM-1. There were more and bigger oil bodies in both the leaf and fiber cells. LD biogenesis and degradation, as well as their interactions with other organelles, are tightly coupled to cellular metabolism and are critical to buffer the levels of toxic lipid species. Thus, LDs facilitate the coordination and communication between different organelles and act as vital hubs of cellular metabolism . The change in the LDs further confirmed the disruption of the lipid metabolism in the fiber cells of li-1 . This provided a novel clue to reveal the regulatory mechanism of fiber growth and development in the future.
The growth and development of fiber can be divided into four distinct and overlapping stages: initiation, elongation (primary wall formation), SCW deposition, and maturation. Following initiation (0~2 DPA), the fiber cells start to elongate. During the fast elongation period (8~12 DPA), the cell size and membrane area expand rapidly, and various metabolic activities need to be increased accordingly. This process requires a lot of lipids . At the stage of SCW deposition (15~40 DPA), cell expansion stopped and cellulose synthesis, transportation, and deposition proceeded steadily. The cellulose synthase complex is located in the plasma membrane. The lipid composition and membrane features at this stage are conducive to the synthesis and organization of cellulose . In this study, we analyzed the difference in lipid groups between the fast elongation stage (10 DPA) and the SCW formation stage of fibers (20 DPA). Most of the lipid components (16/23 lipid sub-classes and 56/82 lipid molecular species) were higher in the elongating fibers, and only a few lipid components (7/23 lipid sub-classes and 26/82 lipid molecular species) were higher in the SCW formation stage. Among them, glycerophospholipids (GP), sphingolipids (SP), and glycerolipids (GL) were enriched in the elongated fibers, which was similar to previous studies. It was reported that the content of polar lipids such as PC, PE, PI, PA, and PG was the highest in the elongating fibers (7–10 DPA) . These results suggested that the accumulation of these lipids may be necessary for fiber elongation. The lipids of the 10-DPA fiber and ovule of the wild type, and the 10-DPA ovule of the fuzzless – lintless mutant were detected by targeted lipidomics. The result showed that phosphatidylinositol (PI) was enriched in fiber cells and PI (34:3) was the highest lipid molecule species . Consistently, in our study, the lipid intensity of PI and molecular species PI (18:3/18:3) and PI (16:0/18:3) molecules was significantly higher in the 10-DPA fiber cells than in the 20-DPA fiber cells of TM-1. The PI intensity of the 10-DPA fibers from the li-1 mutant was also significantly higher than that of the 10-DPA fibers from the wild-type. However, the increased PI molecules were PI (16:0/18:1), PI (16:0/18:2), and PI (18:2/18:2), which indicated that the PI molecules enriched in the mutant 10-DPA fibers were different from those enriched in the wild-type 10-DPA fibers. Phospholipase D (PLD) can hydrolyze phospholipid to phospholipid acid (PA). GhPLDα1 was highly expressed in the 20-DPA fibers, which may lead to a decrease in the phospholipid content and an increase in PA in 20-DPA fibers . Consistently, most PA was enriched in the 20-DPA fibers. On the other hand, phospholipid acid can promote the production of hydrogen peroxide and induce the synthesis of SCW . These results indicate that phospholipids play some roles in the regulation of fiber elongation and SCW synthesis. The intensity of saccharlipid (SL), sterol ester (ST), and wax ester (WE) was enriched at the stage of SCW deposition, which indicated that the enrichment of these lipid components may promote cellulose synthesis and SCW formation. Phytosterols play important roles in the development of fiber cells . Sterols are comprised of free sterols and conjugated sterols such as sitosterol ester (SiE), campesterol ester, stigmasterol (StE) ester, sterol glycosides (SGS), and acetyl sterol glycosides (ASGS). Conjugated sterols play a role in the dynamic balance of sterols and in the synthesis of WEs . Three lipid classes of steryl esters, SiE, AGlcSiE, and StE, were detected in the study, among which, the AGlcSiE intensity was the highest in the rapid elongating fibers, and the intensity of SiE significantly increased in the stage of SCW synthesis. It was suggested that SiE might be associated with cellulose synthesis and SCW formation. Sterol glycosides and acetylsterol glycosides are enriched in elongating fibers. They may play roles in maintaining the balance between sterols and sphingolipids in fiber cell elongation . Sterols and sphingolipids are two key components of lipid rafts, which are functional regions of membranes . Sphingolipids are necessary for plant growth and development, in response to biotic stress or abiotic stress . Sphingolipids also played a role in cotton fiber elongation. FB1 (Fumonisin B1), an inhibitor of sphingolipid synthesis, significantly inhibited fiber elongation and altered the activity of lipid raft activity in fiber cells . In this study, CerG1, CerG2, CerG1 (d18:2/18:0), CerG1 (d18:2/18:1), CerG1 (d36:2), and CerG2 (d34:4) are enriched in the elongating fiber cells. Moreover, these molecules are closely associated with the AGlcSiE lipid molecule , which indicated that these molecules play a role in the regulation of fiber cell elongation. On the other hand, VLCFA is a component of sphingolipid molecules and plays an important role in fiber cell elongation. Application of ACE inhibited fiber elongation while VLCFA promoted cell elongation by promoting ethylene synthesis . These results indicated that there is a close relationship between sphingolipids, sterol, and membrane lipid rafts during the process of fiber elongation. The future study of these relationships may be an important aspect to reveal the regulatory mechanism in the growth and development of fibers.
Ligon lintless-1 ( li-1 ) is a dominant mutant of Gossypium hirsutum , which has the phenotype of damaged vegetative growth and short and thick fibers. Although the gene identification and gene expression profile of the fibers of the li-1 mutant have been studied for many years, the regulatory mechanism of the fiber growth deficiency is still unclear . Compared with the wild type, unexpectedly, most lipid classes and lipid molecule species have a higher rather than lower intensity in mutant fibers (20/25 lipid species, 60/68 lipid molecular species), and only 5 lipid species and 8 lipid molecule species are lower in mutant fibers. In the fibers of the li-1 mutant, the lipid classes and lipid molecule species of TG and SL were enriched. Consistently, GL and SL were enriched in the stage of SCW synthesis in the wild-type fibers. The high intensity of GL and SL in the li-1 fiber may promote SCW synthesis and inhibit fiber elongation. Consequently, the li-1 fiber is shorter and thicker than the wild-type fiber. Sphingolipids and GPs were the major lipids in li-1 cells. LPC and PLE were enriched in the 10-DPA fibers during fiber development in the wild type but decreased in the li-1 fiber cells. These two GPs may play a key role in fiber elongation. FB1 is an inhibitor of ceramide synthase that is the center of sphingolipid metabolism and biology . The exogenous application of FB1 strongly inhibited fiber cell growth and its fiber phenotype is similar to that of the li-1 fiber . FB1 treatment resulted in a decrease in the total GIPC and all GIPC molecular species . Consistent with the FB1 treatment, the total GIPC and all molecular species were reduced in the li-1 mutant fibers ( F). Furthermore, the expression levels of two genes encoding ceramide synthases Gh_D12G0217 and Gh_A07G0513 were significantly reduced . These results revealed that GIPC synthesis was inhibited in the li-1 fibers and GIPC might play an important role in fiber cell elongation. There are 10 homologous genes encoding ceramide synthase in the upland cotton genome. In our previous studies, overexpression of GhCS1 (Gh_D07G0583) promoted the synthesis of ceramide molecules containing dihydroxy LCB and VLCFA and inhibited the initiation and elongation of fiber cells . Consistently, the most ceramide molecules containing dihydroxy LCB and VLCFA were significantly increased in the li-1 mutant fibers ( D). Taken together, given sphingolipids are regarded as major regulators of lipid metabolism , sphingolipid balance might be an important factor for the normal growth of fiber cells. During fiber development, SiE was enriched in the 20-DPA fibers while acetylglycosyl sterol ester and StE were enriched in the 10-DPA fibers. However, the lipid intensity of three sterol esters in the mutant fiber was higher than that in the wild-type fiber, and the intensity of the SiE molecule (SiE (18:3)) was 59.56 times that in the wild-type 10-DPA fibers. This change was further confirmed by the targeted lipidomics of phytosterols. Compared with TM-1, the contents of campesteryl ester, sitosteryl ester, and total steryl ester were strikingly increased in the li-1 fibers by 21.8-, 48.7-, and 45.5-fold, respectively ( B). In plant cells, too high or too low levels of sterols will inhibit plant growth, and sterol esters play a role in regulating the level of free sterols . The abnormal enrichment of sterol esters in the li-1 fibers may be a reason for their elongation suppression. Lipid droplets (LDs), also known as oil bodies or lipid bodies, are the “youngest” cellular organelle and are composed of a neutral lipid core surrounded by a phospholipid monolayer draped with hundreds of different proteins . Since the hydrophobic core of LDs is mainly composed of TG and steryl esters , we further investigated the LDs in the leaf and fiber cells. The number and size of oil bodies in the li-1 mutant obviously differ from those of TM-1. There were more and bigger oil bodies in both the leaf and fiber cells. LD biogenesis and degradation, as well as their interactions with other organelles, are tightly coupled to cellular metabolism and are critical to buffer the levels of toxic lipid species. Thus, LDs facilitate the coordination and communication between different organelles and act as vital hubs of cellular metabolism . The change in the LDs further confirmed the disruption of the lipid metabolism in the fiber cells of li-1 . This provided a novel clue to reveal the regulatory mechanism of fiber growth and development in the future.
4.1. Plant Materials and Growth Conditions Gosspium hirsutum ligon lintless-1 ( li-1 ) mutant and its wild-type Texas Marker-1 (TM-1) were kindly provided by the institute of Cotton Research, Chinese Academy of Agricultural Science, and were cultivated in a field with regular management. Flowers were tagged on the day of anthesis (0 DPA, days post-anthesis). 4.2. RNA Extraction and qRT-PCR Assay The total RNA of 10-DPA and 20-DPA fiber cells from TM-1 and li-1 mutant plants was extracted by using a Plant Total RNA Extraction kit (Tiangen, Beijing, China). An amount of 1.0 μg total RNA was used to synthesize cDNA using a Reverse Transcription Kit with Genomic DNA Remover (Takara, Kusatu, Japan). Quantitative real-time reverse transcription PCR (RT-qPCR) was performed on a CFX96 Optical Reaction Module (Bio-Rad, Hercules, CA, USA) using Novostar-SYBR Supermix (Novoprotein, Shanghai, China) according to the manufacturer’s instructions. The primers for each gene were indicated in . Cotton GhHISTONE3 was used as an internal control. Their sequences are in the primer list in . Each analysis was repeated with three biological replicates. Calculation method adopts 2 −ΔΔCt . 4.3. Sample Preparation and Lipid Extraction and Lipidomics The 10-DPA and 20-DPA fibers of the TM-1 (wild type) and the 10-DPA fibers of the li-1 mutant were isolated from ovules and frozen in liquid N2. Lipids were extracted according to the MTBE (methyl tert-butyl ether) method. Briefly, samples were ground into fine powder in liquid nitrogen. The 100 mg samples were homogenized with 200 μL water and 240 μL precooled methanol. Then, 800 μL of MTBE was added and the mixture underwent ultrasound for 20 min at 4 °C followed by sitting still for 30 min at room temperature. The solution was centrifuged at 14,000× g for 15 min at 10 °C and the upper organic solvent layer was obtained and dried under nitrogen stream. The dried extraction was dissolved in 200 μL isopropanol by vortexing for 2 min and followed by centrifugation at 14,000× g for 15 min at 10 °C and the upper organic solvent layer was obtained and dried under nitrogen stream. The dried extraction was dissolved in 200 μL isopropanol by vortexing for 2 min and followed by centrifugation at 14,000× g for 15 min at 10 °C. The supernatant was subjected to analysis. Lipid analysis was performed using a UHPLC-MS/MS system consisting of a Shimadzu Nexera UHPLC system (manufactured by Shimadzu Corporation, Kyoto, Japan). Reverse-phase chromatography using a CSHC18 column (1.7 μm, 2.1 mm × 100 mm, Waters (Waters Corporation, Milford, MA, USA)) was chosen for LC separation. The lipid extracts were re-dissolved in 200 μL 90% isopropanol–acetonitrile, centrifuged at 14,000× g for 15 min, and finally, 3 μL of sample was injected. Solvent A was acetonitrile–water (6:4, v / v ) with 0.1% formic acid and 0.1 Mm ammonium formate and solvent B was acetonitrile–isopropanol (1:9, v / v ) with 0.1% formic acid and 0.1 mM ammonium formate. The initial mobile phase was 30% solvent B at a flow rate of 300 μL.min −1 . It was held for 2 min, and then linearly increased to 100% solvent B in 23 min, followed by equilibrating at 5% solvent B for 10 min. Mass spectra were acquired by Q-Exactive Plus in positive and negative mode, respectively. ESI parameters were optimized and preset for all measurements as follows: source temperature, 300 °C; capillary temp, 350 °C; the ion spray voltage was set at 3000 V; S-Lens RF Level was set at 50%; and the scan range of the instruments was set at m/z 200–1800. “Lipid Search” is a search engine for the identification of lipid species based on MS/MS math. LipidSearch contains more than 30 lipid classes and more than 1,500,000 fragment ions in the database. Both mass tolerance for precursor and fragment were set to 5 ppm. OPLS-DA was applied to the data by using SIMCA-P Software 14.1 (Umetrics, Umeå, Sweden) to discover the lipid differences. The screening criteria were variable importance for the projection (VIP) > 1 and p value < 0.05. 4.4. The Nile Red Stain Fresh fibers and leaves were placed in 2 mL EP tubes, fixed with paraformaldehyde, and placed at 4 °C for at least 4 h, overnight: washed twice with PBS (phosphate), dyed with 2 mL Nile red for about 1 h (tinfoil shielded from light), washed twice with PBS, and stored in PBS at 4 °C for later use. 4.5. Bioinformatic Analysis Refer to the RNA-seq analysis method. Total RNA was extracted from the 10-DPA fiber cells of the li-1 mutant and its wild-type TM-1. They were frozen in liquid nitrogen and sent to Shanghai Meiji Biological Company for Illumina Sequencing, meiji biological cloud web site ( https://cloud.majorbio.com/ (accessed on 6 January 2024)) in the United States for the sequencing analysis results. The DEGseq difference Different analysis was used, at a significance level p-adjust difference multiple ≥2, multiple test correction method BH (FDR) with Correction with Benjamini/Hochberg, Differential Expression was obtained Genes), and then functional annotation analysis and functional enrichment analysis were performed on the obtained differentially expressed genes. 4.6. Statistical Analysis All investigations were performed using three or six separate biological replicates. The results were presented as mean ± standard error (SE). A one-way analysis of variance was performed using SPSS 22 (IBM, New York, NY, USA) to determine significant differences. Statistical discrepancies were with p -values.
Gosspium hirsutum ligon lintless-1 ( li-1 ) mutant and its wild-type Texas Marker-1 (TM-1) were kindly provided by the institute of Cotton Research, Chinese Academy of Agricultural Science, and were cultivated in a field with regular management. Flowers were tagged on the day of anthesis (0 DPA, days post-anthesis).
The total RNA of 10-DPA and 20-DPA fiber cells from TM-1 and li-1 mutant plants was extracted by using a Plant Total RNA Extraction kit (Tiangen, Beijing, China). An amount of 1.0 μg total RNA was used to synthesize cDNA using a Reverse Transcription Kit with Genomic DNA Remover (Takara, Kusatu, Japan). Quantitative real-time reverse transcription PCR (RT-qPCR) was performed on a CFX96 Optical Reaction Module (Bio-Rad, Hercules, CA, USA) using Novostar-SYBR Supermix (Novoprotein, Shanghai, China) according to the manufacturer’s instructions. The primers for each gene were indicated in . Cotton GhHISTONE3 was used as an internal control. Their sequences are in the primer list in . Each analysis was repeated with three biological replicates. Calculation method adopts 2 −ΔΔCt .
The 10-DPA and 20-DPA fibers of the TM-1 (wild type) and the 10-DPA fibers of the li-1 mutant were isolated from ovules and frozen in liquid N2. Lipids were extracted according to the MTBE (methyl tert-butyl ether) method. Briefly, samples were ground into fine powder in liquid nitrogen. The 100 mg samples were homogenized with 200 μL water and 240 μL precooled methanol. Then, 800 μL of MTBE was added and the mixture underwent ultrasound for 20 min at 4 °C followed by sitting still for 30 min at room temperature. The solution was centrifuged at 14,000× g for 15 min at 10 °C and the upper organic solvent layer was obtained and dried under nitrogen stream. The dried extraction was dissolved in 200 μL isopropanol by vortexing for 2 min and followed by centrifugation at 14,000× g for 15 min at 10 °C and the upper organic solvent layer was obtained and dried under nitrogen stream. The dried extraction was dissolved in 200 μL isopropanol by vortexing for 2 min and followed by centrifugation at 14,000× g for 15 min at 10 °C. The supernatant was subjected to analysis. Lipid analysis was performed using a UHPLC-MS/MS system consisting of a Shimadzu Nexera UHPLC system (manufactured by Shimadzu Corporation, Kyoto, Japan). Reverse-phase chromatography using a CSHC18 column (1.7 μm, 2.1 mm × 100 mm, Waters (Waters Corporation, Milford, MA, USA)) was chosen for LC separation. The lipid extracts were re-dissolved in 200 μL 90% isopropanol–acetonitrile, centrifuged at 14,000× g for 15 min, and finally, 3 μL of sample was injected. Solvent A was acetonitrile–water (6:4, v / v ) with 0.1% formic acid and 0.1 Mm ammonium formate and solvent B was acetonitrile–isopropanol (1:9, v / v ) with 0.1% formic acid and 0.1 mM ammonium formate. The initial mobile phase was 30% solvent B at a flow rate of 300 μL.min −1 . It was held for 2 min, and then linearly increased to 100% solvent B in 23 min, followed by equilibrating at 5% solvent B for 10 min. Mass spectra were acquired by Q-Exactive Plus in positive and negative mode, respectively. ESI parameters were optimized and preset for all measurements as follows: source temperature, 300 °C; capillary temp, 350 °C; the ion spray voltage was set at 3000 V; S-Lens RF Level was set at 50%; and the scan range of the instruments was set at m/z 200–1800. “Lipid Search” is a search engine for the identification of lipid species based on MS/MS math. LipidSearch contains more than 30 lipid classes and more than 1,500,000 fragment ions in the database. Both mass tolerance for precursor and fragment were set to 5 ppm. OPLS-DA was applied to the data by using SIMCA-P Software 14.1 (Umetrics, Umeå, Sweden) to discover the lipid differences. The screening criteria were variable importance for the projection (VIP) > 1 and p value < 0.05.
Fresh fibers and leaves were placed in 2 mL EP tubes, fixed with paraformaldehyde, and placed at 4 °C for at least 4 h, overnight: washed twice with PBS (phosphate), dyed with 2 mL Nile red for about 1 h (tinfoil shielded from light), washed twice with PBS, and stored in PBS at 4 °C for later use.
Refer to the RNA-seq analysis method. Total RNA was extracted from the 10-DPA fiber cells of the li-1 mutant and its wild-type TM-1. They were frozen in liquid nitrogen and sent to Shanghai Meiji Biological Company for Illumina Sequencing, meiji biological cloud web site ( https://cloud.majorbio.com/ (accessed on 6 January 2024)) in the United States for the sequencing analysis results. The DEGseq difference Different analysis was used, at a significance level p-adjust difference multiple ≥2, multiple test correction method BH (FDR) with Correction with Benjamini/Hochberg, Differential Expression was obtained Genes), and then functional annotation analysis and functional enrichment analysis were performed on the obtained differentially expressed genes.
All investigations were performed using three or six separate biological replicates. The results were presented as mean ± standard error (SE). A one-way analysis of variance was performed using SPSS 22 (IBM, New York, NY, USA) to determine significant differences. Statistical discrepancies were with p -values.
This study analyzed the lipid changes in the 10-DPA and 20-DPA fiber cells of wild-type (TM-1), as well as the lipid differences in the 10-DPA fibers between the TM-1 and li-1 mutant. The predominant lipid classes enriched in fiber cells are GP (LPC, PA, PC, PE, PG, PI, and PS), SP (So, Cer, CerG1, CerG2), GL (DG and TG), SL (MGDG), ST (AGlcSiE), and Co. Compared to the rapid elongation stage of fibers, all SPs and most GPs, TG, and AGlcSiE were reduced, while CL LPA, MG, DGMG, MGMG, SiE, and WE increased in the secondary cell wall deposition stage of fibers. Dramatically, the contents of most lipid classes in the 10-DPA fibers were far higher in the li-1 mutant than in TM-1, among which, the content of all STs strikingly increased, especially SiE and StE, which were increased by 64.37 and 3.99 times, respectively. Only LPC, LPE, CerG2, SM, and FA were decreased in the li-1 fiber cells. The detailed analysis of the sphingolipid and sterol profile showed that the contents of S1P, Cer, and GluCer significantly increased while those of Sph and GIPC drastically decreased in the li-1 fibers. Similarly, the content of all sterols and conjugated sterols significantly increased in the li-1 fibers, especially the content of SiE and total steryl esters, which increased by 48.7 and 45.5 times, respectively. The results from the transcriptomes showed that differentially expressed genes were enriched in the pathways of the lipid metabolism. Furthermore, the number and size of lipid droplets were altered in the li-1 fiber cells and leaves. These results revealed that the lipid metabolism was disrupted drastically in the li-1 short fiber cells and fiber cell elongation was blocked in the li-1 mutant, which was not due to a lack of lipids, but rather lipid over-accumulation (obesity). This provides a new insight into the regulatory mechanism of fiber cell growth and development.
|
The role of telepathology in diagnosis of pre-malignant and malignant cervical lesions: Implementation at a tertiary hospital in Northern Tanzania | 0d1e610e-3760-4c2d-ae91-32c309bb0965 | 9009664 | Pathology[mh] | Cervical cancer is one of the most preventable and treatable malignant diseases. Yet it is the fourth most commonly detected cancer in women worldwide, with more than half a million new cases and 300,000 deaths in 2018 . The disease is disproportionally dispersed, with low- and middle-income countries (LMICs) accounting for more than 90% of the disease burden. This uneven distribution reflects ineffective screening and poor pathological service which are prevailing problems in most LMICs. East Africa is among the regions with highest incidence rate, about ≥42.7 per 100,000 and yet women especially in rural areas do not have access to cervical cancer screening programs . Cervical cancer is the most common cancer (38.4%) and the main cause of female cancer deaths (34.3%) among Tanzanian women . The age-standardized incidence rate is 54 per 100,000 women, which is almost double the average age-standardized rate for Africa (27.6 per 100,000 women) . High incidence and mortality from cervical cancer in Tanzania, suggest an urgent need for improved screening, diagnostic and treatment approaches. As the number of newly diagnosed cervical cancer patients in Sub-Saharan Africa has increased rapidly, timely access to screening and diagnostic services are necessary pillars for scaling up the cancer control, . However, there is a gap in the complete cancer care cycle in most low-and-middle income countries (LMICs) due to inadequate pathology resources as a result of extreme shortages of pathologists, deficiency of infrastructures and reagents, poor specimen handling and storage . Due to severe shortage of pathologists, many cancer patients do not have timely access to diagnostic service and treatment, and thus die at home or present at the hospital with advanced disease with poor prognosis . Tanzania has limited number of well-trained and experienced pathologists; most of them are general surgical pathologists who are not sub-specialized, working in teaching hospitals in large cities . Likewise, there is a challenge in specimens transport, costs and long turn-around time which altogether contribute to the delayed diagnosis . Pathologists in LMICs may have difficulties in diagnosing rare, complex or challenging cancer cases because in these settings advanced diagnostic tests such as molecular or immunohistochemistry are usually not readily available. Thus, error in pathology diagnosis is not uncommon, which often may result in wrong treatment . Telepathology is the practice of pathology using telecommunications to transmit digital images and data between two or more sites remotely located from each other . The use of telepathology increases access to specialized pathology services, and the technology can be used in routine surgical pathology, consultations, quality assurance, education and research . However, there is scanty data on the use of telepathology to support cancer care in Tanzania . We therefore conceived the present study to investigate the feasibility and the role of implementing telepathology by assessing the agreement in the diagnoses of cervical biopsies made conventionally on glass slides and on scanned images telepathology platform across the participating pathologists. Study design and study setting This was a cross-sectional study conducted from January to December 2020 at Kilimanjaro Christian Medical Center (KCMC), which is one of the major four referral and consultant hospitals in Tanzania. At the end of 2019, Tanzania’s population was estimated to be at 55.9 million, the KCMC hospital serves about 17 million people in Northern part of the country and Kilimanjaro region . The hospital is a University Teaching Hospital for Kilimanjaro Christian Medical University College (KCMUCo) and it offers both general and specialized care. For many years, the Pathology Department of KCMC, had no local pathologists; pathology services were supported by occasional visiting pathologists from Radboud University Medical Center, Netherlands. In the year 2016, two junior local (Tanzanians) pathologists joined the Pathology Department. The first author (AM) is one of the two local pathologists. Telepathology platform In response to increasing demands of timely and reliable histological diagnosis of malignant diseases, KCMC acquired a whole slide image scanning (WSI) equipment which was installed in November 2019. The system (Motic EasyScan, USA), is a website based tele-consultation platform ( www.med3.motic.com ), developed for capacity building on cancer diagnosis and management in resources limited setting (Figs and ). After successful installation and training of the KCMC Pathology Department staff, consultant and sub-specialized pathologists were registered and linked in the platform. Continuous communication between KCMC pathologists, Information Technology (IT) personnel and consultant telepathologists was made electronically to address any raised issues with regard to the quality of the images and troubleshooting challenges that hindered the smooth system workflow of the platform. Recruitment procedure A detailed participants enrollment procedure has been described elsewhere . Briefly, women residing in rural Kilimanjaro were invited via public announcements to attend a community based cervical cancer screening with HPVself-sampling using the Evalyn Brush ® . Authorization to conduct the study was obtained from the local government management committees as well as the community stakeholders. Two experienced reproductive health nurses informed women who showed up for screening about the study objectives and eligibility criteria, and they received written informed consent from all participants. Care HPV was used to test the presence of high risk HPV infection. Participants found with positive HPV test were recalled for further investigations which included cervical visual inspection with acetic acid (VIA), Pap smear and cervical biopsies. Consecutive biopsies obtained from the participants were routinely processed and stained with Hematoxylin and Eosin (H&E) at the Pathology Department of KCMC. Data collection H&E stained glass slides of recruited cases were first interpreted conventionally by light microscope, (Olympus BX43 F, Japan). After primary reading the glass slides, they were scanned at 40X original magnification using the Motic whole slide scanner. The images were stored in a local server. Each scanned case was accompanied with relevant anonymized clinical information of the participant. After a washout period of at least three months of the primary glass slide reading (conventional microscopy), the local pathologist reviewed the scanned slides on a 24-inch monitor (Dell, Round Rock, Texas). The microscopic glass slides of the study cases were then shipped to Odense University Hospital (OUH), Denmark where two pathologists (an experienced senior gynecologic pathologist and a junior surgical pathologist) reviewed them independently. After a wash-out period of at least one month of reviewing the glass slides, the OUH pathologists reviewed the scanned images on their personal computer screens in a blinded fashion. Data analysis Results for categorical variables were expressed as absolute numbers and percentages and 95% Confidence Intervals (CI). The agreement between arbitrary pairs of observers (inter-observer agreement) was measured by kappa statistics (κ). Kappa is an index of agreement over and above that which is expected by chance alone and is scored as a number between 0 and 1 . Here, intra-observer reliability describes whether pathological diagnoses rendered using telepathology platform (scanned images) are comparable (non-inferior) to diagnoses made by conventional microscopy e.g. degree of agreement between the two diagnostic methods. Inter-observer agreement describes the degree of agreement across pathologists. Inter-observer reliability demonstrates the level of agreement in establishing diagnoses on conventional microscopy and scanned images across the pathologists. A nomenclature recommended by Landis-Koch was adopted for interpreting the strength of agreement (κ); values >0.75 are regarded as excellent agreement beyond chance, values between 0.40 and 0.75 as fair to good agreement beyond chance and values <0.40 as poor agreement beyond chance . To test the quality of response rates from glass slides and scanned images diagnoses, McNemar’s test was used. An effect was considered statistically significant if the p-value of its corresponding test statistic was 5% (p <0.05). Differences in diagnoses established across the pathologists were evaluated and scored on a three-point scale: “Discordance” corresponded to cases with a difference in diagnosis that would be associated with a difference in patient care, e.g. when initial diagnosis was normal cervix while review diagnosis was cancer; normal versus CIN2, normal versus CIN3. “Partial concordance” corresponded to cases with a minor discrepancy in diagnosis that may not be associated with a difference in patient care, e.g. normal versus CIN1or CIN2 versus CIN3. “Concordance” corresponded to cases where both observers gave identical diagnoses, . The statistical analyses were performed using the software packages StatXact version 16 (MathSoft, Inc, Seattle, WA, USA). Final gold standard diagnosis and ethical considerations In case of discrepancy, the diagnosis established on conventional microscopy by the senior pathologist was considered as gold standard. The local pathologist was legally responsible for the final diagnoses and thus, signed out a final report for each case. Diagnoses rendered by pathologists from OUH were regarded as second opinions since OUH pathologists are not licensed to practice medicine in Tanzania. The three pathologists did not discuss the diagnostic criteria prior to the study but referred to the WHO Classification . The study was approved by the College Research Ethics Review Committee at Kilimanjaro Christian Medical University College and the National Institute for Medical Research of Tanzania; with approval certificates numbers of 2401 and NIMR/HQ/R.8a/Vol.IX/3093 respectively. This was a cross-sectional study conducted from January to December 2020 at Kilimanjaro Christian Medical Center (KCMC), which is one of the major four referral and consultant hospitals in Tanzania. At the end of 2019, Tanzania’s population was estimated to be at 55.9 million, the KCMC hospital serves about 17 million people in Northern part of the country and Kilimanjaro region . The hospital is a University Teaching Hospital for Kilimanjaro Christian Medical University College (KCMUCo) and it offers both general and specialized care. For many years, the Pathology Department of KCMC, had no local pathologists; pathology services were supported by occasional visiting pathologists from Radboud University Medical Center, Netherlands. In the year 2016, two junior local (Tanzanians) pathologists joined the Pathology Department. The first author (AM) is one of the two local pathologists. In response to increasing demands of timely and reliable histological diagnosis of malignant diseases, KCMC acquired a whole slide image scanning (WSI) equipment which was installed in November 2019. The system (Motic EasyScan, USA), is a website based tele-consultation platform ( www.med3.motic.com ), developed for capacity building on cancer diagnosis and management in resources limited setting (Figs and ). After successful installation and training of the KCMC Pathology Department staff, consultant and sub-specialized pathologists were registered and linked in the platform. Continuous communication between KCMC pathologists, Information Technology (IT) personnel and consultant telepathologists was made electronically to address any raised issues with regard to the quality of the images and troubleshooting challenges that hindered the smooth system workflow of the platform. A detailed participants enrollment procedure has been described elsewhere . Briefly, women residing in rural Kilimanjaro were invited via public announcements to attend a community based cervical cancer screening with HPVself-sampling using the Evalyn Brush ® . Authorization to conduct the study was obtained from the local government management committees as well as the community stakeholders. Two experienced reproductive health nurses informed women who showed up for screening about the study objectives and eligibility criteria, and they received written informed consent from all participants. Care HPV was used to test the presence of high risk HPV infection. Participants found with positive HPV test were recalled for further investigations which included cervical visual inspection with acetic acid (VIA), Pap smear and cervical biopsies. Consecutive biopsies obtained from the participants were routinely processed and stained with Hematoxylin and Eosin (H&E) at the Pathology Department of KCMC. H&E stained glass slides of recruited cases were first interpreted conventionally by light microscope, (Olympus BX43 F, Japan). After primary reading the glass slides, they were scanned at 40X original magnification using the Motic whole slide scanner. The images were stored in a local server. Each scanned case was accompanied with relevant anonymized clinical information of the participant. After a washout period of at least three months of the primary glass slide reading (conventional microscopy), the local pathologist reviewed the scanned slides on a 24-inch monitor (Dell, Round Rock, Texas). The microscopic glass slides of the study cases were then shipped to Odense University Hospital (OUH), Denmark where two pathologists (an experienced senior gynecologic pathologist and a junior surgical pathologist) reviewed them independently. After a wash-out period of at least one month of reviewing the glass slides, the OUH pathologists reviewed the scanned images on their personal computer screens in a blinded fashion. Results for categorical variables were expressed as absolute numbers and percentages and 95% Confidence Intervals (CI). The agreement between arbitrary pairs of observers (inter-observer agreement) was measured by kappa statistics (κ). Kappa is an index of agreement over and above that which is expected by chance alone and is scored as a number between 0 and 1 . Here, intra-observer reliability describes whether pathological diagnoses rendered using telepathology platform (scanned images) are comparable (non-inferior) to diagnoses made by conventional microscopy e.g. degree of agreement between the two diagnostic methods. Inter-observer agreement describes the degree of agreement across pathologists. Inter-observer reliability demonstrates the level of agreement in establishing diagnoses on conventional microscopy and scanned images across the pathologists. A nomenclature recommended by Landis-Koch was adopted for interpreting the strength of agreement (κ); values >0.75 are regarded as excellent agreement beyond chance, values between 0.40 and 0.75 as fair to good agreement beyond chance and values <0.40 as poor agreement beyond chance . To test the quality of response rates from glass slides and scanned images diagnoses, McNemar’s test was used. An effect was considered statistically significant if the p-value of its corresponding test statistic was 5% (p <0.05). Differences in diagnoses established across the pathologists were evaluated and scored on a three-point scale: “Discordance” corresponded to cases with a difference in diagnosis that would be associated with a difference in patient care, e.g. when initial diagnosis was normal cervix while review diagnosis was cancer; normal versus CIN2, normal versus CIN3. “Partial concordance” corresponded to cases with a minor discrepancy in diagnosis that may not be associated with a difference in patient care, e.g. normal versus CIN1or CIN2 versus CIN3. “Concordance” corresponded to cases where both observers gave identical diagnoses, . The statistical analyses were performed using the software packages StatXact version 16 (MathSoft, Inc, Seattle, WA, USA). In case of discrepancy, the diagnosis established on conventional microscopy by the senior pathologist was considered as gold standard. The local pathologist was legally responsible for the final diagnoses and thus, signed out a final report for each case. Diagnoses rendered by pathologists from OUH were regarded as second opinions since OUH pathologists are not licensed to practice medicine in Tanzania. The three pathologists did not discuss the diagnostic criteria prior to the study but referred to the WHO Classification . The study was approved by the College Research Ethics Review Committee at Kilimanjaro Christian Medical University College and the National Institute for Medical Research of Tanzania; with approval certificates numbers of 2401 and NIMR/HQ/R.8a/Vol.IX/3093 respectively. A total of 175 H&E stained glass slides and their respective 175 scanned images were included in this analysis. The diagnoses were grouped into five categories: normal, CIN1, CIN2, CIN3 and cervical cancer. When diagnoses established by the local pathologist on conventional light microscopy were compared to scanned images, complete agreement was observed in 87.4% of the cases with intraobserver kappa statistic strength (ƙ) of 0.73; CI (0.65–0.79). However, the agreement varied widely among diagnostic categories. Complete agreement was 94.5%, 60%, 25%, 60% and 50% for the normal, CIN1, CIN2, CIN3 and cancer diagnostic categories respectively. For the senior gynecologic pathologist, the overall complete agreement between conventional light microscopy and telepathology was 85.7%; with intraobserver variability (ƙ) value of 0.76; CI (0.69–0.82). Agreement among diagnostic categories were 93.9%, 70.5%, 42.8%, 64.3% and 57.1% for the normal, CIN1, CIN2, CIN3 and cancer diagnostic categories respectively. On the other hand, the overall agreement of diagnoses established by the junior pathologist on conventional microscopy compared to scanned images was 90.9%. The intraobserver variability (ƙ) value for the junior pathologist was 0.73; CI (0.65–0.79). Agreement among diagnostic categories were 97.1%, 78.6%, 25%, 75% and 50% for the normal, CIN1, CIN2, CIN3 and cancer diagnostic categories respectively, . When diagnoses established on conventional microscopy by the senior pathologist were compared to the local pathologist, the overall agreement was 96%; with interobserver value (k) of 0.93; CI (0.87–1.00). Across diagnostic categories, the rate of agreement were 97.7%, 100%, 57.1%, 92.3% and 87.5% for the normal, CIN1, CIN2, CIN3 and cancer respectively. Similarly, on telepathology platform, the overall concordance for diagnoses established by senior pathologist compared to local pathologist was 96.6%. The interobserver reliability (k) value was 0.94; CI (0.88–1.00). Similarly, the concordance rates were 98.4%, 94.7%, 71.4%, 93.3% and 100% for the normal, CIN1, CIN2, CIN3 and cancer respectively . To assess the agreement of diagnoses established on glass slides compared to the scanned images across the three pathologists, a symmetric agreement matrix was created in which each diagnosis from each pathologist was compared with the corresponding diagnoses made by the other two observers. This matrix resulted in 1050 paired comparisons in total (3 observers, 2 comparisons, and 175 slides. The overall concordance across the three pathologists was 87.7%; k-value 0.54; CI (0.49–0.57). Across the diagnostic category, the concordance rates were 93.7%, 62.7%, 38.5%, 71.1% and 90.6% for normal, CIN1, CIN2, CIN3 and cancer respectively, the data summarized in . When diagnoses established on conventional microscopy were classified on a three-point scale; the rate of concordance was 66.3%, 66.9% and 82.3% between the three pathologists, and the rate of discordance was 2.3%, 4.6% and 1.7%. On scanned images, the rate of concordance was nearly comparable among the three pathologists with 73.3% for senior versus local pathologist, 76.6% for senior versus junior pathologist and 78.3% for junior versus local pathologist, and the rate of discordance was 1.1%, 3.4% and 4.0% . In this study, good agreement (87.4%; κ = 0.73) was found when comparing diagnoses established on conventional microscopy and scanned images by local pathologist; and across all 3 study pathologists, the overall concordance between conventional microscopy and scanned images was high (87.7%). The high intraobserver agreement of histological evaluation of cervical biopsies by the participating pathologists between conventional and scanned images shown in this study suggests that the diagnoses of cervical premalignant and malignant lesions are highly reproducible using both glass and digital formats. This implies that the use of digital images can be used for primary diagnosis and thus address the challenge of shortage of pathologists in Tanzania by providing alternative means for the health facilities without pathologists to access pathology healthcare services. Similar findings have been reported in a study conducted in Spain which demonstrated almost perfect agreement between the diagnoses established on conventional microscopy and scanned images, (κ = 0.91; CI (0.88–0.95)) . The results of our study are comparable with studies on feasibility and validation of telepathology conducted in Africa which have reported a slightly higher concordance rate between diagnoses established on conventional microscopy and scanned images. These studies reported good feasibility regarding the use of telepathology and also found scanned images to be of good quality, however internet speed was found to be a limiting factor. Unlike the current study which focused on cervical specimens, the former studies involved a wide spectrum of specimens (different organ systems); and immunohistochemistry tests were used whenever it was necessary . The findings of our study suggest that through telepathology, specific or definitive diagnoses can be achieved; and the discrepancies observed in our study are within the range of generally acceptable interobserver variability in surgical pathology practice . In this study, the overall concordance for the diagnoses established on conventional microscopy compared to scanned images across the three pathologists was excellent (87.7%). All participating pathologists established nearly comparable intra-observer diagnostic accuracy between conventional and scanned images suggesting that telepathology can potentially be a reliable tool to support primary cancer diagnosis in Tanzania. However, across diagnostic categories, relatively low concordances were observed in CIN2 lesions (38.5%), but it is well known that this diagnosis in particular has low reproducibility. Moreover, our study was not primarily intended to compare diagnostic capabilities between pathologists in Tanzania and Denmark, but rather to demonstrate the feasibility and role of implementing telepathology in supporting healthcare in Tanzania. Factors which may explain the observed discrepancy include poor quality of the glass slides or scanned images in some cases, and borderline lesions which can be challenging to interpret without the use of immunohistochemistry. High quality slides and images preparation are necessary for optimizing diagnostic accuracy for both conventional and digitalized slides. This is among the key areas that need to be considered in digital pathology. When the diagnoses established on convention microscopy by local pathologist are compared to the senior pathologist, four discrepant cases were identified and diagnosed as CIN3/normal, cancer/CIN3, CIN2/cancer and CIN2/CIN3 whereas only two discrepant cases were found on the scanned slides diagnosed as CIN2/CIN3 and CIN3/cancer. The majority of the partial discordances (72.7%) were related to discrepancies in the diagnosis of CIN1 lesions versus normal/reactive cervical epithelium which has minor clinical implications. Across diagnostic categories, these partial discrepancies that are without consequences for the clinical care, resulted in a lower κ value than the overall value. Studies using conventional microscopy have reported a substantial variation among and within pathologists’ observers in the interpretation of various histopathological lesions on H&E-stained tissue sections . The use of p16 immunohistochemistry has been associated with reduction of interobserver disagreement in the interpretation of cervical lesions and improves detection of high grade precancerous lesions associated with HPV infection; as p16 is overexpressed in almost all high grade lesions and negative or normal in reactive lesions. Recently, p16 has been recommended in the evaluation of cervical biopsies since it reduces the interobserver variability, particularly in cases of professional disagreement . In this study, both the local and OUH pathologists had little or no previous experience in the use of telepathology, however this did not affect the reproducibility. This may suggest that if telepathology is widely introduced in Tanzania, only minor difficulties should be expected. Tanzania has a shortage of more than 534 medical specialists, a situation which is worse in rural areas . Yet, the current rate of training suggests that it may take centuries before adequate number of specialist pathologists is obtained . In addition, as a developing country, challenges in medical services are likely to exist for a long time to come. At present, it is practically impossible for many patients to obtain pathology services in a timely manner since the services are only available in tertiary hospitals . Thus, in order to scale up cancer control, implementation of telepathology should be considered. Telepathology infrastructures may facilitate timely access to diagnostic and specialized healthcare services through networking. Moreover, telepathology can reduce travel costs for a number of patients who are referred to high level referral hospitals for diagnostic services . However, with the increasing cancer burden, telepathology cannot be an ultimate solution for the shortage of pathologists. The government and other stakeholders in healthcare should also invest in training new pathologists who are able to provide essential and specialized services in the country. The use of telepathology in surgical pathology practice has received extensive attention and widespread acceptance. In high-income countries, digital revolution is toward sweeping the field and the practice of anatomic pathology; as laboratories are switching from the conventional microscope to digital slide scanners for primary diagnosis . Studies from LMICs have highlighted good feasibility in implementing telepathology . In Tanzania, there are indeed several research projects on telepathology and these have given us important insights on feasibility and clinical utility . However, poor participation of indigenous researchers suggests an important limitation in these studies. Moreover, sustainability factors have not been critically evaluated. For instance, to date, there have been no large studies showing the success of telepathology which have involved various health centers within the country. For successful performance of this promising technological innovation, several factors should be considered. These include operating costs, data security as well as user acceptance. Additionally, continuous training of pathologists regarding principles and limitation of digital pathology, sustainable internet connectivity, continuous bi-directional communication, and support from the institutional leadership are key to the success of telepathology programs . A study comparing the cost for telepathology and a visiting pathologist services has revealed that establishing and running telepathology services were cheaper compared to hiring a visiting pathologist . The main costs of the telepathology service were related to installation of the technology, whereas the main costs of the visiting pathologist service were payroll costs. A neuropathology consultation study involving Muhimimbili National Hospital (Tanzania) and University of Colorado (USA) documented the initial high cost of telepathology equipment, especially robotic and WSI telepathology as a challenge . The results of this study indicate that telepathology can be used not only for primary diagnosis, but also as an international consultation platform that can improve patient care in Tanzania by facilitating access to pathology expertise. Once introduced, telepathology can revolutionize pathology services, as it is an important emerging adjunct to conventional light microscopy. It will enable remote diagnosis and improve collaboration among pathologists. The important strength of the present study is that it is one among few validation studies carried out in Tanzania. However, the study has some limitations. Firstly, the time needed to establish diagnoses on both conventional and telepathology platform was not recorded. Therefore, the actual turn-around times could not be established. Secondly, cost analysis of implementing the telepathology program was out of the scope of this study. Thirdly, infrastructural obstacles in using immunohistochemistry in limited resources setting; the intra and interobserver variablility most likely would have been improved by the use of p16 immunohistochemistry. Fourth, our study included exclusively cervical biopsies with a relatively small sample size. Thus, for generalizability purposes, larger studies involving many pathologists and with a wide range of organ systems (to assess the clinical impact and the effect of telepathology on surgical pathology) are recommended. Lastly, but not least, despite good agreement rates between telepathology and conventional microscopy established in this study, the findings should be interpreted cautiously since the kappa statistics had rather wide confidence intervals. In addition, there are some disputes of the value of using kappa statistics due to difficulties in interpreting indices of agreement . Therefore, setting an acceptable value of kappa should depend on the clinical context. This study focused on the performance of telepathology in relation to cervical cancer diagnosis. The fight against cervical cancer has been given increasing priority lately with the WHO Director-General announcing a global call for action towards the elimination of invasive cervical cancer as a public health problem. Through cost-effective, evidence based interventions, including human papillomavirus vaccination of girls, screening and treatment of precancerous lesions, and improving access to diagnosis and treatment of invasive cancers . In that relation it has been acknowledged that there is a need for more innovative technologies for detection of CIN2+ . Our results illustrate how telepathology can be used to increase access to appropriate diagnostic service and identification of precancerous lesions and cervical cancer. This may lead to improved diagnosis and treatment and ultimately improved survival of women who are at risk of dying from cervical cancer. In summary, our study demonstrate that diagnoses of cervical premalignant and malignant lesions in biopsies are highly reproducible using both glass and digital formats of the slides, implying that telepathology is non-inferior to conventional light microscopy for primary diagnosis. The comparable diagnostic concordance between Tanzanian and Danish pathologists suggests that telepathology service can be a reliable tool to support primary cancer diagnosis in resource-limited settings that have limited numbers of pathologists. Studies assessing the utility of telepathology on cytological specimens including Pap smears are recommended. Once validated, the technology has the potential to scale up cancer control in Tanzania. S1 Data Deidentified raw dataset for the study participants (scml.dta). (DTA) Click here for additional data file. S1 Appendix Definition of the key terminologies . (DOCX) Click here for additional data file. |
Herbal medicines use and associated factors among pregnant women in Debre Tabor town, north West Ethiopia: a mixed method approach | 9ca0c142-fbd8-4641-a669-034d558a7957 | 8547058 | Pharmacology[mh] | Traditional medicine is comprehensive in nature on the grounds that it delivers numerous issues identified with socio- cultural, economic and ecological context of a community . For instance, one of the working definition of “traditional medicine” by World Health Organization is that: “it is the sum total of all knowledge and practices, whether explicable or not used in the diagnosis, prevention and elimination of physical, mental or social imbalances and relying exclusively on practical experience and observation handed down from generation to generation, whether orally or in writing . Herbal medicines, which is part of traditional medicines are characterized as plant-determined material or preparations saw to have remedial advantages; they frequently contain crude or processed ingredients from at least one plants. Herbal medications incorporate herbs, home remedies, herbal preparations, and finished home-grown items that contain portions of plants or other plant materials as active ingredients and used by the general population as well as pregnant women . Pregnancy is a condition associated with immense physiological alterations resulting in many pregnancy-related problems, including nausea, vomiting, constipation, and heartburn . These ailments usually result in pregnant women self-medicating using traditional medicine, especially herbs . Furthermore, pregnant women, in developing countries, use herbal medicines due to its easy accessibility, affordability, lack access to health care and belief that herbs are safer for the fetus than modern medicine because they are natural products . The use of herbal medicines has increased in most countries in Africa and Asia as in many other developed countries. Approximately 65–80% of the world’s population use traditional medicine as their primary form of health care, including use during pregnancy . In sub-Saharan Africa, up to 80% of the population uses TM to meet their health care needs, including use during pregnancy . Like most African countries, Ethiopia relies heavily on indigenous medicine for its primary health care services . Despite the increased consumption of herbal medicines among pregnant women all over the globe, majority of them are unaware of the potential side effects and a potential teratogenicity of some herbal products . In Ethiopia, more than 80% of the population use traditional medicine. A study done in Hossana town, Southern Ethiopia showed that 73.1% of pregnant women use herbal medicines and the most common herbs used were ginger (55.8%), garlic (69.8%), and tenaadam (26.4%) . Similar study conducted in Nekemte Hospital, western Ethiopia reported that the most commonly used herbs, by pregnant women, were ginger (44.36%) and Garlic (37.32%). Nausea (23.90%) and morning sickness (21.05%) were the most common reasons for herbal use in pregnancy . Another study conducted in Gondar, Ethiopia, found ginger (40.7%) and garlic (19%) were the two most commonly used herbs in pregnancy. Common cold and inflammation were the most common reasons for herbal use . Several studies have also reported the association between different socio-demographic characteristics of pregnant women, and utilization of herbal medicines during pregnancy. The most important variables listed by scholars includes: low level of education, rural residence, no occupation, older, and positive attitude towards the use of herbal medicines . In spite of the studies done on herbal medicine, there is a limited data on the prevalence and correlates of herbal medicine use among pregnant women using mixed method approach. Therefore, the objective of this study was to assess the prevalence of herbal medicine use and associated factors among pregnant women in Debre tabor town, Northwest Ethiopia using mixed study design.
Study design and setting A Community based cross-sectional explanatory sequential mixed methods approach was conducted to assess the use of herbal medicines and associated factors among pregnant women in Debre Tabor Town from September 1 to 30, 2019. The town was selected as the study area because it has unique cultural, environmental landscape, and variety of plant species that do affect the pattern of herbal medicine uses. Debre Tabor town is located in south Gondar zone, northwest Ethiopia, 667 km away from Addis Ababa (the capital city of Ethiopia). Based on the information from Debre Tabour town Administrative Bureau, the town has 10 kebeles and with an estimated total population of 111,029 (Fig. ). Conceptual framework To understand factors that motivate the use of herbal medicine among the study community, a conceptual framework was adapted from literatures . It suggests that herbal medicine uses by pregnant women depends on four core components which includes: sociodemographic factor, obstetric related factors, health service-related factors and attitude towards the use of herbal medicines. This conceptual framework also effectively integrates different factors which may influence the use of herbal medicine by pregnant women (Fig. ). Sample size determination and procedure The source populations were all pregnant women residing in Debre Tabor town, while those pregnant women who were registered at the health extension workers registration book during the data collection period were taken as a study population. The sample size was calculated using a single population proportion formula with the assumption of 95% confidence interval, 5% margin of error and 48.6% prevalence of herbal medicine use among pregnant women and 10% for possible non response was taken to determine a final sample size of 267. By dividing the total number of pregnant of the ten kebele (650) with the total sample size (267), an interval of two was used to select household during the one-month data collection period by using systematic random sampling technique. The first household was selected using the lottery method. Then the next household was selected with an interval of two. If the study participant was not available at the first visit, this household was revisited once the same day or following day. If not available again, the study participant was considered as non-respondent. Pregnant women who were willing to share the information and available at the time of data collection were included in this study. Qualitative Approach: The health extension workers also helped to select 12 participants for one focus group discussion (FGD) and six individuals were selected purposively for in-depth interviews from the FGD based on their knowledge about local herbs and ability to describe their experiences in the focus group. Data collection tools and techniques Data collection was performed by three data collectors (BSC nurses) through interviewer-administered questionnaire. The data collectors were properly trained on the purpose of the study, the content of the questionnaire, interviewing techniques, how to approach the respondents, and securing their permission for interview prior to the data collection process. The data collection tool was adapted from different literature after review of the published studies and prepared in English. This was translated to local language ( Amharic ) and then back translated to English in order to ensure consistency. The data collection instrument was pretested on 26 pregnant women who were not included in the final analysis and relevant modifications were done before the commencement of actual data collection. The final questionnaire constituted 33 items that were divided into five main parts . The first section includes questions about socio-demographic data, the second section includes questions about related to the use of herbal medicines, the third section includes questions about some obstetric related factors, the fourth section includes questions about accessibility of health facility and finally, the last section assessed the attitude of sample population regarding use of herbal medicine during pregnancy with five Likert-scale including seven questions. Each question had five choices (ranging from strongly agree to strongly disagree). Item analysis was done, and the internal consistency reliability had a Cronbach’s alpha of 0.82. The answer scores for each question choice and question in both groups were added up and their means and standard deviations calculated. The question scores ranged from 7 to 34 representing the most negative and the most positive attitude respectively. The higher the question score, the more positive the attitude is. After computing the median of all respondents’ responses, the median score of each respondent was dichotomized as have a positive attitude or negative attitude. A score of ≥19 was defined as a “positive attitude towards the use of herbal medicines during pregnancy,” and a score of < 19 was defined as “negative attitude towards the use of herbal medicines during pregnancy.” Positive attitude ≥median. Negative attitude <median. In this study, respondents were considered as herbal medicine users if they have taken herbal medicine(s) through oral, intra-vaginal or topical routes during gestational period. Other preparations that are consumed as routine meal preparations and those that are taken as nutrients were excluded. The qualitative data collection was conducted using an interview guide through probing questions by the principal investigator. Interviews lasted 30–90 min using an audio recorder as well as a filled note was taken by one trained note taker. Participants were briefed about the aim of the study by the principal investigator. Verbal informed consent was obtained from the participants and place of the interview was arranged between the principal investigator and the interviewee, the FGD and in-depth interviews were conducted at the public places in the villages. Data processing and analysis Quantitative data was entered into EpiData version 4.2.0.0 and exported to Statistical Package for the Social Sciences (SPSS) software version 25.0 for analysis. Descriptive statistics (frequencies, percentages, mean, and standard deviation) and inferential statistics (bivariate and multivariate analyses) were calculated using bivariate and multivariate logistic regression with a 95% confidence interval (CI). Bivariate logistic regression was used to measure the association between independent variables and herbal medicine use. First bivariate logistic regression was performed to identify candidate variables for multiple logistic regressions. Those variables with a p -value of below 0.25 in the bivariate analysis were fitted to multiple logistic regressions. Model fitness was tested using the Hosmer and Lemeshow’s test and it was insignificant. Multi-collinearity was checked using variance inflation factor (VIF). Covariates with a p-value of below 0.05 in multivariate logistic regression were considered statistically significant factors with the dependent variable (herbal medicine use). Finally, the crude and adjusted odds ratio (OR) with 95% confidence interval (95% CI) were computed to measure the strength of the association between the outcome and the independent variables. A focus group discussion and interviews were audio recorded and transcribed verbatim in Amharic. Texts were read independently by the PI and another professional who speaks the local language and codes were developed in reference to the research questions. Each of the codes were organized into higher-order conceptual themes. These individual codes and themes were discussed at group meetings until consensus was reached on basic themes and subthemes across a focus group and interviews. Finally, the themes were incorporated into a conceptual model of the participants and their use of herbal medicines and associated factors among pregnant women . Sections of original transcripts and key quotes considered to be illustrative of the emerging themes were translated into English to facilitate discussion with the full research team. Data analysis was supported by the use of NVivo 10 computer software. Data quality assurance The quality of the quantitative data was assured by pre-testing the questionnaire on 10% of the sample size (26 pregnant women) in a Town which is different from the study area prior to the start of the actual study to test the fitness of the questionnaire for the study settings. Training about the data collection tool as well as data collection procedures was given to data collectors and supervisors for a total of 1 day prior to the data collection process. The principal investigator was verifying the data during the data collection and every questionnaire was checked every day after data collection before data entry. Data was kept in the form of a file in a private secure place and confidentiality of respondents was ensured by not recording names or any personal identity. The transcripts of qualitative data were shared with research participants to confirm the verbatim accurately reflected their experiences. The data was assured by an expert from the department of social and administrative pharmacy who confirmed the interpretations accurately. Moreover, a conceptual framework was used to guide the study, methodological triangulation (the data collected in the quantitative part and the qualitative part were compared and contrasted) and more than one investigator was involved in this study. Moreover, to ensure reliability of the qualitative tool or research team credibility, transferability, dependability and confirmability aspects of the research were taken into account . Issues of reflexivity: GT status as an insider The first author’s (GT) “native” status offered both opportunities and limitations for the study. He approached this study as an “ Amharic ” speaker and tradition bearer, a member of the “ Amhara ” elite, and also as a senior pharmacy professional. He was able to use existing networks and contacts within the indigenous institutions and local health officials, thereby gaining access to a very wide cross-section of people. He carefully reflected on how the data collection process influenced his own perceptions, and how other people respond to him. He was also faced with the challenge of being perceived as a powerful individual due to his position as a member of the elite and a senior university lecturer. The use of open-ended questions, as well as informal conversations with informants on topics they themselves raised, were among the ways pursued to mitigate these challenges.
A Community based cross-sectional explanatory sequential mixed methods approach was conducted to assess the use of herbal medicines and associated factors among pregnant women in Debre Tabor Town from September 1 to 30, 2019. The town was selected as the study area because it has unique cultural, environmental landscape, and variety of plant species that do affect the pattern of herbal medicine uses. Debre Tabor town is located in south Gondar zone, northwest Ethiopia, 667 km away from Addis Ababa (the capital city of Ethiopia). Based on the information from Debre Tabour town Administrative Bureau, the town has 10 kebeles and with an estimated total population of 111,029 (Fig. ).
To understand factors that motivate the use of herbal medicine among the study community, a conceptual framework was adapted from literatures . It suggests that herbal medicine uses by pregnant women depends on four core components which includes: sociodemographic factor, obstetric related factors, health service-related factors and attitude towards the use of herbal medicines. This conceptual framework also effectively integrates different factors which may influence the use of herbal medicine by pregnant women (Fig. ).
The source populations were all pregnant women residing in Debre Tabor town, while those pregnant women who were registered at the health extension workers registration book during the data collection period were taken as a study population. The sample size was calculated using a single population proportion formula with the assumption of 95% confidence interval, 5% margin of error and 48.6% prevalence of herbal medicine use among pregnant women and 10% for possible non response was taken to determine a final sample size of 267. By dividing the total number of pregnant of the ten kebele (650) with the total sample size (267), an interval of two was used to select household during the one-month data collection period by using systematic random sampling technique. The first household was selected using the lottery method. Then the next household was selected with an interval of two. If the study participant was not available at the first visit, this household was revisited once the same day or following day. If not available again, the study participant was considered as non-respondent. Pregnant women who were willing to share the information and available at the time of data collection were included in this study. Qualitative Approach: The health extension workers also helped to select 12 participants for one focus group discussion (FGD) and six individuals were selected purposively for in-depth interviews from the FGD based on their knowledge about local herbs and ability to describe their experiences in the focus group.
Data collection was performed by three data collectors (BSC nurses) through interviewer-administered questionnaire. The data collectors were properly trained on the purpose of the study, the content of the questionnaire, interviewing techniques, how to approach the respondents, and securing their permission for interview prior to the data collection process. The data collection tool was adapted from different literature after review of the published studies and prepared in English. This was translated to local language ( Amharic ) and then back translated to English in order to ensure consistency. The data collection instrument was pretested on 26 pregnant women who were not included in the final analysis and relevant modifications were done before the commencement of actual data collection. The final questionnaire constituted 33 items that were divided into five main parts . The first section includes questions about socio-demographic data, the second section includes questions about related to the use of herbal medicines, the third section includes questions about some obstetric related factors, the fourth section includes questions about accessibility of health facility and finally, the last section assessed the attitude of sample population regarding use of herbal medicine during pregnancy with five Likert-scale including seven questions. Each question had five choices (ranging from strongly agree to strongly disagree). Item analysis was done, and the internal consistency reliability had a Cronbach’s alpha of 0.82. The answer scores for each question choice and question in both groups were added up and their means and standard deviations calculated. The question scores ranged from 7 to 34 representing the most negative and the most positive attitude respectively. The higher the question score, the more positive the attitude is. After computing the median of all respondents’ responses, the median score of each respondent was dichotomized as have a positive attitude or negative attitude. A score of ≥19 was defined as a “positive attitude towards the use of herbal medicines during pregnancy,” and a score of < 19 was defined as “negative attitude towards the use of herbal medicines during pregnancy.” Positive attitude ≥median. Negative attitude <median. In this study, respondents were considered as herbal medicine users if they have taken herbal medicine(s) through oral, intra-vaginal or topical routes during gestational period. Other preparations that are consumed as routine meal preparations and those that are taken as nutrients were excluded. The qualitative data collection was conducted using an interview guide through probing questions by the principal investigator. Interviews lasted 30–90 min using an audio recorder as well as a filled note was taken by one trained note taker. Participants were briefed about the aim of the study by the principal investigator. Verbal informed consent was obtained from the participants and place of the interview was arranged between the principal investigator and the interviewee, the FGD and in-depth interviews were conducted at the public places in the villages.
Quantitative data was entered into EpiData version 4.2.0.0 and exported to Statistical Package for the Social Sciences (SPSS) software version 25.0 for analysis. Descriptive statistics (frequencies, percentages, mean, and standard deviation) and inferential statistics (bivariate and multivariate analyses) were calculated using bivariate and multivariate logistic regression with a 95% confidence interval (CI). Bivariate logistic regression was used to measure the association between independent variables and herbal medicine use. First bivariate logistic regression was performed to identify candidate variables for multiple logistic regressions. Those variables with a p -value of below 0.25 in the bivariate analysis were fitted to multiple logistic regressions. Model fitness was tested using the Hosmer and Lemeshow’s test and it was insignificant. Multi-collinearity was checked using variance inflation factor (VIF). Covariates with a p-value of below 0.05 in multivariate logistic regression were considered statistically significant factors with the dependent variable (herbal medicine use). Finally, the crude and adjusted odds ratio (OR) with 95% confidence interval (95% CI) were computed to measure the strength of the association between the outcome and the independent variables. A focus group discussion and interviews were audio recorded and transcribed verbatim in Amharic. Texts were read independently by the PI and another professional who speaks the local language and codes were developed in reference to the research questions. Each of the codes were organized into higher-order conceptual themes. These individual codes and themes were discussed at group meetings until consensus was reached on basic themes and subthemes across a focus group and interviews. Finally, the themes were incorporated into a conceptual model of the participants and their use of herbal medicines and associated factors among pregnant women . Sections of original transcripts and key quotes considered to be illustrative of the emerging themes were translated into English to facilitate discussion with the full research team. Data analysis was supported by the use of NVivo 10 computer software.
The quality of the quantitative data was assured by pre-testing the questionnaire on 10% of the sample size (26 pregnant women) in a Town which is different from the study area prior to the start of the actual study to test the fitness of the questionnaire for the study settings. Training about the data collection tool as well as data collection procedures was given to data collectors and supervisors for a total of 1 day prior to the data collection process. The principal investigator was verifying the data during the data collection and every questionnaire was checked every day after data collection before data entry. Data was kept in the form of a file in a private secure place and confidentiality of respondents was ensured by not recording names or any personal identity. The transcripts of qualitative data were shared with research participants to confirm the verbatim accurately reflected their experiences. The data was assured by an expert from the department of social and administrative pharmacy who confirmed the interpretations accurately. Moreover, a conceptual framework was used to guide the study, methodological triangulation (the data collected in the quantitative part and the qualitative part were compared and contrasted) and more than one investigator was involved in this study. Moreover, to ensure reliability of the qualitative tool or research team credibility, transferability, dependability and confirmability aspects of the research were taken into account .
The first author’s (GT) “native” status offered both opportunities and limitations for the study. He approached this study as an “ Amharic ” speaker and tradition bearer, a member of the “ Amhara ” elite, and also as a senior pharmacy professional. He was able to use existing networks and contacts within the indigenous institutions and local health officials, thereby gaining access to a very wide cross-section of people. He carefully reflected on how the data collection process influenced his own perceptions, and how other people respond to him. He was also faced with the challenge of being perceived as a powerful individual due to his position as a member of the elite and a senior university lecturer. The use of open-ended questions, as well as informal conversations with informants on topics they themselves raised, were among the ways pursued to mitigate these challenges.
Quantitative results Socio-demographic characteristics of respondents Out of 267 pregnant women invited to participate, 262 of them completed the survey giving a response rate of 98.2%. The age range of respondents was from 18 to 46 with a mean of 32.68 years (SD = ± 6.47 years). Of the total number of respondents, 178 (67.9%) of the study participants were a follower of Orthodox Christian followed by Muslims 66 (25.2%). In terms of educational level, 69 (26.3%) of respondents had completed secondary school. The socio-demographic and pregnancy related characteristics of respondents are summarized in Table . Prevalence and reasons of herbal medicine use during pregnancy The prevalence of herbal medicine use among pregnant women in Debre Tabor Town was 95 (36.3%), with more than half of them (54.7%) used in the third trimester (Fig. ). The most common reason for herbal medicines used during pregnancy was the ease of availability when they need them ( n = 80, 84.2%). Similarly, the most common reason for not using herbal medicines during pregnancy among non-users was not properly processed ( n = 119, 71.3%) (Table ). Regarding respondent’s discussion with health care providers (HCPs) about HM use during pregnancy, the majority of the respondents ( n = 90, 94.7%) didn’t disclose their use of herbal medicines with health care providers, only few of them ( n = 5, 5.3%) discussed use of herbal medicines with doctors/midwives. The most common reason for the non-disclosure was doctors/midwives did not ask this query ( n = 47, 52.2%). With regard to the source of information about the use of herbal medicines during pregnancy, most participants (69%) used herbal medicines based on advice from family/friends (Fig. ). Factors associated with the use of herbal medicines during pregnancy Results of the bivariate analysis showed that age group, educational status, previous use of HM, presence of health problems, drug availability, distance to health facilities (HFs), and attitude towards the use of HMs as candidates for multivariate analysis at p -value < 0.25 (Table ). Accordingly, educational status ( p -value ≤0.05), previous use of herbal medicines ( p -value ≤0.05), presence of health problems ( p -value ≤0.005), drug availability ( p -value ≤0.001) and distance to the health facilities ( p -value ≤0.001) were found to have a significant association in multivariate logistic regression analysis (Table ). The odds of using herbal medicines during pregnancy who can’t read and write were 9.32 times higher than those women who attend more than diploma (AOR: 9.32 95% CI ((2.34, 37.10)). Pregnant women who were with previous experience of using herbal medicines were 3.14 times more likely to use herbal medicine as compared to those who hadn’t previous experience with herbal medicines (AOR: 3.14, 95% CI ((1.38–7.16). Respondents who had health problems were 3.26 times higher than those who hadn’t health problems to use herbal medicines (AOR: 3.26, 95% CI: (1.50–7.09). The odds of using herbal medicines during pregnancy were 6.15 times higher than for those who residing greater than or equal to 5kms from home to the nearest health facility as compared to those women who residing less than 5kms (AOR:6.15, 95% CI: (2.49–15.23). There was also significant association between the use of herbal medicines during pregnancy and drug availability in the health facilities. The odds of using herbal medicine during pregnancy were 9.87 folds higher if drugs were not available as compared to if drugs were available in the health facility (AOR: 9.87, 95% CI: (4.32–22.55). In this study there was no significant association between HM use during pregnancy and age group, residence and respondents’ attitude towards the use of herbal medicines. Qualitative findings The qualitative study was conducted to elicit information about the use of herbal medicine during pregnancy. Focus group was done with one group of 12 pregnant women and 6 individuals for in-depth interviews from FGD. Their ages ranged from 22 to 42 years. Two major themes were emerged in qualitative data analysis. These were reported as facilitators of herbal medicine use and commonly used herbal medicines. Facilitators for the use of herbal medicines during pregnancy Major reasons mentioned by respondents as facilitators were: cultural beliefs to strengthening the pregnancy, previous experience with herbal medicines, distance to modern healthcare, beliefs that herbal medicines are effective in treating many ailments, presence of health problems, and dissatisfaction with modern health service. Cultural beliefs to strengthening the pregnancy The participants belief that once someone is pregnant, she needs herbal medicine to ‘strengthen the pregnancy. By strengthening, the women meant preventing the pregnancy from miscarriage. Herbal medicine played a very central role in the care of pregnancy because it was believed to stabilize the pregnancy during the early period. As one respondent in the in-depth interview said; “ I don’t know but others say that during the fourth and half month they used this medicine and they call it strengthened in order to prevent a miscarriage” (Participant #2) Previous experience with herbal medicine use Some of the respondents in the FGD had previous exposure to herbal medicines. After evaluating the effect of the herbal medicines, they decided to use them again. Because they perceive herbs were natural and safer than conventional medicines. As one respondent in the in-depth interview noted; “I had taken traditional medicines before this time and I have checked its recuperation ability for me so if I got a disease, I will not go to the hospital rather use traditional herbal medicine confidentially”. ( Participant #6) Distance to modern healthcare Focus group participants stated that during pregnancy, especially after the third trimesters, they usually experience weakness or tiredness. Therefore, the distance between their home and health facility had a decisiveness role in using and not using herbal medicines. If the distance between home and health facility was slightly far it was difficult to go to the hospital. So, when they felt some illness, they used herbs from backyards. A respondent from the in-depth interview had to say: “Occasionally, I used cultural medicine when I got sickened. Because the distance between my home and health center is so far, I have faced tiredness while I have gone to the health facility, especially during my pregnant situation. Therefore, I used traditional treatment like leaves”. ( Participant #1) Beliefs that herbal medicines are effective in treating many ailments Most of the participant in the FGD explained that some diseases, such as yewofbeshita (Herpes Zoster) were treated by only using herbal medicines. Regarding this a participant from FGD had to say: “ The recuperation ability of HM from my sickness over some diseases So I take herbal medicines if I do not relief from my sickness, when I take modern medication” example yewofbeshita” (Participant #3) Presence of health problems Pregnant women with chronic illnesses were quite high in the use of herbal medicines. Most of the pregnant women in the FGD think that herbal medicines were effective to treat chronic diseases. This was strengthened by one of the participants in the in-depth interview as follows, “U knows … . traditional medicines are very important. I use it because I have high blood pressure and physicians order medicine for this disease to use forever throughout my life every day. This is too tedious to take in such ways every day. So I will prefer to take traditional herbal medicines rather than this because I have seen a change when I use it.” ( Participant #5) Dissatisfaction with modern medicine use Some respondents articulated that they were dissatisfied with the result of modern medicine use because health care providers didn’t give them due attention. Regarding this a respondent in the in-depth interview stated: “I am not feeling comfortable in modern medicine once I take it after I have gone to the health center by paying for transportation. I do not feel comfortable in essence, the health professional said that you are ok and simply give Panadol as an analgesic but I was in a series of sick conditions. So as the sickness condition increases, I choose to take herbal medicine, in such a way that I have seen changes within a day” ( Participant #4). Herbal medicines commonly used by pregnant women Participants during focus group discussion indicated the use of different herbal medications to manage some pregnancy related minor ailments, such as nausea and vomiting, abdominal cramp, fever and common cold. Most of the participants agreed that linseed as the most commonly used herb as it was believed to have a facilitator effect of labor. Herbal medicines used by study participants and most common indications were illustrated in Table .
Socio-demographic characteristics of respondents Out of 267 pregnant women invited to participate, 262 of them completed the survey giving a response rate of 98.2%. The age range of respondents was from 18 to 46 with a mean of 32.68 years (SD = ± 6.47 years). Of the total number of respondents, 178 (67.9%) of the study participants were a follower of Orthodox Christian followed by Muslims 66 (25.2%). In terms of educational level, 69 (26.3%) of respondents had completed secondary school. The socio-demographic and pregnancy related characteristics of respondents are summarized in Table . Prevalence and reasons of herbal medicine use during pregnancy The prevalence of herbal medicine use among pregnant women in Debre Tabor Town was 95 (36.3%), with more than half of them (54.7%) used in the third trimester (Fig. ). The most common reason for herbal medicines used during pregnancy was the ease of availability when they need them ( n = 80, 84.2%). Similarly, the most common reason for not using herbal medicines during pregnancy among non-users was not properly processed ( n = 119, 71.3%) (Table ). Regarding respondent’s discussion with health care providers (HCPs) about HM use during pregnancy, the majority of the respondents ( n = 90, 94.7%) didn’t disclose their use of herbal medicines with health care providers, only few of them ( n = 5, 5.3%) discussed use of herbal medicines with doctors/midwives. The most common reason for the non-disclosure was doctors/midwives did not ask this query ( n = 47, 52.2%). With regard to the source of information about the use of herbal medicines during pregnancy, most participants (69%) used herbal medicines based on advice from family/friends (Fig. ). Factors associated with the use of herbal medicines during pregnancy Results of the bivariate analysis showed that age group, educational status, previous use of HM, presence of health problems, drug availability, distance to health facilities (HFs), and attitude towards the use of HMs as candidates for multivariate analysis at p -value < 0.25 (Table ). Accordingly, educational status ( p -value ≤0.05), previous use of herbal medicines ( p -value ≤0.05), presence of health problems ( p -value ≤0.005), drug availability ( p -value ≤0.001) and distance to the health facilities ( p -value ≤0.001) were found to have a significant association in multivariate logistic regression analysis (Table ). The odds of using herbal medicines during pregnancy who can’t read and write were 9.32 times higher than those women who attend more than diploma (AOR: 9.32 95% CI ((2.34, 37.10)). Pregnant women who were with previous experience of using herbal medicines were 3.14 times more likely to use herbal medicine as compared to those who hadn’t previous experience with herbal medicines (AOR: 3.14, 95% CI ((1.38–7.16). Respondents who had health problems were 3.26 times higher than those who hadn’t health problems to use herbal medicines (AOR: 3.26, 95% CI: (1.50–7.09). The odds of using herbal medicines during pregnancy were 6.15 times higher than for those who residing greater than or equal to 5kms from home to the nearest health facility as compared to those women who residing less than 5kms (AOR:6.15, 95% CI: (2.49–15.23). There was also significant association between the use of herbal medicines during pregnancy and drug availability in the health facilities. The odds of using herbal medicine during pregnancy were 9.87 folds higher if drugs were not available as compared to if drugs were available in the health facility (AOR: 9.87, 95% CI: (4.32–22.55). In this study there was no significant association between HM use during pregnancy and age group, residence and respondents’ attitude towards the use of herbal medicines.
Out of 267 pregnant women invited to participate, 262 of them completed the survey giving a response rate of 98.2%. The age range of respondents was from 18 to 46 with a mean of 32.68 years (SD = ± 6.47 years). Of the total number of respondents, 178 (67.9%) of the study participants were a follower of Orthodox Christian followed by Muslims 66 (25.2%). In terms of educational level, 69 (26.3%) of respondents had completed secondary school. The socio-demographic and pregnancy related characteristics of respondents are summarized in Table .
The prevalence of herbal medicine use among pregnant women in Debre Tabor Town was 95 (36.3%), with more than half of them (54.7%) used in the third trimester (Fig. ). The most common reason for herbal medicines used during pregnancy was the ease of availability when they need them ( n = 80, 84.2%). Similarly, the most common reason for not using herbal medicines during pregnancy among non-users was not properly processed ( n = 119, 71.3%) (Table ). Regarding respondent’s discussion with health care providers (HCPs) about HM use during pregnancy, the majority of the respondents ( n = 90, 94.7%) didn’t disclose their use of herbal medicines with health care providers, only few of them ( n = 5, 5.3%) discussed use of herbal medicines with doctors/midwives. The most common reason for the non-disclosure was doctors/midwives did not ask this query ( n = 47, 52.2%). With regard to the source of information about the use of herbal medicines during pregnancy, most participants (69%) used herbal medicines based on advice from family/friends (Fig. ).
Results of the bivariate analysis showed that age group, educational status, previous use of HM, presence of health problems, drug availability, distance to health facilities (HFs), and attitude towards the use of HMs as candidates for multivariate analysis at p -value < 0.25 (Table ). Accordingly, educational status ( p -value ≤0.05), previous use of herbal medicines ( p -value ≤0.05), presence of health problems ( p -value ≤0.005), drug availability ( p -value ≤0.001) and distance to the health facilities ( p -value ≤0.001) were found to have a significant association in multivariate logistic regression analysis (Table ). The odds of using herbal medicines during pregnancy who can’t read and write were 9.32 times higher than those women who attend more than diploma (AOR: 9.32 95% CI ((2.34, 37.10)). Pregnant women who were with previous experience of using herbal medicines were 3.14 times more likely to use herbal medicine as compared to those who hadn’t previous experience with herbal medicines (AOR: 3.14, 95% CI ((1.38–7.16). Respondents who had health problems were 3.26 times higher than those who hadn’t health problems to use herbal medicines (AOR: 3.26, 95% CI: (1.50–7.09). The odds of using herbal medicines during pregnancy were 6.15 times higher than for those who residing greater than or equal to 5kms from home to the nearest health facility as compared to those women who residing less than 5kms (AOR:6.15, 95% CI: (2.49–15.23). There was also significant association between the use of herbal medicines during pregnancy and drug availability in the health facilities. The odds of using herbal medicine during pregnancy were 9.87 folds higher if drugs were not available as compared to if drugs were available in the health facility (AOR: 9.87, 95% CI: (4.32–22.55). In this study there was no significant association between HM use during pregnancy and age group, residence and respondents’ attitude towards the use of herbal medicines.
The qualitative study was conducted to elicit information about the use of herbal medicine during pregnancy. Focus group was done with one group of 12 pregnant women and 6 individuals for in-depth interviews from FGD. Their ages ranged from 22 to 42 years. Two major themes were emerged in qualitative data analysis. These were reported as facilitators of herbal medicine use and commonly used herbal medicines.
Major reasons mentioned by respondents as facilitators were: cultural beliefs to strengthening the pregnancy, previous experience with herbal medicines, distance to modern healthcare, beliefs that herbal medicines are effective in treating many ailments, presence of health problems, and dissatisfaction with modern health service.
The participants belief that once someone is pregnant, she needs herbal medicine to ‘strengthen the pregnancy. By strengthening, the women meant preventing the pregnancy from miscarriage. Herbal medicine played a very central role in the care of pregnancy because it was believed to stabilize the pregnancy during the early period. As one respondent in the in-depth interview said; “ I don’t know but others say that during the fourth and half month they used this medicine and they call it strengthened in order to prevent a miscarriage” (Participant #2)
Some of the respondents in the FGD had previous exposure to herbal medicines. After evaluating the effect of the herbal medicines, they decided to use them again. Because they perceive herbs were natural and safer than conventional medicines. As one respondent in the in-depth interview noted; “I had taken traditional medicines before this time and I have checked its recuperation ability for me so if I got a disease, I will not go to the hospital rather use traditional herbal medicine confidentially”. ( Participant #6)
Focus group participants stated that during pregnancy, especially after the third trimesters, they usually experience weakness or tiredness. Therefore, the distance between their home and health facility had a decisiveness role in using and not using herbal medicines. If the distance between home and health facility was slightly far it was difficult to go to the hospital. So, when they felt some illness, they used herbs from backyards. A respondent from the in-depth interview had to say: “Occasionally, I used cultural medicine when I got sickened. Because the distance between my home and health center is so far, I have faced tiredness while I have gone to the health facility, especially during my pregnant situation. Therefore, I used traditional treatment like leaves”. ( Participant #1)
Most of the participant in the FGD explained that some diseases, such as yewofbeshita (Herpes Zoster) were treated by only using herbal medicines. Regarding this a participant from FGD had to say: “ The recuperation ability of HM from my sickness over some diseases So I take herbal medicines if I do not relief from my sickness, when I take modern medication” example yewofbeshita” (Participant #3)
Pregnant women with chronic illnesses were quite high in the use of herbal medicines. Most of the pregnant women in the FGD think that herbal medicines were effective to treat chronic diseases. This was strengthened by one of the participants in the in-depth interview as follows, “U knows … . traditional medicines are very important. I use it because I have high blood pressure and physicians order medicine for this disease to use forever throughout my life every day. This is too tedious to take in such ways every day. So I will prefer to take traditional herbal medicines rather than this because I have seen a change when I use it.” ( Participant #5)
Some respondents articulated that they were dissatisfied with the result of modern medicine use because health care providers didn’t give them due attention. Regarding this a respondent in the in-depth interview stated: “I am not feeling comfortable in modern medicine once I take it after I have gone to the health center by paying for transportation. I do not feel comfortable in essence, the health professional said that you are ok and simply give Panadol as an analgesic but I was in a series of sick conditions. So as the sickness condition increases, I choose to take herbal medicine, in such a way that I have seen changes within a day” ( Participant #4).
Participants during focus group discussion indicated the use of different herbal medications to manage some pregnancy related minor ailments, such as nausea and vomiting, abdominal cramp, fever and common cold. Most of the participants agreed that linseed as the most commonly used herb as it was believed to have a facilitator effect of labor. Herbal medicines used by study participants and most common indications were illustrated in Table .
The present study determined the prevalence and factors associated with the use of herbal medicines during pregnancy among 262 women in Debre Tabor Town. The finding of this study reported that the prevalence of herbal medicines use among pregnant women is 36.3%. This finding is lower than reports from Zimbabwe (69.9%), Iraq (56.7%) and Hossana, Southern Ethiopia (73.1%) . The lower prevalence of HMs use in our study might have been due to the difference in the populations studied, sample size difference, the time of the study and differences in socio-cultural contexts. A study conducted in Zimbabwe reported that Zimbabwean culture and traditions encourage pregnant women to use traditional medicines to either treat pregnancy-related illnesses or to facilitate delivery as they are believed to be safe . However, our finding is higher than the prevalence reported in Kenya (12%) and Northern Uganda (20%) . The possible justification for the difference might be due to differences in accessibility, affordability and socio-cultural context. This study also found that family/friends were the most frequently cited source of information about the use of herbal medicines and users tended to trust the benefits of use if recommended by close acquaintances. Family and friends represent the social and cultural environment in which pregnant women live and in part influence to their use of herbal medicines during pregnancy which is similar to another finding conducted in Nairobi, Kenya . This could be linked to the very well-constructed social capital values (social support) of the study communities. We also discovered the degree of disclosure between herbal medicine users and their health care providers. The result was alarming because only 5.3% of pregnant women disclosed herbal medication use with their doctors. More than 90% of the respondents did not discuss use of herbal medicines with health care providers. The reasons for non-disclosures were: doctors/midwives did not ask (40.3%), forget to inform (36%), afraid of doctors/midwives’ response (15.1%) and it was not important to talk (8.6%). This finding was in line with a survey conducted on Iraqi women, who stated that doctors did not ask (50.53%) and afraid of a doctor’s response (5.3%) were perceived hindrances for not reporting . Moreover, the study conducted in Nekemt , Ethiopia stated only 14.29% of the women reported to have received health advices from healthcare workers . The lack of communication between the health care providers and pregnant women who are using herbal medicine may have a harmful effect on the mother as well as the fetus. Therefore, health care providers should advise about the harmful effects of taking herbal medicines during ANC visit . The odds of herbal medicine use during pregnancy were 6.15 times higher for women living more than 5kms from the nearest health facility than those who live less than 5kms. This finding was in line with the study conducted in Gonder , Ethiopia . Moreover, our qualitative finding supported this result; respondents explained that if the distance between their home and health facility was slightly far it was difficult for them to go to the hospital and they resort to use herbs from the backyard. However, this finding was different from the study in Northern Uganda, which reported that there was not a significant association between distance from the health facilities and the use of herbal medicine during pregnancy . The difference may be due to a variation in transportation access in the two countries. This study also indicated that the use of herbal medicines during pregnancy was 9.87 times higher, if drugs are not availability in the health facility. This finding was in line with the study conducted in Nairobi, Kenya . However, a study conducted in Northern Uganda reported there was no significant association between the use of herbal medicines during pregnancy and the availability of drugs in the health facility ( p -value = 0.08) . This may imply that the unavailability of medicines in the health facility necessitates the use of herbal medicines among pregnant women and their existence has influenced the result significantly. Pregnant women who were illiterate (cannot read and write) were 9.32 times more likely to use herbal medicine as compared to those who did attend diploma and above. Similar findings were reported from the study conducted in Gondar and Nekemte , Ethiopia; and Ghana which stated that use of herbal medicines during pregnancy and educational status has significantly related . This may be due to the fact that as they become more educated the rate of herbal drug usage was decreased, because they are likely to know the side effects of herbal medicine usage during pregnancy . This study also found that the odds of using herbal medicines during pregnancy were 3.14 times higher to those participants who had previous experience of herbal medicines use as compared to those who hadn’t prior use. Similarly, a study conducted in Saudi reported that there was a significant association between the use of herbal medicines during pregnancy and prior use . Our qualitative finding also articulated prior use of herbal medicines, by pregnant women, was facilitator of herbal medicines use. After assessing and reassessing the effect of the herbal medicines they decided to use them again. Because they perceived herbs were natural and safer than conventional medicines. The findings also indicated that respondents who had health problems were 3.26 times more likely to use herbal medicines during pregnancy than those who hadn’t health problems. This result was similar to the study conducted in Gondar , Ethiopia . This was also augmented by our qualitative finding; which reported that pregnant women with chronic illnesses were quite high in the use of herbal medicines. Respondents in the FGDs mentioned that Ginger,Tenaadam , Telba , NechiBahrzaf , and Moringaas a specific example of herbal medicine use during pregnancy. The pattern of herbal medicine use in our study was almost similar to the study done in Hosanna town, southern Ethiopia, where garlic, ginger, tenaadam , and damakasse were reported to be the commonest herbs used by pregnant women However, a study conducted in Norway reported Echinacea, chamomile, cranberry, and iron-rich herbs as traditional medicines used by pregnant women . The difference in patterns across different countries may be due to differences in accessibility and geographical distribution of herbs. In this study respondents mentioned facilitate labour, pneumonia, nausea and vomiting, abdominal cramp, fever, common cold, hypertension, as a specific example of indications treated with herbal medicines during pregnancy. This finding was similar to the study conducted in Bangladesh . Strength and limitation of the study This study was used mixed-method approaches that provide a better understanding of the use of herbal medicines during pregnancy. The principal investigator was native to the study community and this minimizes linguistic and cultural barriers, otherwise, insider bias. Like all self-reported exposure assessments, under reporting is very likely. As it is cross-sectional, it fails to show seasonal variability in the use of herbal medicine Moreover, the study was not able to look at the effectiveness or safety and side effects of the herbal medicines that were mentioned.
This study was used mixed-method approaches that provide a better understanding of the use of herbal medicines during pregnancy. The principal investigator was native to the study community and this minimizes linguistic and cultural barriers, otherwise, insider bias. Like all self-reported exposure assessments, under reporting is very likely. As it is cross-sectional, it fails to show seasonal variability in the use of herbal medicine Moreover, the study was not able to look at the effectiveness or safety and side effects of the herbal medicines that were mentioned.
Herbal medicine use during pregnancy was a common experience, and it’s linked to educational status, prior use of herbal medicines, drug availability, presence of health problems and distance to the health facilities. Pregnant women depend mainly on family/ friends as a source of information about herbal medicine use. Ginger ( Zingiber officinale ),Tenaadam ( Ruta chalepensis ), Telba (Linumusitatissimum), NechiBahrzaf (Eucalyptus globulus) and Moringa (Moringa stenopetala were the most commonly used herbs among pregnant women, and the most popular indication) and the most common indication for use were pneumonia, nausea and vomiting, abdominal cramp, fever, common cold, and hypertension. Given the high prevalence of herbal medicine and the low rate of disclosure, health care professionals should be willing to explore herbal medicine use with their pregnant patients, since it will result in a better health outcome. Moreover, a detailed study on commonly used herbs to establish the efficacy, safety and side effects of these herbs to ensure the well-being of the mother and foetus would be recommended.
Additional file 1.
|
Single-arm trials for domestic oncology drug approvals in China | bd4e9557-aaa4-4c9e-ad98-5df890520172 | 10690876 | Internal Medicine[mh] | We conducted a retrospective analysis of all new Chinese domestic oncology drugs that received regulatory approval on the basis of SATs between 2018 and 2022. As of December 31, 2022, the National Medical Products Administration (NMPA) had granted approval for 81 new oncology indications for domestic drugs. Of those, 34 (42%) were established through SATs : 21 (62%) of which pertained to new molecular entities (NMEs) or original biologics, and 13 (38%) of which pertained to supplementary indications. Notably, 97% of SAT approvals (33 of 34) used the objective response rate (ORR) or complete response (CR) as the primary endpoint . Major cytogenetic response (McyR) and major hematologic response (MaHR) were achieved for chronic phase and accelerated phase chronic myeloid leukemia (CML). Approvals had varying disease sites, and treatments for lymphomas and lung tumors were those most frequently approved. Lymphomas accounted for nearly one-third (32%) of approvals, and were closely followed by lung cancers (12%). The dawn of pan-tumor approvals arrived in 2020 with envafolimab, an anti-programmed cell death protein 1/programmed death-ligand 1 (PD1/PDL1) antibody, which was followed by 3 approvals for microsatellite instability high (MSI-H)/deficient mismatch repair (dMMR) solid tumors. Immune checkpoint inhibitors, particularly PD1/PDL1, constituted the largest category (53%) and included the world’s first approval for a PD1/cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) dual antibody (cadonilimab). Tyrosine kinase inhibitors (TKIs) played a critical role in the treatment of lung and lymphatic system cancers, and included Bruton’s tyrosine kinase (BTK), anaplastic lymphoma kinase (ALK), epidermal growth factor receptor (EGFR), BCR-ABL1, and C-met. The approval landscape also included 2 antibody-drug conjugates (ADCs), 2 poly ADP ribose polymerase (PARP) inhibitors, and 2 chimeric antigen receptor T-cell (CAR-T) therapies.
SATs play a critical role in expediting drug development and providing early treatment accessibility. However, the use of SATs requires a careful assessment of the delicate balance between therapeutic benefits and potential risks. The limitations of SATs introduce an element of uncertainty that requires subsequent confirmatory studies. To address this uncertainty, mitigate effects on the benefit-risk assessment, and ensure the responsible use of SATs, Center for Drug Evaluation (CDE) of NMPA (China) issued the “Technical Guidelines on the Applicability of Single Arm Clinical Trials for Use in Support of Oncology Drug Marketing Applications” in March of 2023 . This guidance rigorously defines the conditions under which SATs can be applied, outlining 6 key application scenarios in which their use may be appropriate . Two of these scenarios relate to unmet medical needs, addressing cases in which no available therapy exists within the study population or for rare cancers. In our study, “no available therapy” was defined as the absence of approved treatment options or standard therapies at the time of submission of a new drug application (NDA). For indications with more than 2 approved treatments, drugs submitted before the first drug approval for the same indication are considered contemporaneous and are subject to the same assessment criteria. However, if no approved treatment has undergone confirmatory studies, the treatment is categorized as lacking available therapeutic options. Of the total 19 indications we analyzed, 10 indications, representing 20 drugs, had no available therapy at the time of NDA submission . Four drugs (3 for urothelial cancer and 1 for gastric cancer) had existing treatments that had limited efficacy, as seen in cases such as second-line treatment for advanced urothelial carcinoma or third-line treatment for advanced gastric cancer, in which standard chemotherapy has limited efficacy. In addition, 10 drugs were submitted when imported drugs had already received marketing approval, but domestic alternatives were not yet available, thus highlighting the urgent need for domestically available alternatives. In cases of rare cancer, most proposed indications (74%, 25 of 34) are for rare cancers, and some (12%, 4 of 34) are for targeted populations enriched with specific biomarkers [such as EGFR T790M or human epidermal growth factor receptor 2 (HER2)], which inherently have a small patient population. Only a minority (15%, 5 of 34) of the proposed indications were for non-rare cancers. The remaining 4 scenarios are more closely associated with clear clinical benefit evidence, encompassing well-defined therapeutic mechanisms, outstanding efficacy, clear external control data, and controllable safety risk. Outstanding efficacy is particularly important with respect to historical controls with robust evidence from meta-analyses, systematic reviews, real-world data, and other sources of evidence-based medicine. Our study findings revealed that all approved single-arm studies used sample size calculations based on statistical assumptions, and specifically used historical control data as the reference for the null hypothesis regarding ORR values . Sample assumptions varied across tumors. The median sample size in approved single-arm studies was 102 (minimum: 28, maximum: 249). The sample size of most rare cancers were below 100. Most analyzed drugs (82%, 28 of 34) had preset null hypothesis ORR values, in accordance with the required efficacy standard set by the NMPA, whereas a minority (18%, 6 of 34) set the null hypothesis values above than those required by the NMPA. Further analysis revealed that the lower limits of the 95% confidence intervals for the actual efficacy values of all drugs were consistently higher than their respective null hypothesis values, and nearly half (41%, 14 of 34) of the drugs demonstrated lower limits of the 95% confidence intervals for actual efficacy values exceeding the statistically estimated alternative hypothesis regarding the ORR. Intriguingly, half (50%, 17 of 34) of the drugs had actual efficacy values surpassing twice the NMPA’s required efficacy. In general, the standard set by the NMPA would be the same, or within a similar range, for the same indication, but might vary by tumor type. Moreover, a general, gradual, and modest increase in the ORR over time was observed within the same indication . Some individual cases showed lower ORR values than those of previously approved drugs, most of which were drugs developed contemporaneously with similar evaluation criteria. For instance, serplulimab, used as a second-line treatment for MSI-H or dMMR colorectal cancer (CRC) and/or solid tumors, exhibited slightly lower efficacy than envafolimab and tislelizumab, which were developed in parallel. However, the efficacy of serplulimab still surpassed the NMPA required efficacy by more than twofold. After categorization by origin and nature of the tumors, we observed that the ORR treatment efficacy for hematological malignancies exceeded that for solid tumors, in agreement with clinical experience. Thus, most domestic drugs approved through single-arm studies exhibit clinically significant improvements over previous treatment approaches. Because most domestic drugs are concentrated in PD1/PDLI related immune checkpoint inhibitors or imported drugs of the same class that have been approved in China, the relevant drug treatment mechanism is relatively clear, and known safety-associated risks are also disclosed for imported drugs. The main adverse reaction characteristics of domestically approved PD1 drugs are similar to those of similar products or those in other indications. The safety profiles of “first-in-class” drugs in China, such as the C-met kinase inhibitor (savolitinib) and HER2-ADC (disitamab vedotin), align with those reported in international clinical studies. Similarly, in the case of global “first-in-class” drugs, no significant difference in adverse reactions has been observed for PD1/CTLA-4 dual antibody (cadonilimab), compared with PD1 and CTLA-4 monotherapies. All domestic drugs have a corresponding risk control plan at the time of NDA submission, coupled with an extensive regime of post-marketing safety surveillance; consequently, the overall safety risks of domestic oncology drugs are generally controllable.
Drugs granted conditional approval on the basis of SATs in oncology often undergo subsequent confirmatory trials, which generally fall into 3 distinct categories: randomized controlled trials (RCTs), extended single-arm studies with larger sample sizes, and real-world studies. We found that only 1 drug, relmacabtagene autoleucel, which was recently approved as a third-line treatment for r/r FL, has not yet initiated its confirmatory trial, whereas all other drugs in our study have commenced confirmatory trials . A substantial majority (85%, 28 of 33) of the approved drugs had RCTs selected as the chosen approach for the confirmatory trials. Within this subset, less than half (39%, 11 of 28) used active controls, several (32%, 9 of 28) involved comparison with standard care, and the remaining were placebo controlled; almost all studies (96%, 27 of 28) used a primary clinical endpoint of progression-free survival (PFS), whereas only 1 study [of disitamab vedotin used as a third-line treatment for HER2 GC/GEJC] used OS as the primary endpoint. In contrast, for pan-tumor studies, indications such as MSI-H or dMMR solid tumors, or for studies including rare cancer types, such as MET Ex14 skipping NSCLC or r/r DLBCL, the design of confirmatory studies tends to favor extended single-arm studies essentially constituting a continuation of the original study. Notably, we identified no examples of use of real-world evidence as the basis for confirmatory trials among approved domestic drugs. In addition, a substantial proportion (71%, 24 of 34) of participants in these confirmatory trials moved into the front-line treatment population. The remaining cases involved continuation of pivotal studies for specific indications, such as MSI-H or dMMR solid tumors, as well as the maintenance of r/r cHL, r/r DLBCL and CML, or HER2 GC/GEJC regimens in line with previous regimens. Notably, nearly half (48%, 14 of 29) of the approved drugs had carefully planned confirmatory trials in the development phase, which began before the formal NDA submission. As of July 2023, 9 domestic oncology drugs have converted full approvals, defined as successful completion of confirmatory studies and submitting NDAs or supplemental applications. Additionally, tislelizumab, intended as a first-line treatment for hepatocellular carcinoma (HCC), is currently undergoing review, and its NDA was submitted in December 2022 . All 10 of these drugs were initially granted conditional approval for populations with no available therapy or imported alternatives, followed by conducting confirmatory studies in front-line populations, and ultimately had fully approved conversions and new approvals for their front-line indications. Notably, the initiation of the confirmatory trials of these drugs was conducted well, and occurred before the NDA submission for the previous indication. The time for full approval conversion ranged from as short as 2 months (for camrelizumab, used as a first-line treatment for NPC) to no longer than 4.6 years (for zanubrutinib, used as a first-line treatment for r/r CLL/SLL).
Domestic oncology drugs supported by SATs for approval in China have all been granted conditional approval in the past 5 years. Considering the timeframe for completing post-marketing confirmatory trials within 5 years of being granted conditional approval, we anticipate that a wave of drugs might be submitted for post-approval supplementary applications in the next 2 years. Our analysis suggested that several companies might be at risk of delays in filing these supplementary applications. According to the FDA’s experience, approximately 50% of drugs completed confirmatory clinical trials, and 12% of indications were eventually withdrawn. The median time for conversion to full approval was 3.1 years, and the timing of completion of confirmatory trials was influenced by whether a confirmatory trial was initiated at the time of accelerated approval. Consequently, the FDA recommends early discussions regarding the design and initiation of confirmatory trials, preferably initiated before the time of application for accelerated approval or the completion of most patient enrollment . The FDA may also withdraw conditional approval because of failure or delay of the completion of a confirmatory trial. However, in China, procedures for conditional drug approval were recently released in August 2023 and are currently in the phase of solicitation of opinions . This guidance clarified the completion timeline for confirmatory studies, established comprehensive evaluation criteria, and outlined the process for withdrawal or conversion to full approval. The overarching concept aligns with that of the FDA, which is to withdraw drugs that do not meet the specified conditions. In China, certain criteria may be more stringent; notably, only a single drug with the same mechanism is permitted to receive conditional approved for a specific indication.
Effective post-marketing risk management is particularly important for drugs approved on the basis of SATs. Because of limitations in trial design, and constraints on the premarket sample size and observation period, fully identifying the safety risks of a product might not be possible. These risks include a low incidence of adverse reactions, potential long-term safety risks, or a risk of drug ineffectiveness. Consequently, certain drugs may require extended safety monitoring after market approval. For instance, relmacabtagene autoleucel, a cell gene therapy, requires a 15-year follow-up period to observe potential long-term oncogenic risks. The assessment of therapeutic effectiveness in SAT often relies on comparisons with historical control data. However, achieving a consistent baseline between a historical control group and SAT participants can be challenging. Factors such as demographic variables, comorbidities, disease characteristics (such as severity, symptoms, and progression), initiation of treatment follow-up, concurrent therapies, and collected clinical observations can introduce biases into the evaluation of therapeutic efficacy. To address these issues, the FDA encourages the inclusion of real-world studies as a valid external control, and supplementary evidence from SATs to substantiate the safety and efficacy of a drug . Nonetheless, in designing external control trials, careful consideration must be paid to factors that ensure the comparability of data with the SAT, including the timing and frequency of data collection, therapeutic regimen, healthcare practices, and criteria for assessing patient outcomes. The acceptance of the inherent risk of uncertainty associated with SAT changes as post-marketing evidence accumulates, and the criteria for the use of SATs should be adjusted periodically in light of emerging evidence, as exemplified in April 2021, when the FDA’s Oncology Drugs Advisory Committee reevaluated 6 PI3K inhibitors that had initially received accelerated approval through single-arm pathways. Some of these drugs were withdrawn from the market because of safety concerns, and subsequent PI3K inhibitors were required to undergo RCT studies for approval . Correspondingly, recent guidelines for SATs issued by the China’s CDE emphasize that the applicable conditions for supporting anti-tumor drug approval in SATs should be “patient-centered” and “clinically value-oriented” , , . Moreover, the acceptance of uncertainty risks must adapt to the evolving landscape of drug development and clinical practice, while ensuring that the benefits consistently exceed the uncertainties, to ultimately optimize the benefit-risk ratio.
Click here for additional data file.
|
Молекулярные онкологические консилиумы и тераностика | 8954edba-cb80-487a-adb1-6513ff6a80a7 | 9939957 | Internal Medicine[mh] | В первую очередь на МОК должны обсуждаться те случаи опухоли, терапия которых в соответствии с клиническими рекомендациями и доказательным опытом неизвестна или недостаточно эффективна. Это столь же актуально при неоднозначном гистологическом заключении или несоответствии его клинической картине и течению болезни. По данным одного из обзоров, включавшего 440 случаев ЗНО с четким гистологическим заключением, необходимость в МОК чаще всего возникала при саркомах (21,4%), раке молочной железы (20%), опухолях головного мозга (15,5%), онкогинекологии (14,1%), раке легкого (7,3%) и колоректальном раке (6,4%). Важно понимать, что для обсуждения на МОК потенциально подходят все типы и локализации ЗНО, причем гистологический тип опухоли не является дискриминирующим критерием. Более того, редкие (орфанные) ЗНО, которые совокупно представляют примерно четверть всех случаев, гораздо более актуальны для обсуждения на МОК, чем часто встречающиеся карциномы. Молекулярный портрет опухоли должен рассматриваться как обязательная интегративная часть гистологического диагноза при ЗНО, у которых отсутствуют стандартизованные рекомендации по терапии.
По данным анализа публикаций, существует множество вариантов молекулярных анализов. В большинстве случаев (57,4%) для обсуждения на МОК выполнялось таргетное мультигенное секвенирование следующего поколения (Targeted Multigene Next Generation Sequencing, MG-NGS), которое дополнялось полноэкзомным секвенированием (Whole-Exome Sequencing, WES) в 16,4%, секвенированием РНК (RNA sequencing) — в 13,1%, матричной сравнительной геномной гибридизацией (Array Comparative Genomic Hybridization, aCGH) — в 4,9%, полногеномным секвенированием (Whole Genome Sequencing, WGS) и секвенированием по Сэнгеру (Sanger sequencing, SS) — по 3,3%, анализом мРНК (mRNA) — в 1,6% случаев. Наиболее подходящим анализом с целью обнаружения полезных в практическом смысле генетических нарушений является таргетное NGS наиболее важных рак-ассоциированных генов, в идеале в сопряжении с анализом на клинически релевантные генетические фьюжины (например, слиянием киназных генов). Этот анализ широко распространен и связан с относительно коротким временем на секвенирование и посильной биоинформатикой.
МОК – важнейший инструмент побуждения «клинико-молекулярных» обсуждений и поощрения коммуникации между коллегами различных специальностей, что значительно ускоряет развитие персонализированной медицины. Как минимум в составе МОК обязательно должны быть следующие ключевые специалисты: клинические онкологи (по данным литературы в 100% МОК), патологоанатомы (90,6%). Также велика роль генетиков (68,7%), особенно при обсуждении вопросов в отношении герминальных мутаций, биоинформатиков (37,5%) и молекулярных биологов (25%), особенно при интерпретации массивных наборов данных в результате использования технологий NGS, например при WGS. Включение в состав МОК специалистов лучевой диагностики и эндоскопии (при необходимости) также очень желательно. Менее часто участвуют фармакологи (21,9%) и специалисты по биоэтике (9,4%), подключение которых особенно актуально при использовании экспериментальных препаратов. В 12,5% в МОК включались клиницисты-исследователи со знанием «молекулярной медицины».
Во многом этот период определяется индивидуальной клинической ситуацией. В реальной клинической практике, по данным литературы, данный срок варьирует от 12 до 86 дней и в среднем составляет 38 дней. Минимальный период ожидания решения МОК в клинической практике, по данным литературы, был в среднем 16 дней. Компромиссный срок проведения МОК с учетом клинической необходимости и времени на проведение масштабных молекулярных исследований находится в пределах 28 дней (4 нед). Разумеется, чем этот период меньше, тем выше шансы на изменение клинической ситуации к лучшему путем назначения эффективной таргетной терапии. Чаще всего причиной задержки молекулярных исследований являются организационные, технические и «человеческие» проблемы, связанные с возможностью получения парафиновых блоков и их транспортировкой в лабораторию, низкое качество тканевого материала в парафине (поздняя передача в лабораторию патанатомии макропрепарата, передерживание его в формалине, использование чистого формалина (а не забуференного раствора), дефекты проводки и заливки, низкое качество парафина), а также его контаминация на этапе пробоподготовки. На качество пробоподготовки в патологоанатомической лаборатории невозможно повлиять обратным порядком, поэтому столь важно оперировать или выполнять биопсию в тех медицинских учреждениях, которые имеют высокую валидированную репутацию, в том числе по части качества парафиновых блоков и морфологической верификации. Помимо этого, важнейшим фактором является контроль взятия на молекулярные исследования клеточного материала из опухоли. Это осуществляется путем контролируемого взятия биологического материала под контролем гистологической верификации источника. Отличным способом уменьшить срок выполнения молекулярно-генетических исследований является так называемый «рефлекс-тест» (reflex test). Этот способ заключается в том, что молекулярно-генетический тест назначается врачом-патологоанатомом уже при изучении гистологических стеклопрепаратов, когда срабатывает «рефлекс» выяснить молекулярный профиль определенного гистологического паттерна (например, рак поджелудочной железы с медуллярным гистологическим строением, который обычно ассоциирован с микросателлитной нестабильностью). В случае наличия биобанка молекулярный анализ может быть ускорен и стать более качественным с учетом того, что криоконсервированный образец опухоли предпочтительнее для молекулярно-генетических исследований, чем парафиновый блок. Идентификация опухолевой природы криообразца осуществляется при помощи гистологического исследования участка, откуда взят образец опухоли для криоконсервации. Довольно сложно оценивать рентабельность молекулярно-ориентированной терапии в онкологии, так как «лавры» всегда достаются таргетному препарату или их комбинации. Сама по себе активность МОК (в основном трудозатраты) составляет не более 5% стоимости молекулярной диагностики, а в стоимости общих лечебно-диагностических затрат занимает около 0,3%. Чаще всего стоимость современных таргетных препаратов намного превышает затраты на молекулярно-генетические исследования и МОК, но без выполнения последних бывает невозможно переломить негативный клинический сценарий. Также молекулярно-генетическое тестирование МОК открывает новые горизонты для включения пациентов в клинические исследования и программы открытого доступа фармкомпаний по новым таргетным препаратам. В отсутствие таковой возможности пациенты, социальные службы и фонды могут инициативно на основании рекомендаций МОК приобретать зарегистрированные в стране противоопухолевые препараты, а также ввозить из-за рубежа по индивидуальным жизненным показаниям. Включение в формализованные (on-label) клинические испытания во всем мире является предпочтительным путем получения таргетной терапии как минимум по двум причинам: При анализе мутаций могут быть обнаружены герминальные онкомутации, требующие проведения генетического скрининга на предмет носительства мутации среди кровных родственников пробанда, что, в свою очередь, позволяет улучшить раннюю диагностику опухоли и выполнять, к примеру, превентивное хирургическое лечение, если таковое осуществимо. Мультидисциплинарная парадигма МОК позволяет различным специалистам улучшить коммуникацию в интересах не только обсуждаемого, но и многих других пациентов. Конструктивная дискуссия диагностов (радиологов, рентгенологов, патоморфологов, специалистов лабораторной диагностики) с лечебниками (хирургами, радиотерапевтами, химиотерапевтами, специалистами радионуклидной терапии) по поводу предметного клинического случая повышает потенциал современной биомедицины в конкретном медицинском учреждении, учит командной работе и создает благоприятные условия для развития научно-практического сотрудничества. Более того, накопление информации и обмен опытом стандартизируют работу МОК, расширяют ее компетенции и производительность, увеличивают доказательный опыт применения on-label и off-label таргетных препаратов, диагностических технологий, в том числе целенаправленного геномного и постгеномного секвенирования. В дополнение к опухолевым клинико-морфологическим факторам выяснение молекулярного профиля опухоли может предоставлять возможности использования более информативных диагностических методов и коррекции лечебной стратегии на более эффективную. Накопление «молекулярно-клинического» опыта и пополнение доказательной базы требуют совершенствования технологий и принципов, непрерывно обогащаемых собственным и зарубежным опытом. В качестве примеров клинической значимости молекулярного профиля опухоли и роли молекулярной визуализации (ПЭТ/КТ, ОФЭКТ/КТ) в планировании стратегии ведения пациента можно привести опухоли эндокринной системы, например, болезнь Кушинга и рак щитовидной железы. При болезни Кушинга (опухоль гипофиза, продуцирующая АКТГ) агрессивность клинического течения значительно варьирует. По данным полноэкзомного секвенирования частота мутаций в клетках опухоли варьирует от 20 до 60%. При более углубленных молекулярных исследованиях обнаружены дополнительные мутации в гене глюкокортикоидного рецептора NR3C1, а также онкогена BRAF, деубиквитиназа-кодирующем гене USP48, гене TP53, которые встречались гораздо реже. Более того, в дальнейшем было выявлено, что кортикотрофные опухоли с диким и мутантным типами гена USP8 имели совершенно различные транскриптомные профили и как результат обладали разной гормональной активностью и клинической агрессивностью. При недифференцированной (анапластической) карциноме щитовидной железы, крайне агрессивной ЗНО, наличие мутации BRAF в клетках опухоли позволяет успешно применять комбинацию селективных BRAF и MEK-киназных ингибиторов для системной терапии болезни. Визуализация очагов медуллярного рака щитовидной железы (нейроэндокринной опухоли) с помощью соматостатин-рецепторной сцинтиграфии (ОФЭКТ/КТ с тектротидом или ПЭТ/КТ с 68Ga-DOTA-TATE/NOC) позволяет не только улучшить стадирование опухоли, но также является основанием для коррекции терапевтической стратегии добавлением в алгоритм лечения аналогов соматостатина и пептид-рецепторной радионуклидной терапии (177Lu-DOTA-TATE, LutatheraTM). В идеале терапевтический выбор должен опираться на наличие таргетированных драйверных мутаций, анализируемых целенаправленно, или агностических1 (выявленных при поисковых исследованиях) молекулярных биомаркеров, что сегодня составляет авангард прецизионной персонализированной онкологии, эндокринологии и других областей современной биомедицины. Практически значимые агностические молекулярные мишени стали предиктивными биомаркерами для соответствующих таргетных препаратов, эффективных без относительности к локализации и гистологии карциномы. Разработаны дизайны «корзинных» (Basket) и «зонтичных» (Umbrella) клинических исследований для ускорения разработки новых таргетных противоопухолевых препаратов. В связи с жесткими критериями включения в такие исследования набор пациентов был низким, что отчасти послужило поводом для широкого внедрения технологий прецизионной молекулярной онкологии в клиническую практику по всему миру. В результате резко возросло количество онкологических пациентов, являющихся потенциальными кандидатами на изучение молекулярного профиля опухоли на фоне растущего числа потенциальных терапевтических молекулярных мишеней. Сложность интерпретации молекулярных данных также увеличивается. Адекватно транслируемая и интерпретируемая информация о комплексе с генетическими и транскриптомными особенностями опухоли предоставляет клиницистам новые возможности моделирования индивидуализированной лечебно-диагностической стратегии. Это, разумеется, коренным образом меняет привычный функционал и потенциал возможностей традиционных онкологических консилиумов, повышая их персонализированную молекулярно-генетическую осведомленность. МОК может оказаться полезен в выборе наиболее подходящего биологического образца для молекулярного анализа, например, гистологического блока или жидкостной биопсии крови, технологий молекулярно-генетических анализов и преимуществ конкретных тестов. Жидкостная биопсия крови с целенаправленной оценкой профиля геномных нарушений позволяет обеспечить мониторинг уже известных препарат-таргетируемых драйверных мутаций. В клинической практике сегодня применяется жидкостная биопсия крови для обнаружения драйверных мутаций в циркулирующей опухолевой ДНК у пациентов с раком легких. Данная технология активно исследуется и при других солидных ЗНО. МОК особенно важен у пациентов с распространенными ЗНО, а именно: Опубликованный опыт реальной клинической практики применения рекомендованных МОК режимов лечения на основе молекулярного генно-транскриптомного профиля опухоли свидетельствует о лучших результатах безрецидивной и общей выживаемости пациентов в сравнении с лечением, назначаемым врачом без учета молекулярного профиля опухоли. Необходимы дальнейшее накопление опыта и проведение рандомизированных контролируемых клинических испытаний для более основательных и доказательных выводов. Однако ни у кого из специалистов нет сомнений, что МОК является мощнейшим источником развития прецизионной персонализированной медицины. Необходимо продолжить работу по стандартизации исовершенствованию принципов работы МОК, стимулировать создание и пополнение мультидисциплинарных банков данных пациентов, многоцентровые исследования, обмен и тиражирование успешного опыта. Интеграция научной и клинической практики ускоряет комплексное и прикладное исследование генетики рака, а также приближает создание и исследование новых противоопухолевых препаратов и их действенных комбинаций. Созданием эффективной сетевой системы МОК позволит преодолеть неоднородность критериев выбора пациентов для молекулярных исследований, совершенствовать на более высоком доказательном уровне технологии геномных и транскриптомных анализов, а также оптимизировать выбор терапевтической и диагностической стратегии для каждого пациента. Важно обеспечить транспарентность работы и кросс-валидацию результатов молекулярного тестирования в целях обеспечения надлежащего качества исследований, особенно это касается анализов на основе NGS, современных программ обучения и обмена опытом. При обсуждении на МОК лечебной тактики вполне уместны и поощряются дискуссии в отношении накопленного клинического опыта, биоинформатики, преаналитических и аналитических моментов, включая выбор типа биологического материала (ткань или жидкостная биопсия), молекулярно-генетических анализов (полногеномный или целенаправленный) и т.п.
Выбор биологического образца является важнейшим вопросом для информативности последующих молекулярных анализов. Прежде всего очень важно обсудить на МОК, какой именно биоматериал (цитологический или гистологический) является наиболее информативным и будет направлен на молекулярные исследования. В эру прецизионной медицины количество биомаркеров для диагностических, предиктивных и прогностических целей постоянно растет. Однако в некоторых клинических случаях имеется возможность получения лишь небольшого количества биоматериала (например, при пункции под УЗИ-, МСКТ- или ПЭТ/КТ-контролем). Для молекулярных анализов, как правило, достаточно небольшого количества биоматериала, но необходимо обеспечить полноценную пригодность и сохранность извлеченного биопсийного образца для молекулярных исследований, особенно если объем биообразца небольшой. Рекомендуется изначально озаботиться обеспечением надлежащей пробоподготовки (недопущение контаминации, надлежащая маркировка, хранение в контейнере с забуференным раствором формалина, надлежащие условия транспортировки и хранения, качество консервации, минимизация периода доставки в лабораторию для молекулярных исследований). Все этапы пробоподготовки должны выполняться квалифицированным и опытным персоналом, согласованно и ответственно. Патологоанатомы играют ключевую роль, экспертную и координирующую, так как морфологическая верификация опухоли и той ее части, которая направляется на молекулярное исследование, определяет качество и точность последнего. Специализирующиеся на молекулярных исследованиях патологи могут оказать помощь в решении случайных или систематических проблем, связанных с дефектами пробоподготовки. Поэтому активное участие патологоанатома в составе МОК крайне желательно. Первичная морфологическая оценка биологического образца необходима для: По существу патологоанатом должен оценить качественную и количественную адекватность тканевых образцов, особенно фокусируя внимание на пределах детекции молекулярных методов и нижнем пороге содержания нуклеиновых кислот в образце для анализа во избежание ложнонегативных или ненадежных результатов молекулярных исследований. Эти данные должны быть включены в отчет по молекулярным исследованиям для МОК. Более того, на МОК должны рассматриваться вопросы диагностической приоритетности цитологических и гистологических образов, а именно: Жидкостная биопсия является валидной альтернативой молекулярному анализу, что особенно актуально при невозможности получения образца опухолевой ткани или его низком качестве. Жидкостная биопсия может содержать различные опухоль-содержащие жидкости, включающие циркулирующую опухолевую ДНК (цоДНК), опухолевые клетки и экзомы. Анализ цоДНК в плазме крови является наиболее широко используемой альтернативой опухолевой ткани для изучения генотипа солидных опухолей. После многочисленных попыток определения мутаций EGFR при немелкоклеточном раке легкого чувствительность и специфичность определения цоДНК методом NGS были значительно повышены. В связи с неинвазивностью и хорошей воспроизводимостью (стандартная процедура взятия образца венозной крови) жидкостная биопсия имеет важное значение при динамическом мониторинге опухоли, а также для преодоления проблем опухолевой гетерогенности и определения минимальной остаточной болезни (MRD, Minimal Residual Disease). C другой стороны, жидкостная биопсия крови не обладает еще достаточной чувствительностью и имеет высокий риск ложноотрицательных результатов. Метод жидкостной биопсии крови еще больше, чем биопсия ткани опухоли, зависит от качества преаналитической подготовки, начиная от процедуры забора образца крови, центрифугирования, экстракции ДНК и ее хранения. Жидкостная биопсия крови является чрезвычайно перспективным и бурно развивающимся направлением молекулярной онкологии, а ее роль будет только расти в ближайшем и отдаленном будущем.
Клиническая эффективность МОК определяется оперативностью и результативностью оптимального комплекса диагностических мероприятий и терапевтических решений на их основе, позволяющих улучшать выживаемость и качество жизни онкологических пациентов. Важными составляющими успешности в работе МОК являются: (а) получение надлежащего качества и количества биопсийного материала; (б) снижение уровня преаналитических ошибок, внедрение и стандартизация технологий тканевого сэмплинга/анализа; (в) информационная технологическая инфраструктура, позволяющая осуществлять оперативную и удобную для работы МОК инфраструктуру на многоцентровом уровне, а также медицинские банки данных (современные клинические регистры пациентов) для доказательного анализа эффективности и безопасности лечения большого числа клинических случаев; (г) непрерывное повышение эффективности и качества медицинской помощи путем совершенствования тераностики на основе «молекулярно-навигационных» технологий. Прецизионная медицина, особенно тестирование опухолевой ткани на наличие драйверных мутаций с помощью NGS с целью улучшения терапии, является огромным преимуществом в лечении рака и рассматривается как оптимальная опция персонализированной медицины.
Источники финансирования. Работа выполнена по инициативе автора без привлечения финансирования. Конфликт интересов. Автор декларирует отсутствие явных и потенциальных конфликтов интересов, связанных с содержанием настоящей статьи. Участие авторов. Автор одобрил финальную версию статьи перед публикацией, выразил согласие нести ответственность за все аспекты работы, подразумевающую надлежащее изучение и решение вопросов, связанных с точностью или добросовестностью любой части работы. 1. Агностический — происходит от древнегреческого ἄγνωστος — «не основанный на уверенном знании». Агностический принцип применяется для поиска и анализа статистическими методами множества молекулярных мишеней. Осуществляется в целях обнаружения новых биомаркеров после изучения совокупности всех доступных генетических вариантов нарушений (мутаций) с интересующими признаками.
|
Analysis of status quo and influencing factors of oral health literacy among the elderly in Hangzhou | cb09387b-111f-4f78-8e3e-9f0ede38e7fd | 11827154 | Health Literacy[mh] | Oral health is an important sign of a person’s physical and mental health. Additionally, oral diseases are closely related to systemic diseases such as stroke, heart disease, diabetes, and digestive system disorders . Previous research findings have revealed that individuals with poor oral health often exhibit inadequate oral hygiene habits and health beliefs . Oral health literacy (OHL) is the ability to obtain and process fundamental oral health information to make informed decisions regarding oral health. The level of OHL is directly correlated with overall health, quality of life, and happiness . In addition, social support and self-efficacy have also been widely discussed in health behavior and health literacy. Social support is defined as the emotional and practical help an individual receives when facing health challenges, while self-efficacy is the individual’s belief in his or her ability to cope with these challenges .Existing studies have shown that both of them can affect an individual’s health behavior and health management ability. Social support and self-efficacy are considered to be two important factors affecting OHL. There is an interaction between them. Social support can improve self-efficacy, and the improvement of self-efficacy can further promote the level of OHL . Therefore, it is speculated that social support and self-efficacy may affect individuals’ health outcomes by improving their functional health literacy. Despite the importance of this metric, due to multiple influencing factors, the oral health of the elderly tends to be suboptimal . Previous studies have shown that OHL contributes to poor oral health in these aging populations , indicating that OHL must be improved. However, current literature lacks reports on the status of OHL and influencing factors among the elderly. Existing OHL studies predominantly target students and residents , using self-developed questionnaires or impact assessment scales rather than a comprehensive OHL scale, leading to a lack of authority and specificity. This study utilized the Simplified Version ofOral Health Literacy Scale (HeLD-14) translated by Yan et al. Although different dimensions of health literacy have been explored, systematic research on how social support and self-efficacy affect functional health literacy is still insufficient, so this study will explore the effects of social support and self-efficacy on functional health literacy.To investigate the OHL status and influencing factors among the elderly in Hangzhou, providing a scientific basis for improving OHL and enhancing oral health in the elderly population. Subjects This cross-sectional study employed convenience sampling to select 230 elderly individuals residing in Hangzhou, Zhejiang Province, from August 2022 to January 2023. Only participants who were permanent residents of Hangzhou, Zhejiang Province, ≥ 60 years old, and were willing to participate in the study were included. Any participants whose communication was problematic or who had a history of mental health disorders were excluded from this study. All protocols have been approved by the Affiliated Hospital of Stomatology of Zhejiang University School of Medicine (Zhejiang University Stomatology Review No. 071 of 2023), of which Article 7 shows that the informed consent form is exempted, and the purpose of the study will be orally informed when filling out the questionnaire, which will be orally approved. Survey tools The researchers designed a General Information Questionnaire based on retrieval, covering age, gender, marital status, educational level, health-care payment methods, monthly household income, number of co-residents, place of residence, and history of oral health education (9 items). The Simplified Version of the Oral Health Literacy Scale (HeLD-14) was employed in this study. HeLD-14 was developed by Jones et al. based on the complete version (HeLD-29) and translated by Yan et al. This scale comprised seven dimensions with a total of 14 items. The items were scored on a Likert 5-point scale, with 1 point representing “no difficulty” and 5 points representing “unable to do”. During the analysis, 5 points were converted into 0 points, 4 points into 1 point, 3 points into 2 points, 2 points into 3 points, and 1 point into 4 points before calculating the total score. The total score ranges from 0 to 56 points; the higher the score, the higher the OHL level. The original scale demonstrated a Cronbach’s α coefficient of 0.87 , showing good reliability and validity in large sample studies in Australia . In the current study, the overall Cronbach’s α coefficient was 0.87. Next, a Social Support Rating Scale (SSRS), developed by Xiao , was utilized to assess social support. This scale includes three dimensions—objective support (3 items), subjective support (4 items), and utilization of support (3 items). The total score ranges from 12 to 66, with higher scores indicating more comprehensive support. This study’s overall Cronbach’s α coefficient for this scale was 0.83. The General Self-Efficacy Scale (GSES), developed by Schwarzer et al., was used to assess the overall level of individual self-efficacy. This scale consisted of 10 items rated from 1 (completely incorrect) to 4 (completely correct). The maximum score on this scale is 40, with higher scores indicating higher general self-efficacy levels . The Cronbach’s α coefficient for this scale was 0.82 in the original study and 0.89 in this study. Assignment methods of independent variables. Age:<70 years = 1 70 ~ 80 years = 2 > 80 years = 3. Educational level: Junior high school and below = 1 High school/technical secondary school = 2 University and above = 3. Monthly household income:<6000 yuan = 1 6000–8000 yuan = 2 > 8000 yuan = 3. Place of residence: Urban area = 1 Rural area = 2. History of receiving oral health education: Yes = 1 No = 2. Data collection methods Trained nurses with over five years of experience in dental wards were selected as surveyors. After obtaining consent, questionnaires were distributed, and surveyors provided a unified explanation of the study’s purpose. The participants were then individually interviewed to complete the questionnaires. The surveyors checked for completeness, and any missing information was immediately addressed. A total of 230 questionnaires were distributed, with 222 valid responses, resulting in a questionnaire effective recovery rate of 96.52%. Statistics All participant data were entered into Excel and cross-verified by two individuals. Statistical analysis was conducted using SPSS 25.0 software. Descriptive statistics, such as frequencies and percentages , were used to compare qualitative data. For normally distributed quantitative data, mean values ± standard deviations (‾x ± s) were compared using the independent sample t-test and analysis of variance (ANOVA) for group comparisons. For non-normally distributed quantitative data, median values (interquartile range) [M(QR)] were compared using the Mann-Whitney U test or the Kruskal-Wallis test for group comparisons. In this study, the basic data of patients did not meet the normal distribution. Kendall’s tau-b correlation analysis was used to assess the relationship between variables. Multiple linear regression analysis was employed to identify factors influencing OHL in the elderly, with significance set at P < 0.05. This cross-sectional study employed convenience sampling to select 230 elderly individuals residing in Hangzhou, Zhejiang Province, from August 2022 to January 2023. Only participants who were permanent residents of Hangzhou, Zhejiang Province, ≥ 60 years old, and were willing to participate in the study were included. Any participants whose communication was problematic or who had a history of mental health disorders were excluded from this study. All protocols have been approved by the Affiliated Hospital of Stomatology of Zhejiang University School of Medicine (Zhejiang University Stomatology Review No. 071 of 2023), of which Article 7 shows that the informed consent form is exempted, and the purpose of the study will be orally informed when filling out the questionnaire, which will be orally approved. The researchers designed a General Information Questionnaire based on retrieval, covering age, gender, marital status, educational level, health-care payment methods, monthly household income, number of co-residents, place of residence, and history of oral health education (9 items). The Simplified Version of the Oral Health Literacy Scale (HeLD-14) was employed in this study. HeLD-14 was developed by Jones et al. based on the complete version (HeLD-29) and translated by Yan et al. This scale comprised seven dimensions with a total of 14 items. The items were scored on a Likert 5-point scale, with 1 point representing “no difficulty” and 5 points representing “unable to do”. During the analysis, 5 points were converted into 0 points, 4 points into 1 point, 3 points into 2 points, 2 points into 3 points, and 1 point into 4 points before calculating the total score. The total score ranges from 0 to 56 points; the higher the score, the higher the OHL level. The original scale demonstrated a Cronbach’s α coefficient of 0.87 , showing good reliability and validity in large sample studies in Australia . In the current study, the overall Cronbach’s α coefficient was 0.87. Next, a Social Support Rating Scale (SSRS), developed by Xiao , was utilized to assess social support. This scale includes three dimensions—objective support (3 items), subjective support (4 items), and utilization of support (3 items). The total score ranges from 12 to 66, with higher scores indicating more comprehensive support. This study’s overall Cronbach’s α coefficient for this scale was 0.83. The General Self-Efficacy Scale (GSES), developed by Schwarzer et al., was used to assess the overall level of individual self-efficacy. This scale consisted of 10 items rated from 1 (completely incorrect) to 4 (completely correct). The maximum score on this scale is 40, with higher scores indicating higher general self-efficacy levels . The Cronbach’s α coefficient for this scale was 0.82 in the original study and 0.89 in this study. Assignment methods of independent variables. Age:<70 years = 1 70 ~ 80 years = 2 > 80 years = 3. Educational level: Junior high school and below = 1 High school/technical secondary school = 2 University and above = 3. Monthly household income:<6000 yuan = 1 6000–8000 yuan = 2 > 8000 yuan = 3. Place of residence: Urban area = 1 Rural area = 2. History of receiving oral health education: Yes = 1 No = 2. Trained nurses with over five years of experience in dental wards were selected as surveyors. After obtaining consent, questionnaires were distributed, and surveyors provided a unified explanation of the study’s purpose. The participants were then individually interviewed to complete the questionnaires. The surveyors checked for completeness, and any missing information was immediately addressed. A total of 230 questionnaires were distributed, with 222 valid responses, resulting in a questionnaire effective recovery rate of 96.52%. All participant data were entered into Excel and cross-verified by two individuals. Statistical analysis was conducted using SPSS 25.0 software. Descriptive statistics, such as frequencies and percentages , were used to compare qualitative data. For normally distributed quantitative data, mean values ± standard deviations (‾x ± s) were compared using the independent sample t-test and analysis of variance (ANOVA) for group comparisons. For non-normally distributed quantitative data, median values (interquartile range) [M(QR)] were compared using the Mann-Whitney U test or the Kruskal-Wallis test for group comparisons. In this study, the basic data of patients did not meet the normal distribution. Kendall’s tau-b correlation analysis was used to assess the relationship between variables. Multiple linear regression analysis was employed to identify factors influencing OHL in the elderly, with significance set at P < 0.05. Scale and dimension scores Table shows that OHL levels of the older adults were moderate(M = 37), the lowest score was observed to be in the dimension of acceptance capacity(M = 5), social support(M = 22) was at a low level and the objective support dimension has the lowest average score (M = 6), the level of self-efficacy(M = 26) was moderate. Preliminary analysis Table illustrates the socio-demographic characteristics and distribution of OHL among the elderly. The average age of the participants was 72.49 (standard deviation = 10.06) years, with a predominance of females (53.6%). The majority of the participants were married (98.2%), and a significant proportion had an educational background of junior high school or below (78.4%). Most had medical insurance or coverage for medical expenses (72.1%), and more than half (66.2%) had a monthly household income below 6000 yuan. Many of the participants lived with more than two cohabitants (51.4%) and the majority (84.7%) resided in urban areas. A significant proportion of the elderly (68.5%) had not received any oral health education. The results indicated significant differences in OHL in terms of age ( χ 2 value = 9.195, P = 0.010), educational level ( χ 2 value = 14.751, P = 0.001), monthly household income ( χ 2 value = 11.814, P = 0.003), place of residence ( χ 2 value = -2.052, P = 0.040), and history of receiving oral health education ( χ 2 value = -3.556, P < 0.001). Correlation analysis between variables The results of the correlation analysis between the variables is presented in Table . The total OHL score in the elderly was found to be significantly correlated with the total social support score ( r = 0.222), the subjective support dimension ( r = 0.217), and the support utilization dimension ( r = 0.114). The dimension of acceptance ability in OHL showed significant correlations with the total social support score ( r = 0.215), objective support dimension ( r = 0.154), and subjective support dimension ( r = 0.182). The understanding dimension in OHL was significantly correlated with the total social support score ( r = 0.255), objective support dimension ( r = 0.204), and subjective support dimension ( r = 0.230). The support dimension in OHL was found to be significantly associated with both the support utilization dimension ( r = 0.124) and the total self-efficacy score ( r =-0.234). The economic burden dimension in OHL was significantly correlated with the total social support score ( r = 0.126) and the subjective support dimension ( r = 0.176), while the medcal care dimension was significantly associated with both the total social support score ( r = 0.137) and the support utilization dimension ( r = 0.135). The communication dimension of OHL was significantly correlated with the total social support score ( r = 0.203), subjective support dimension ( r = 0.225), and support utilization dimension ( r = 0.139). The application dimension was significantly correlated with the total social support score ( r = 0.188), subjective support dimension ( r = 0.228), and support utilization dimension ( r = 0.168). Determinants of OHL in the elderly Next, the significant variables (age, educational level, monthly household income, place of residence, history of receiving oral health education, social support) in the univariate and correlation analyses were taken as independent variables, and the OHL total score of the elderly was taken as the dependent variable for multiple linear regression analysis. This analysis resulted in a Durbin-Watson score of 1.689, indicating independence among the observed values. Collinearity diagnostics revealed a tolerance range of 0.758 to 0.936 and a variance inflation factor between 1.069 and 1.320, suggesting the absence of multicollinearity among the independent variables. This study met the acccpeted multiple linear regression analysis requirements for robust analysis: F = 12.910, P < 0.001. The regression analysis showed that monthly household income ( t = 2.342, p = 0.020), place of residence ( t =-3.572, p <0.001), oral health education ( t =-3.192, p = 0.002), and social support ( t = 4.859, p <0.001) were the main influencing factors of OHL in the elderly, accounting for 24.4% of the variation (Table ). Table shows that OHL levels of the older adults were moderate(M = 37), the lowest score was observed to be in the dimension of acceptance capacity(M = 5), social support(M = 22) was at a low level and the objective support dimension has the lowest average score (M = 6), the level of self-efficacy(M = 26) was moderate. Table illustrates the socio-demographic characteristics and distribution of OHL among the elderly. The average age of the participants was 72.49 (standard deviation = 10.06) years, with a predominance of females (53.6%). The majority of the participants were married (98.2%), and a significant proportion had an educational background of junior high school or below (78.4%). Most had medical insurance or coverage for medical expenses (72.1%), and more than half (66.2%) had a monthly household income below 6000 yuan. Many of the participants lived with more than two cohabitants (51.4%) and the majority (84.7%) resided in urban areas. A significant proportion of the elderly (68.5%) had not received any oral health education. The results indicated significant differences in OHL in terms of age ( χ 2 value = 9.195, P = 0.010), educational level ( χ 2 value = 14.751, P = 0.001), monthly household income ( χ 2 value = 11.814, P = 0.003), place of residence ( χ 2 value = -2.052, P = 0.040), and history of receiving oral health education ( χ 2 value = -3.556, P < 0.001). The results of the correlation analysis between the variables is presented in Table . The total OHL score in the elderly was found to be significantly correlated with the total social support score ( r = 0.222), the subjective support dimension ( r = 0.217), and the support utilization dimension ( r = 0.114). The dimension of acceptance ability in OHL showed significant correlations with the total social support score ( r = 0.215), objective support dimension ( r = 0.154), and subjective support dimension ( r = 0.182). The understanding dimension in OHL was significantly correlated with the total social support score ( r = 0.255), objective support dimension ( r = 0.204), and subjective support dimension ( r = 0.230). The support dimension in OHL was found to be significantly associated with both the support utilization dimension ( r = 0.124) and the total self-efficacy score ( r =-0.234). The economic burden dimension in OHL was significantly correlated with the total social support score ( r = 0.126) and the subjective support dimension ( r = 0.176), while the medcal care dimension was significantly associated with both the total social support score ( r = 0.137) and the support utilization dimension ( r = 0.135). The communication dimension of OHL was significantly correlated with the total social support score ( r = 0.203), subjective support dimension ( r = 0.225), and support utilization dimension ( r = 0.139). The application dimension was significantly correlated with the total social support score ( r = 0.188), subjective support dimension ( r = 0.228), and support utilization dimension ( r = 0.168). Next, the significant variables (age, educational level, monthly household income, place of residence, history of receiving oral health education, social support) in the univariate and correlation analyses were taken as independent variables, and the OHL total score of the elderly was taken as the dependent variable for multiple linear regression analysis. This analysis resulted in a Durbin-Watson score of 1.689, indicating independence among the observed values. Collinearity diagnostics revealed a tolerance range of 0.758 to 0.936 and a variance inflation factor between 1.069 and 1.320, suggesting the absence of multicollinearity among the independent variables. This study met the acccpeted multiple linear regression analysis requirements for robust analysis: F = 12.910, P < 0.001. The regression analysis showed that monthly household income ( t = 2.342, p = 0.020), place of residence ( t =-3.572, p <0.001), oral health education ( t =-3.192, p = 0.002), and social support ( t = 4.859, p <0.001) were the main influencing factors of OHL in the elderly, accounting for 24.4% of the variation (Table ). OHL levels can be used as a metric of general health. This study shows that OHL among the elderly in Hangzhou is at a moderate level, lower than that of older adults with dental defects . This observation could be attributed to individuals with dental defects seeking medical attention for oral health issues, leading to increased opportunities for oral health education. The acceptance capability dimension for Oral Health Literacy (OHL) comprised two key items, namely, Item 1: “Do you pay attention to your teeth or oral health?” and Item 2: “Do you take time to engage in activities that promote the well-being of your teeth or oral health?” Notably, among the elderly population in Hangzhou, these two dimensions exhibited the lowest scores. This result suggests a lack of attention to dental health among the elderly. These low-scoring parameters could be attributed to this demographic being born before the 1960s when dental medicine in China was in its early stages. Nationwide oral health education activities in China began in September 1989 with the first National Tooth Day. The results of this study indicate that elderly individuals with higher monthly household incomes have higher OHL levels. This observation aligns with findings from studies by Su et al. , Zhang et al. , and Vanwormer et al. The reason may be that economically constrained families experience more material and psychological pressure, dedicating more time and energy to increasing income and focusing less on their health. Higher-income families, on the other hand, may have more access to dental services and opportunities to communicate with dentists, potentially contributing to improved OHL. Next, we show that elderly individuals in urban areas have higher OHL levels than those in rural areas. This disparity may exist despite significant improvements in oral hygiene and notable progress in preventing and managing oral diseases. Inequalities persist between urban and rural areas in the distribution, accessibility, utilization, treatment outcomes, health insurance coverage, oral health-related quality of life, and prevalence of oral diseases. People in rural areas generally have lower economic incomes, less awareness of health conditions, more cavities, fewer teeth, no health insurance, and spend less on dental treatment. The rural-urban gap is often associated with lower levels of education, which, in turn, are related to lower levels of health knowledge dissemination and inadequate utilization of health management services. These factors impact oral health management and service provision, highlighting the need to address oral health inequalities and improve urban-rural disparities. Therefore, we recommend population-specific oral health promotion programs, measures to increase access to oral health services in rural areas, and integrating oral health into existing primary health-care services. Our data indicate that elderly individuals who have received oral health education have higher OHL levels, consistent with the findings of Wang et al. . Research suggests that receiving oral health education can increase knowledge of oral hygiene and preventive care, alter attitudes toward oral health, promote the adoption of oral health behaviors, and is one of the widely implemented educational measures in clinical settings. Additionally, this study demonstrated that elderly individuals with higher social support scores have higher OHL levels, similar to the findings of Suwakhon et al. Social support, including obtaining health information from health-care providers, receiving assistance from family members, and gaining information from the media, positively influences the self-management of oral health in the elderly. Previous studies have shown a significant correlation between social support and good behaviors related to dental care among the elderly, such as wearing dentures . Social support can improve individuals’ ability to manage their health by boosting confidence and motivation. Although self-efficacy was considered to be an important factor influencing health literacy, this study did not find a significant relationship between self-efficacy and functional oral health literacy.According to the correlation analysis, there was a slight negative correlation between the total score of self-efficacy and the support dimension of OHL ( r = -0.234), but the correlation with other OHL dimensions was not significant.There may be several reasons for this: First, measures of self-efficacy may not accurately capture its true impact on oral health literacy.Second, other factors, such as social support and environmental factors, may have a greater impact on health literacy among older adults in the specific cities we studied.Specifically, there was a significant correlation between the various dimensions of OHL and social support in older adults.For example, the correlation coefficient between the OHL total score and the social support total score is r = 0.222.This suggests that social support may play a more important role in the health literacy of older adults, even outweighing the impact of self-efficacy.Therefore, future research needs to further explore the interaction between self-efficacy and other factors to better understand their combined impact on oral health literacy. Our data indicate that the OHL levels among the elderly in Hangzhou need improvement. It is essential to focus on elderly individuals with low monthly household incomes, residing in rural areas, lacking oral health education, and having low social support. Planned, targeted, multi-channel, and specialized oral health education interventions should be implemented to enhance oral hygiene awareness and promote OHL in this demographic. Below is the link to the electronic supplementary material. Supplementary Material 1 |
Transitioning your course to online: High-yield modifications for success | d5679d4c-5ef4-4648-980f-90a2083118fb | 9760338 | Pharmacology[mh] | The COVID-19 pandemic is forcing medical educators to transition in-person courses to an online format with limited preparation, time, and resources, creating challenges related to student engagement, content delivery, and assessment. We transitioned our required graduate clinical pharmacology and pathophysiology courses to an online format several years ago. In this article, we leverage our experiences as well as hundreds of student evaluations to highlight high-yield modifications that are key to teaching online courses successfully.
As we transitioned to online teaching, we initially retained familiar aspects of the in-person course, including the syllabus, recordings of lectures made in the lecture hall that lasted ≥ one hour, handouts in a variety of formats, and ad hoc reminders posted to the course site. Communication with students was not a priority as we assumed that they could seek assistance. There were few ways for students to engage with content actively, and assessment tended to favor traditional exams. We identified challenges using substantial student feedback and refined our online course design to align with best practices. Our approach to online teaching is focused on four pillars: organization, course content, communication, and assessment. We use asynchronous delivery to provide course content in modules on a weekly basis over the semester, allowing students to learn on their own schedule. This is an ideal approach to use during a pandemic that challenges students to juggle many competing demands and responsibilities. Asynchronous course delivery requires student motivation. Graduate professional students tend to be older, engaged, and motivated as the courses relate directly to their current or future career; however, motivation may be an issue for some student populations or with specific courses (e.g., undergraduates taking required courses unrelated to their major or interests). Sound organization of the course is fundamental, and a learning management system (LMS) provides the framework. In the absence of regular on-campus meetings, the online course structure holds everything together. Organization and attention to detail are of paramount importance because even a simple glitch such as a broken link can result in a barrage of emails, frustration, and distraction from learning. We divide classes with large enrollment into manageable sections ≤25 students (subdivided into groups of four to six for collaborative group work). A single faculty member serves as coordinator for each course with responsibility for course oversight, organization, and consistency across the multiple sections. The course coordinator is additionally responsible for orienting and mentoring new and adjunct faculty to the assigned course and standardized processes. An instructor's manual can be beneficial for orienting faculty and ensuring a high-quality, consistent student experience. Challenges for faculty, particularly those new to online teaching, include maintaining enhanced, ongoing communication with students (rather than communicating in periodic weekly classroom meetings) and designing a strategically organized course that optimizes content presentation and the learner's experience. An enhanced syllabus that details expectations thoroughly and a comprehensive schedule grid listing all topics, due dates/times, and assessment details can be accessed together at one place on the course site. As our students are located across different time zones, it is essential to specify date, time, and time zone for live meetings and assignment deadlines. Students are provided access to assessments during a pre-specified window of time to provide flexibility and allowance for different time zones. A simple, standardized content layout organized by week or module provides students with reassuring consistency and predictability. For example, each weekly module is organized on a single page within the LMS and includes objectives, embedded lectures, weekly/module assignments, and supplemental resources . Students consistently comment on our course organization and layout, emphasizing its value in their online learning experience. The online environment provides an easy mechanism for rich content delivery in a variety of multi-media forms. To promote student engagement with content, we provide readings, additional resources, recordings of lectures (divided into short clips of ≤20 min), handouts (on white/light background), and ample practice study questions or case studies with answers. Students resoundingly appreciate being introduced to resources such as clinical practice guidelines or web-based tools that help them apply their knowledge. Students engage in weekly discussion forums that promote higher-level skills as they discuss case study problems or questions about course-related content with one another. In an online course, faculty must move outside the traditional role of content expert to become a learning coach/facilitator. This shift in faculty role requires more frequent but less direct engagement so as to encourage but not dominate or stifle class discussion, thus supporting a learner-centered environment with increased opportunities for learner engagement. Online course faculty communicate frequently in a variety of ways. We provide weekly class updates, optional live meetings using chatrooms or video conferencing, open discussion forums for initial introductions and content questions/answers, prompt feedback on assessments, and collection of mid-course student evaluative feedback. At the beginning of each semester, we record a video introduction that provides a course overview and highlights expectations for netiquette in course discussions and emails. To keep communications manageable for faculty, we maximize group communications and reserve individual communication for personal issues. In our pharmacology course, the “Pharm Chat” room is a popular communication tool that serves multiple functions and is easily accessible through one click on the course menu bar. A chat room tool is available as an option within our LMS; it allows real-time, text-only communication between course site members, and we use it to host a weekly live chat for interested students. Additionally, the chat room tool has an archive capability so that students can refer to previous chat discussions or post content/course questions throughout the week. Interactions that faculty might view as inconsequential (e.g., simple weekly updates, check-in announcements, brief comments accompanying assignment grades) can be valued highly by students and can strengthen faculty-student connections. Faculty accessibility and communication are essential to success in an online environment in which connection and community are emphasized. Online assessment can take many forms, including tests, quizzes, case study assignments, discussion posts, and team projects. We have found that the use of multiple forms of assessment promotes student engagement with the course material. For example, we utilize unfolding case studies that require students to make connections between course modules. Although our courses are asynchronous, we assign group projects (four to six students per group) to encourage students to become engaged and feel connected with their classmates. In our pathophysiology course, students prepare an end-of-semester disease exemplar presentation that requires them to apply key pathophysiological concepts to a common disease. Each student group completes a narrated slide presentation that is posted to the discussion forum so that other groups can provide a scholarly response. Although we allow each group to choose their preferred means of communication, we provide a dedicated discussion forum in the LMS so that communications and documents can be stored in a single place. While we use traditional exams in our pharmacology course, we also use multiple alternative assessment approaches such as an open-resource quiz that requires students to use current immunization and antimicrobial prescribing guidelines to prescribe appropriate therapy in multiple case scenarios. Students frequently comment that this assessment is a valuable experience that helps them learn how to use relevant resources. In our advanced pathophysiology course, students are permitted to use course resources during their exams. To promote integrity during exams and ensure that students are well-prepared and understand the concepts, we (1) use higher level questions requiring an application of knowledge, (2) avoid the use of fact-based questions, (3) set aggressive time limits, and (4) randomize the question order.
In summary, our experiences have allowed us to identify specific key essentials for successful online teaching: (1) Meticulous course organization and structure are fundamental to learning; (2) The online environment can be leveraged to deliver content or assessment in multiple formats; (3) Frequent communication is essential to increase engagement, promote connections, and foster motivation; and (4) Creation of varied and frequent opportunities for learner engagement are key. Prioritizing these high-yield modifications can facilitate online success.
None.
|
Intra‐ and Postoperative Complications in 4565 | 0b58b120-437f-4dd5-ae07-f7443504ed0c | 11794054 | Surgical Procedures, Operative[mh] | Introduction Hysterectomies are globally one of the most frequently performed gynaecologic procedures. In the United States, one in three women will have their uterus removed before the age of 60, giving a yearly rate of 400 000 hysterectomies . vNOTES (vaginal natural transluminal endoscopic surgery) hysterectomy is a combination of a vaginal hysterectomy and endoscopy via the vagina. The method gives the benefits of a vaginal approach to the abdomen with no scars and faster recovery together with the benefits of endoscopic visual overview . The HALON trial was a non‐inferiority single‐centre RCT comparing vNOTES hysterectomy and laparoscopic hysterectomy showing no conversions in any arm. Surgical time, pain and hospital stay were lower in the vNOTES group versus the laparoscopic group. A prospective study with the first 1000 vNOTES procedures of which 730 were hysterectomies registered in the International NOTES Society (iNOTESs) registry showed a conversion rate of 0.4% and an overall complication rate of 5.2% after vNOTES hysterectomy (intraoperative 1.4%, postoperative 3.8%) . The study included surgical data from a single surgeon. vNOTES surgery is on the rise within gynaecology but is still a relatively new technique. The aim of the study is to describe the intra‐ and postoperative hysterectomy complications and conversion rates registered in the iNOTESs registry, mirroring clinical reality with both learning curve data and data from experienced surgeons.
Methods Data on 4565 hysterectomies performed 2015 to January 2024 extracted from the iNOTESs registry were included. The international NOTES society (iNOTESs) was established in 2015 and holds the international iNOTESs intra‐ and postoperative case registry and complication database. The iNOTESs is founded by research groups, each specialising in surgeries performed via a specific natural orifice. The registry is an initiative where the founders aim to collect data for all NOTES surgeries. Surgical data are registered prospectively and unidentified by the operating vNOTES surgeons. The most common procedure performed is vNOTES hysterectomy, followed by vNOTES adnexal surgery. Data regarding, but not limited to; type of procedure, patient demographics, surgical time, conversion rate, intraoperative and postoperative complications within 6 weeks were extracted, also classified by Clavien Dindo . In the iNOTESs database, each patient is registered with an random unique file number r, and no information is registered regarding name, personal ID number, address, etc., and it is not possible to identify any individual in the registry. Only the surgeon has access to the unique file number which is linked to the patient, after a two factor security authentication. In case of a complication, the surgeon will identify the patients unique file number, and register postoperative data in the postoperative complication registry.. Only the cases with complications are registered in the postoperative complication database; the cases with normal postoperative outcome are not registered. Consent was obtained from the patients for the use of data for the registry. All vNOTES surgeons are certified and have participated in a mandatory structured, standardised vNOTES course to be allowed to register their vNOTES cases in the database. The target group for the vNOTES course are surgeons experienced in both vaginal and laparoscopic hysterectomy. Thus, the vast majority of the surgeons were passed their learning curves in vaginal and laparoscopic hysterectomy. The course is standardised by an international expert panel and is hosted with the support Applied Medical in most western countries; it includes standardised lectures, live surgery, hands‐on simulation model training and often post course proctoring. The surgeons are invited to send in a video of their 10th vNOTES hysterectomy in order to be certified. Log in codes to the registry are first given to the surgeons after completion of the surgical vNOTES course. The surgeons reporting data to the registry are reminded continuously to register all of their cases and not just the ones that have good or bad outcome. This is stressed when the surgeon qualifies to receive log in codes. Core outcome set : To our knowledge, there is no current core outcome set for surgical outcome after hysterectomy; therefore, no core outcome set has been used. Patient involvement: No patients have been involved in designing or conducting the study. 2.1 Statistics Descriptive data are presented in frequencies ( n ) and percent . All outcomes (dichotomous outcomes with categorised variables) have been analysed with a chi‐squared test. As a sensitivity analysis stratification for surgical experience (categorised in classes; 0–9, 10–49, 50–99, 100–499, 500–999, 1000 + hysterectomies) was performed. Analysis was performed with the statistical software package IBM SPSS version 29.
Statistics Descriptive data are presented in frequencies ( n ) and percent . All outcomes (dichotomous outcomes with categorised variables) have been analysed with a chi‐squared test. As a sensitivity analysis stratification for surgical experience (categorised in classes; 0–9, 10–49, 50–99, 100–499, 500–999, 1000 + hysterectomies) was performed. Analysis was performed with the statistical software package IBM SPSS version 29.
Results In the database, 4565 vNOTES hysterectomies were identified, 4084 were performed as Vaginally Assisted NOTES Hysterectomy (VANH) and 240 as Total vNOTES Hysterectomy (TVNH). Table shows background characteristics for patients with and without any intra‐ or postoperative complication or conversion. There was a significant difference in parity, surgeon experience, increased surgical time and use of irrigation among patients with intra‐ or postoperative complications/conversions compared to patients without any complications. Duration of surgery was increased in cases with intraoperative or postoperative complications, and the longest duration of surgery was found among surgeries in need of conversion. When stratifying for surgical experience, no significance difference in complications/conversions was found between the different parity groups. It is unclear if this is due to less power in smaller strata, or due to patient selection. A difference was seen in patient selection depending on surgical experience. Among surgeons with maximum 10 cases, 81% of the patients had a previous vaginal delivery, 7.5% were nullipara and 11.8% had a previous CS. The corresponding rates among surgeons with expertise over 1000 cases were 65% previous vaginal delivery, 19% nullipara and 16% previous CS. When stratified for surgical experience no difference in complications depending on BMI or previous surgery was found, but a difference regarding duration of surgery and the use of suction between patients with or without complications/conversions were found. Table shows data regarding intraoperative complications. In 43 cases, complications occurred when establishing access to the abdomen. In five of these cases, a conversion to a different surgical technique was made due to the complication. Furthermore, in five additional cases complications with the vaginal access or vNOTES port placement were reported. Twenty‐one cases were reported as injuries to normally localised organs. Seven cases of injuries to abnormally localised organs. Three cases were categorised as access related haemorrhages, all converted to multiport laparoscopy. Two urological and one gastrointestinal injury and four patients that bled intraoperatively were reported as access related complications. All access related urological injuries were to the bladder. The most common specified injury was urinary tract injury, in total 60 cystotomies and 1 ureteric injury, of which 36 were defined to be surgery related and 24 access related, giving a rate of 1.39%. Twenty five (42%) of the patients that had a cystotomy had previously undergone a caesarean section. All injuries to organs were repaired intraoperatively. One case of macrohematuria during surgery occurred but further assessment showed no signs of damage to the bladder or structures nearby. Five complications relating to anaesthesia were documented ( n = 5), three of which due to insufficient pneumoperitoneum. One patient developed a diffuse erythema without systemic symptoms after administering anaesthetics, and the patient did not undergo surgery due to inadequate anaesthesia. Two patients with anaesthetic complications were scheduled as TVNH, where only one completed as such. Of the TVNH, one surgery could not reach pneumoperitoneum and one patient retained too much gas postoperative after adequate pneumoperitoneum which had to be emptied manually using an ordinary needle after the vaginal cuff had been closed. The perioperative conversion rate was 72 (1.6%), of which 10 (0.2%) were converted to laparotomy. The reason for conversion was intraoperative complication in 23 (32%) of cases and 51% of the patients that were converted of a BMI over 30, and 82% of the conversions occurred within the 50 cases of the surgeon's learning curve. Table shows data regarding postoperative complications. The rate of postoperative complications was 2.52% ( n = 115). The most common postoperative complications were haemorrhages ( n = 28), vaginal cuff or vault complications ( n = 26), cystitis ( n = 18, all except three treated with oral and/or intravenous antibiotics), and non‐specific infection of other location treated with antibiotics ( n = 14). No cases of vault dehiscence were reported. The overall infection rate was 0.94%, and 47 (1%) patients needed a re‐intervention under general anaesthesia. No complications on a level 4B or higher were reported. No deaths occurred. Three patients were reported receiving care at an ICU ‐unit due to single organ failure, categorised as Clavien Dindo 4A . These patients spent 3, 4 and 15 days hospitalised, respectively. Comparison of complications between TVNH and VANH was not completed since only one patient in the TVNH‐group was registered having a postoperative complication. This was determined as a grade 3B with revision due to postoperative haemorrhage 2 weeks post‐surgery. Three patients had both intraoperative and postoperative complications. The mean time of surgery was 125 min, all performed as VANH, two of the patients were obese (BMI 30 and 37) who both had two previous surgeries to the abdomen and surgeon experience of 10–50 previous vNOTES hysterectomies. The postoperative complications in these three cases were categorised as a Clavien Dindo level one and two. Table shows data regarding the vNOTES surgeons. The vNOTES hysterectomies were performed by 201 surgeons, of which 9.5% had performed more than 50 vNOTES cases. As shown in Table , the data consist of both learning curve data (30% of hysterectomies) and data from experienced surgeons (70% of hysterectomies). Half of the hysterectomies resulting in a cystotomy were operated by inexperienced surgeons (previous vNOTES experience < 50), 24 patients were operated by surgeon with intermediate experience (previous vNOTES experience < 500) and 6 patient were operated by surgeon having experience > 500 vNOTES hysterectomies. One main surgeon performed 30% of all hysterectomies ( n = 1364), with 21 intraoperative complications and 43 postoperative complications, giving a total rate of 4.7% intra‐ and postoperative complications. The vast majority of complications were registered prior to 2020 (15 intraoperative and 33 postoperative). Until 2020, the aforementioned surgeon had performed a total of 861 hysterectomies with an intra‐ and postoperative total complication rate of 5.6%. In the years 2020–2024, the same complication rate was reduced to 3.2%. The intra‐ and postoperative complication rate among all other surgeons was 4.9% (107 intraoperative and 51 postoperative complications). 3.1 Main Findings We present the largest prospectively collected data set showing a rate of 3.2% intra‐ and 2.5% postoperative complications after 4565 registered vNOTES hysterectomies. The vNOTES hysterectomies were performed by 201 surgeons, of which 9.5% had performed more than 50 vNOTES cases, representing 70% ( n = 3181) of the registered cases in the registry. The remaining approximate 30% ( n = 1319) of the hysterectomies mainly represent learning curve data from 90% of the included surgeons. Half of the cystotomies were performed by inexperienced surgeons, and the rate of complications decreased with increasing experience, despite operating a higher rate of patients with CS or nulliparity. 3.2 Strengths and Limitations The population of this study ( n = 4565) is the largest one yet evaluating vNOTES hysterectomy. Since the method was introduced in recent times, limitations in pre‐existing scientific research do remain. It is of high importance to summarise a larger study population generating an overview. In addition, it is conducted in numerous countries making it a valuable multicentre study with a beneficial variety of operating surgeons ( n = > 201). The intraoperative complications are registered at the same instance as the patient is registered for the first time in the database; therefore, the risk of missing data or incorrect data is low. The postoperative complications are registered after 6 weeks postoperatively, or when the complication occurs, and the surgeon needs to log in again to the database in order to register the complications. There is therefore most likely a small under‐registration of postoperative complications. We assume that the vast majority of major postoperative complications will be filled in. Supporting this, the rate of Clavien Dindo 3 complications in our study is in line with a recent RCT comparing same day or next day laparoscopic hysterectomy . Patients having minor postoperative complications, for example urinary tract infections, might seek medical attention at their general practitioner and therefore not be registered in the postoperative registry. The iNOTESs questionnaire contains inquiries of various aspects relevant for analysing new operational methods. A weakness of the current database is that not all patient and surgical variables associated with complications are registered, such as indication, smoking, diabetes or uterus weight. Therefore, it is not possible to analyse any predictive factors for intra‐ or postoperative complications. The vNOTES surgeons are requested to fill in all of their operations, not just the uncomplicated ones or the ones with complications. Despite the request, another potential bias could be that vNOTES surgeons do not want to fill in their complications, and only fill in the uneventful hysterectomies. The risk of selection bias, however, can go in both directions, with surgeons also just adding patients with complications. 3.3 Interpretation The rates of intra‐ and postoperative complications reported in the vNOTES registry are in analogy with corresponding rates for other hysterectomy techniques . The HALON trial was a single‐centre blinded non‐inferiority RCT comparing vNOTES to LH. No difference was seen in readmission, postoperative infection or intraoperative complications, although fewer postoperative complications in total were found in the vNOTES group, 82% of postoperative complications were at Clavien Dindo level 1 or 2. A retrospective study of 2000 vNOTES operations found an overall complication rate of 4.4% with a conversion rate of 0.4%. Two systematic reviews comparing vNOTES hysterectomy and laparoscopic hysterectomy show lower postoperative complication rates, less blood transfusion and no difference in intraoperative complication rates or conversions. The authors concluded that vNOTES may have advantages over conventional laparoscopic hysterectomy techniques . The most common intraoperative complication was cystotomy (1.39%), and 42% of the patients that had a cystotomy had previously undergone a caesarean section. All cystotomies were repaired peroperatively, and none had postoperative complications. In the study by Neumann , reporting data from all VH performed at a hospital in Denmark showed a cystotomy rate of 2.3%, and other studies have shown a range from 1.6% to 1.9% . The risk of cystotomy in VH has been reported higher than those of laparoscopic hysterectomies, but lower risk of ureteric injury . A systematic review showed an incidence of cystotomy of 0.28% in over 144 000 benign gynaecological laparoscopic hysterectomies . In contrast, a systematic review by Wei showed lower risk of urinary tract injury in VH versus LH. The review included six cohorts representing 52 492 women undergoing VH and showed a weighted pooled mean injury rate of 295 cystotomies and 122 ureteric injuries per 100 000 cases, respectively. The corresponding data from LH included 15 cohorts with 50 114 women with a weighted pooled mean injury rate of 997 cystotomies and 262 ureteric injuries per 100 000 cases. The risk of cystotomy in VH and vNOTES should be of equivalent rate due to similarity in entrance method. However, in the vNOTES procedure, in cases with difficult vaginal entrance (multiple CS, nullipara, adhesions, large uteri or myoma) the entrance can be performed endoscopically via the vagina. The Alexis ring is put in the pouch of Douglas and under the vaginal mucosa anteriorly, but the peritoneum in the vesicouterine pouch is not yet opened. Pneumovagina and pneumoperitoneum is created, and the vesicouterine pouch is opened under direct endoscopic visualisation. The possibility to create an anterior colpotomy under direct visualisation could possibly reduce the risk of cystotomies compared to a standard VH. No clear distinction of ureteral injury can be made, although it could be feasible that ureteral injuries are less common in vNOTES than other surgeries since only one injury occurred in 4565 hysterectomies. The Alexis ring presses the ureter laterally, away from the surgical instruments. Also, when performing a vNOTES hysterectomy the specimen is pushed cranially, anteriorly and medially, away from the pelvic sidewall. Surgical advancement has led to a reduction in abdominal hysterectomy (AH) and towards LH and RALH. Several surgical guidelines recommend a vaginal entrance to the abdomen when feasible, as it is associated with shorter surgical time, lower complications and quickest recovery . Despite this, the incidence of vaginal hysterectomies are declining, in Sweden only 11% of hysterectomies are performed as a VH . The Swedish Federation of Obstetrics and Gynecology acknowledges vNOTES as an alternative to TLH and VH, with vNOTES giving an advantage over VH regarding adnexal surgery, when lateral visualisation is needed . Comparative guidelines by the UK committee have declared vNOTES hysterectomy as a successful procedure but states criteria of extended caution when carrying out vNOTES procedures, as it is a relatively new surgical procedure and is viewed to have similar complication rates and readmission rates as other methods. A large RCT, comparing vNOTES with LH or VH, with the aim to include 1000 patients, has recently started to include patients and will in the future give further evidence regarding surgical outcome after vNOTES hysterectomy . Advantages in VH do exist considering decreased risk of vaginal vault dehiscence compared to LH , although the risk of hematomas is increased when performing surgery vaginally. Due to properties of vNOTES, with the possibility of meticulous endoscopic haemostasis, vNOTES could have decreased rates of vault hematomas compared to VH and decreased rate of dehiscence compared to LH, as the vault is sutured vaginally. Supporting this theory and consistent with previous research, rates of infected vault hematomas are similar or lower in vNOTES (0.24%) compared to vaginal hysterectomy (2.2%) . vNOTES subsequently poses no increased risk of infected vault hematomas. There seems to be no evidence of vNOTES leading to increased risk of infection compared to any other route of surgery, rather a possibility of decreased infectious burden. For reasons stated above, vNOTES can be considered a valid alternative when choosing an operative method in benign hysterectomies.
Main Findings We present the largest prospectively collected data set showing a rate of 3.2% intra‐ and 2.5% postoperative complications after 4565 registered vNOTES hysterectomies. The vNOTES hysterectomies were performed by 201 surgeons, of which 9.5% had performed more than 50 vNOTES cases, representing 70% ( n = 3181) of the registered cases in the registry. The remaining approximate 30% ( n = 1319) of the hysterectomies mainly represent learning curve data from 90% of the included surgeons. Half of the cystotomies were performed by inexperienced surgeons, and the rate of complications decreased with increasing experience, despite operating a higher rate of patients with CS or nulliparity.
Strengths and Limitations The population of this study ( n = 4565) is the largest one yet evaluating vNOTES hysterectomy. Since the method was introduced in recent times, limitations in pre‐existing scientific research do remain. It is of high importance to summarise a larger study population generating an overview. In addition, it is conducted in numerous countries making it a valuable multicentre study with a beneficial variety of operating surgeons ( n = > 201). The intraoperative complications are registered at the same instance as the patient is registered for the first time in the database; therefore, the risk of missing data or incorrect data is low. The postoperative complications are registered after 6 weeks postoperatively, or when the complication occurs, and the surgeon needs to log in again to the database in order to register the complications. There is therefore most likely a small under‐registration of postoperative complications. We assume that the vast majority of major postoperative complications will be filled in. Supporting this, the rate of Clavien Dindo 3 complications in our study is in line with a recent RCT comparing same day or next day laparoscopic hysterectomy . Patients having minor postoperative complications, for example urinary tract infections, might seek medical attention at their general practitioner and therefore not be registered in the postoperative registry. The iNOTESs questionnaire contains inquiries of various aspects relevant for analysing new operational methods. A weakness of the current database is that not all patient and surgical variables associated with complications are registered, such as indication, smoking, diabetes or uterus weight. Therefore, it is not possible to analyse any predictive factors for intra‐ or postoperative complications. The vNOTES surgeons are requested to fill in all of their operations, not just the uncomplicated ones or the ones with complications. Despite the request, another potential bias could be that vNOTES surgeons do not want to fill in their complications, and only fill in the uneventful hysterectomies. The risk of selection bias, however, can go in both directions, with surgeons also just adding patients with complications.
Interpretation The rates of intra‐ and postoperative complications reported in the vNOTES registry are in analogy with corresponding rates for other hysterectomy techniques . The HALON trial was a single‐centre blinded non‐inferiority RCT comparing vNOTES to LH. No difference was seen in readmission, postoperative infection or intraoperative complications, although fewer postoperative complications in total were found in the vNOTES group, 82% of postoperative complications were at Clavien Dindo level 1 or 2. A retrospective study of 2000 vNOTES operations found an overall complication rate of 4.4% with a conversion rate of 0.4%. Two systematic reviews comparing vNOTES hysterectomy and laparoscopic hysterectomy show lower postoperative complication rates, less blood transfusion and no difference in intraoperative complication rates or conversions. The authors concluded that vNOTES may have advantages over conventional laparoscopic hysterectomy techniques . The most common intraoperative complication was cystotomy (1.39%), and 42% of the patients that had a cystotomy had previously undergone a caesarean section. All cystotomies were repaired peroperatively, and none had postoperative complications. In the study by Neumann , reporting data from all VH performed at a hospital in Denmark showed a cystotomy rate of 2.3%, and other studies have shown a range from 1.6% to 1.9% . The risk of cystotomy in VH has been reported higher than those of laparoscopic hysterectomies, but lower risk of ureteric injury . A systematic review showed an incidence of cystotomy of 0.28% in over 144 000 benign gynaecological laparoscopic hysterectomies . In contrast, a systematic review by Wei showed lower risk of urinary tract injury in VH versus LH. The review included six cohorts representing 52 492 women undergoing VH and showed a weighted pooled mean injury rate of 295 cystotomies and 122 ureteric injuries per 100 000 cases, respectively. The corresponding data from LH included 15 cohorts with 50 114 women with a weighted pooled mean injury rate of 997 cystotomies and 262 ureteric injuries per 100 000 cases. The risk of cystotomy in VH and vNOTES should be of equivalent rate due to similarity in entrance method. However, in the vNOTES procedure, in cases with difficult vaginal entrance (multiple CS, nullipara, adhesions, large uteri or myoma) the entrance can be performed endoscopically via the vagina. The Alexis ring is put in the pouch of Douglas and under the vaginal mucosa anteriorly, but the peritoneum in the vesicouterine pouch is not yet opened. Pneumovagina and pneumoperitoneum is created, and the vesicouterine pouch is opened under direct endoscopic visualisation. The possibility to create an anterior colpotomy under direct visualisation could possibly reduce the risk of cystotomies compared to a standard VH. No clear distinction of ureteral injury can be made, although it could be feasible that ureteral injuries are less common in vNOTES than other surgeries since only one injury occurred in 4565 hysterectomies. The Alexis ring presses the ureter laterally, away from the surgical instruments. Also, when performing a vNOTES hysterectomy the specimen is pushed cranially, anteriorly and medially, away from the pelvic sidewall. Surgical advancement has led to a reduction in abdominal hysterectomy (AH) and towards LH and RALH. Several surgical guidelines recommend a vaginal entrance to the abdomen when feasible, as it is associated with shorter surgical time, lower complications and quickest recovery . Despite this, the incidence of vaginal hysterectomies are declining, in Sweden only 11% of hysterectomies are performed as a VH . The Swedish Federation of Obstetrics and Gynecology acknowledges vNOTES as an alternative to TLH and VH, with vNOTES giving an advantage over VH regarding adnexal surgery, when lateral visualisation is needed . Comparative guidelines by the UK committee have declared vNOTES hysterectomy as a successful procedure but states criteria of extended caution when carrying out vNOTES procedures, as it is a relatively new surgical procedure and is viewed to have similar complication rates and readmission rates as other methods. A large RCT, comparing vNOTES with LH or VH, with the aim to include 1000 patients, has recently started to include patients and will in the future give further evidence regarding surgical outcome after vNOTES hysterectomy . Advantages in VH do exist considering decreased risk of vaginal vault dehiscence compared to LH , although the risk of hematomas is increased when performing surgery vaginally. Due to properties of vNOTES, with the possibility of meticulous endoscopic haemostasis, vNOTES could have decreased rates of vault hematomas compared to VH and decreased rate of dehiscence compared to LH, as the vault is sutured vaginally. Supporting this theory and consistent with previous research, rates of infected vault hematomas are similar or lower in vNOTES (0.24%) compared to vaginal hysterectomy (2.2%) . vNOTES subsequently poses no increased risk of infected vault hematomas. There seems to be no evidence of vNOTES leading to increased risk of infection compared to any other route of surgery, rather a possibility of decreased infectious burden. For reasons stated above, vNOTES can be considered a valid alternative when choosing an operative method in benign hysterectomies.
Conclusion This prospective international database study has the largest multicentre study population of vNOTES hysterectomies to date, performed by over 200 surgeons. The data consist of both learning curve data (30%) and data from experienced surgeons (70%) The intra‐ and postoperative complication and infection rates reported are lower or in the same range as other minimally invasive techniques, and the conversion rate to laparotomy was very low (0.2%).
A.S., J.W., L.B.F., J.S., A.M., S.E., M.H., J.V., D.H., A.L. and J.B., contributed to the design, background material research and writing the paper and operated the patients. A.S. and A.L. contributed to the statistical design and calculations. All authors have approved the final version and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.
The Swedish National Ethical Board approved the study, reference number 2023–04433‐01, dated 2023‐09‐23.
Jan Baekelandt, Andrea Stuart, Johanna Wagenius, Alvaro Montealegre, Michael Hartmann and Jona Vercammen declare consultancy for Applied Medical.
|
A Diagnostic Chip for the Colorimetric Detection of | 62b25c2b-f164-4774-aeb8-ba3b7cfae67c | 11118137 | Microbiology[mh] | Legionella pneumophila has recently been pinpointed by the World Health Organization as the highest health burden of all waterborne pathogens in the European Union, with many outbreaks reported around the globe every year. Legionella has been found in a wide range of systems, due to its preference for temperatures between 25 and 55 °C, including hot- and cold-water systems, cooling towers, certain types of air conditioning systems, and generally in various water sources such as showers, swimming pools, and fountains, where Legionella can be aerosolized . Infection occurs upon breathing air containing water droplets contaminated with bacteria and is fatal in approximately 10–15% of cases, while this percentage is further increased in patients admitted to intensive care units . As the concentration of Legionella bacteria can double every 90 min under favorable conditions, a very low initial bacterial concentration could increase thousands of times in one day . To prevent outbreaks and deaths, the European Centre for Disease Prevention and Control recommends applying regular checks of man-made water systems and taking appropriate control measures to prevent Legionnaires’ disease at recreation centers, accommodation sites, hospitals, or other settings, where sizeable populations may be exposed . The most established method for bacteria detection is based on bacteria culturing onto agar plates (ISO 11731 for Legionella ). However, this method is characterized by limited sensitivity, a long incubation period (approximately 12 days for Legionella ), false-negative results, and an inability to be implemented for fast and accurate analysis or for multiple-sample analysis in facilities of public interest . Towards this direction, novel methods have been developed to provide results in less than two days. These methods include immunological methods, flow cytometry, optical and electrochemical biosensors, gas sensors, and microfluidic platforms . Nevertheless, two of the most challenging issues, sample preparation/preconcentration and sensitive detection in a portable format, still remain unresolved. Therefore, compact yet simple, rapid, accurate, cost-efficient, real-time, and user-friendly methods are still needed for the on-site detection of waterborne pathogens. This requires further development and tailoring of revolutionary technologies, and this is the gap that the present work promises to fill. The advent of microfluidics and laboratories on a chip (LoC), enabling the on-chip integration of all steps from sample preparation to detection, brings many advantages, such as reduced time, smaller sample and reagent volumes, low contamination risk, automation, multiple and parallel sample detection, low cost, and compatibility with downstream processes . Additionally, all analytical steps, such as sample pre-treatment, chemical reactions, and on-line quantification, can be integrated into a single, portable microfluidic platform for on-site testing applications . However, the detection of waterborne bacteria by microfluidics-based methods is still challenging due to the inherent limitations related to the use of small sample volumes in microchannels, presenting obstacles in terms of desirable sensitivity, selectivity, and stability for microfluidic sensors. The scientific community has proposed portable micro/nano/biosystem prototypes with distinct chip designs for the ultra-fast analysis of pathogens in food, accommodating bacteria capturing and lysis as well as DNA amplification and detection, to cope with the urgent problem of foodborne outbreaks . Among various DNA amplification techniques, LAMP is considered advantageous due to its fast amplification, simple operation at a single temperature, robustness, tolerance to inhibitors, and high sensitivity and specificity. The specificity of LAMP is higher than that of PCR, thanks to the use of four to six different primers that specify the template strand to be amplified. LAMP amplifies nucleic acid at temperatures between 60 and 66 °C by leveraging the Bst polymerase enzyme with high strand displacement activity , and thus, in contrast to PCR, it is able to perform the test in the absence of any sophisticated equipment. One limitation of LAMP for use at the point of interest is the need for a refrigerator for the storage of the reagents. For this reason, the lyophilization of LAMP reagents and their stability have been studied in order to simplify the analysis for the user. Lyophilization has been successfully reported with the addition of sugars in the reaction mixture for the detection of bacteria and viruses . Lyophilized samples are usually stored in aluminum pouches with silica gel at temperatures varying from −20 °C to 37 °C. In previous research, when stored at 37 °C, lyophilized reagents were found to be stable for a maximum of 14 days ; at room temperature (25 °C), they were stable for almost a month ; and at lower temperatures (4 °C, −20 °C), they remained stable from 30 to even 720 days . In integrated LoC platforms, an on-chip sample preparation module is typically connected to a sensing area or a sensor housed on the same substrate or heterogeneously integrated. There are several detection schemes and sensors, such as electrochemical, capacitive, conductivity, and several optical ones, based on fluorescence, turbidity, and colorimetry. Colorimetric methods are easy to use, and, often, detection can be achieved by the naked eye in the form of a color change. Typically, such methods are based on fluorescence enhancement by nanoparticles in order to be easily observable by the naked eye . Rarely, a simple color change may also be used; however, in this case, a significant optical path is needed (in the millimeter range), which is not easily realized in a microfluidic chip (a few tens to a few hundreds of microns deep). Hence, typically, samples are directed to some type of well or large microreactor, where the depth is enough to allow optical observation . Therefore, sample preparation is rarely achieved in the same microfluidic channel as the detection via a color change. With the exception of our previous work , there are no approaches where sample preparation and DNA amplification are both accommodated on the same microfluidic chip, offering a small footprint, user friendliness, and high specificity. With the significant development of various LoC devices, several LAMP-based diagnostic platforms have been developed for the detection of Legionella and other bacteria and have been commercialized . The detection is usually based on fluorescence or turbidity measurements; the cost per test varies from USD 2 to 80, while the cost of the accompanying reader ranges from USD 2.5K to 25K. Significant features are the detection time that ranges from 20 min to 6 h and the limit of detection (LOD) that ranges from 10 2 to 10 5 CFU/mL . The most significant shortcoming is that all of these commercial platforms accommodate DNA amplification only, while most of the steps for sample preparation are performed externally. The present work provides a method for the rapid diagnosis of Legionella spp., in particular of L. pneumophila , in less than 3 h (instead of the 12 days required in the ISO method ) by developing an optimized LoC for rapid, sensitive, specific, and precise colorimetric end-point detection on a chip. This chip is the first to integrate sample preparation and detection in the same microchannel. In addition, the described diagnostic LoC exhibits certain advantages over standard technologies by achieving the following: a reduction in analysis time by 99%; a simplification of fabrication by using a widespread manufacturing technology (i.e., machining); cost-effectiveness using simple bio-processes and small reagent volumes; and a generic product able to be used in many applications based on bacterial DNA detection. The potential impact of this diagnostic chip is huge as governments attempt to tackle the worldwide spread of infectious diseases. In fact, the same technology with the appropriate choice of antibodies and DNA amplification primers can be used for the detection of any pathogenic bacteria other than Legionella .
2.1. Materials Clear plexiglass sheets, 4 mm thick, were purchased from IRPEN (Spain). Plasma treatment was performed in a parallel-plate reactive-ion etcher (from Nextral, France) operating at 13.56 MHz. PMMA samples were treated, typically for 12 min, in O 2 plasma under highly anisotropic etching conditions (O 2 flow: 100 sccm; plasma power: 400 W; pressure: 10 mTorr). After the plasma treatment, an annealing step was performed in an oven at 100 °C for 2 h (i.e., below the glass transition temperature, T g = 109 °C, of the polymer). The chip was sealed with lamination using pressure-sensitive adhesive tape (3M Advanced Polyolefin Microplate Sealing Tape 9795 from 3Μ Co, Olitecn Lab Equipment, Athens, Greece). Anti- Legionella pneumophila rabbit polyclonal antibody was purchased from VWR International, LLC (Radnor PA, USA). Lyophilization (freeze drying) was performed using the FDL-10N-50-TD system from MRC Lab (Harlow, UK). Certified reference material (CRM) of L. pneumophila (CRM12821M LENTICULE ® discs, Supelco (Bellefonte, PA, USA)) was purchased from LIFESCIENCE (Athens, Greece) and kept at −20 °C until used. Ringer’s solution was purchased from MERCK (Darmstadt, Germany). A Triton mix (Triton X-100), purchased from New England Biolabs (NEB, Ipswich, MA, USA), was used as lysis buffer. The WarmStart Colorimetric LAMP 2X Master Mix (M1800 or M1804S with UDG) LAMP amplification kit and DNAse- and RNAse-free water were purchased from New England Biolabs (NEB, Ipswich, MA, USA). 2.2. Bacteria Cultures and Counting Once removed from the freezer, the LENTICULE ® discs were transferred to ambient temperature within 1 h. The CRMs were then rehydrated in 10 mL of 1/40 Ringer’s solution for approximately 10 min; appropriate serial dilutions were conducted in the same liquid media; then, 0.1 mL of the diluted CRMs was surface spread-plated over glycine vancomycin polymyxin B cycloheximide (GVPC) (VWR, Radnor, PA, USA) plates. Inoculated plates were incubated at 36 °C for 8–10 days and the number of colonies obtained was counted. 2.3. Chip Design and Fabrication The chip comprised two straight microfluidic channels, each connected to an inlet and a waste well (see ). One microchannel was used as a control and the other as a sample microchannel (17 mm long). A superhydrophobic valve connected the microfluidic channel to the waste well to prevent sample loss or waste backflow. In detail, a constriction of the microchannel width down to 500 μm was fabricated by CNC milling and was subsequently coated by drop-casting a commercial hydrophobic thin film (Teflon ® AF 1600 from DuPont) to act as a superhydrophobic valve that facilitated the complete filling of the microchannel with a liquid volume equal to 25 and 12 μL for the sample and negative control line, respectively. To push liquid to the waste well, new liquid or air should be injected into the microchannel with pressure exceeding the valve operation threshold. The chip was fabricated on a polymeric poly(methyl methacrylate) (PMMA) substrate. Alternative polymers that have been tested include Cyclo Olefin (COP), PS, and PEEK. The process for the fabrication of the chips is shown schematically in . The microchannels and wells were fabricated by Computer Numerical Control (CNC) milling (step 1). After CNC, the chip was micro/nanotextured by treatment (through a stencil mask) with O 2 plasma (step 2; for processing details, see ). The exposed chip surface was etched and simultaneously nanotextured to acquire a high surface area . Subsequently, the microchannels and waste wells were sealed (step 3) with a pressure-sensitive adhesive film (in a laminator). Following sealing, the capturing antibodies (at a high concentration) were immobilized on the micro/nanotextured surface (step 4) to achieve bacteria capturing and isolation with an extremely high specificity and efficiency. It was proven convenient to lyophilize the antibody solution added into the microchannels (step 4) to allow the long-term usage of the chips (up to 18 months), facilitate their transport at ambient temperature, and simplify the process for the end-user. 2.4. Lyophilization (Freeze-Drying) Process for L. pneumophila Antibody on Chip To enable the easy storage of the chips, the antibodies were immobilized and the functionalized surface was subjected to freeze drying in the presence of BSA . The process followed is described in detail below: L. pneumophila antibodies (4 mg/mL stock initial solution) were immobilized onto the sealed chips through the injection of 25 μL of a 100 μg/mL solution in carbonate buffer (pH 9.2, 50 mM) and overnight incubation at 4–8 °C in a controlled-humidity chamber (75% humidity). Then, the washing of the chip surfaces with 25 μL of 10 mM Tris-HCl, pH 8.25, and 9 g/L NaCl (washing buffer) and blocking using 25 μL of the 10 g/L BSA solution in 0.1 M NaHCO 3 , pH 8.5, for 1 h at RT followed. After blocking, cooling down to around −50 °C took place by lowering the freeze-dryer shelf temperature for approximately 5 h to ensure uniform supercooling across the batch. The antibodies were incubated in the chip overnight (~18 h). Following freeze drying, the chips were stored inside Mylar foil pouches together with desiccant bags and kept at room temperature (RT) until use (for up to 18 months without losing their functionality). 2.5. Sample Preconcentration and Chip Operation Sample preconcentration is necessary in order to reduce the water sample volume from several hundreds of milliliters to a few hundred microliters. For this reason, a filtration-based sample preconcentration protocol was developed ( a). Two hundred milliliters of artificially contaminated water or another liquid sample was filtered using 0.2 μm pore size filters. The bacterial pellet that formed on the filter was resuspended in 200 μL of buffered solution. Subsequently, 100 μL of the resuspended pellet solution was injected into the functionalized chip to initiate antibody-based bacteria capture ( b). Following bacteria capture, a lysis buffer and a LAMP amplification cocktail were introduced. In detail, the chip operation sequence after bacteria capturing was as follows: PBS buffer injection to wash the unbound bacteria (total volume: 30 μL), LAMP reagent injection (total volume: 24 μL), chemical lysis with triton (included in the LAMP reaction) for 10 min, and chip heating on a plate at 65 °C for up to 60 min. DNA detection can be achieved visually by detecting the color change by the naked eye or by a camera. For increased capturing efficiency, the chip was operated in batch mode as follows: A sample volume equal to the microchannel volume (25 μL) was introduced and was allowed to react with the antibodies for an optimized time of 25 min. Then, the sample was pushed to the waste well and a second batch of equal volume was allowed to enter and react with the antibodies. Following bacteria capture on the microchannel surface from 4 batch samples, in situ lysis and specific DNA amplification were performed, followed by chromogenic naked-eye end-point detection to allow us to reach a decision on the potential presence of living bacteria with high sensitivity. 2.6. Amplification Protocol for on-Chip Detection of L. pneumophila The sequences of the primers used in this work for DNA amplification via LAMP were as follows: F3_CGTTACCCACAGAAGAAGC; B3_ACCCTCTCCCATACTCGA; FIP_AGTAATTCCGATTAACGCTCGCAACCGGCTAACT CCGTGC; BIP_GGCGTAAAGGGTGCGTAGGTGACCAGTATTATCT GACCGTCC. The LAMP assay for Legionella detection is shown below: The LAMP reagent mix, with a total volume of 25 μL, contained the following: 12.5 μL of WarmStart Colorimetric Lamp 2X Master Mix (New England BioLabs, Ipswich, MA, USA), 2.5 μL of 10× primer mix (16 μM (each) of Forward Inner Primer (FIP) and Backward Inner Primer (BIP); 2 μM (each) of F3 and B3 primers), 8 μL of DNAse- and RNAse-free water, and 1 μL of Triton mix (10% Triton X-100 + 44 μL of DNAse-free water + 0.5 μL of 1 M Tris buffer pH 8.4 + 0.5 μL of 5% phenol red solution) for chemical lysis and 1 μL of any crude water sample, containing cells or not. After 5–10 min of waiting for the chemical lysis with Triton X-100 to take place, the chip was placed on a heating plate at 66 °C for up to 60 min. Amplification in the presence of target nucleic acids results in the production of protons that decrease the mixture’s pH, resulting in a color change in the phenol red from pink to yellow that is easily detectable by the naked eye. For validating amplification, the DNA amplicons were analyzed using electrophoresis on 2% agarose gel containing GelRed (Biotium, Lab Supplies, Athens, Greece) and visualized under UV light. 2.7. Lyophilization Process (Freeze Drying) for the Amplification Cocktail for L. pneumophila To facilitate the use of the amplification cocktail outside analytical laboratories, i.e., at the point of need or in resource-limited settings, the typical requirement for the storage of the reaction mixture components at −20 °C should be lifted. Thus, to allow room-temperature storage and create a stabilized form of the reaction mixture, lyophilization of the LAMP reagents for the detection of Legionella spp. was attempted. The addition of cryoprotective substances was investigated, as proposed by Klatser et al. and Agel et al. . In this work, it was found that the addition of the non-reducing sugar D-trehalose together with a small concentration of BSA stabilized the reagents during freeze drying. After lyophilization, the amplification cocktail was stored inside Mylar foil pouches together with desiccant bags. The Mylar foil pouches acted as an effective barrier against air moisture with very low water vapor transmission rates, oxygen, and light, resulting in a more stable dehydrated reagent mixture. The lyophilized reaction mixtures in the presence of D-trehalose/BSA, stored inside Mylar foil pouches, retained their activity for prolonged periods, i.e., more than 3 months under dry conditions at room temperature, without compromising the performance of the amplification cocktail. 2.8. Optical Image Analysis In order to achieve a robust diagnosis for the presence of bacteria in the sample based on the color change in the microchannel, automated detection of the color change was attempted. To this end, a computational methodology was designed and realized in a software application, based on the analysis of camera photos of the chip, targeting an assessment of the color changes of the negative control and (sample) detection channels. The proposed methodology consisted of three steps following the acquisition of a photo of the chip after sample analysis. First, the automatic identification and isolation of the regions of interest of the chip photo, i.e., of the control and detection channels, were performed. Second, the color change in these areas (channels) was quantified through the measurement of a specific color index. Third, using standard samples of known bacterial concentrations, the proposed color index was assessed as a robust means on which the diagnosis result can be based. 1st step: To achieve the isolation of the detection and control channels, a machine learning model architecture, called U-Net , was trained using images from samples analyzed on chip. U-Net is a convolutional neural network architecture developed for image processing and object recognition applications. It consists of an “encoder” and a “decoder” that are combined with a connection bridge. The encoder detects high-level features in the image, while the decoder reconstructs an exact match of the original image. This architecture, shown schematically in , is highly efficient for tasks such as the requirement herein, namely the isolation of objects in an image (i.e., the detection and control channels). It must be emphasized that U-net produces a “segmentation map”; i.e., it produces an image with the same dimensions but with active pixels only in the areas of interest (control and detection channels). Therefore, by knowing the positions of the active pixels in the result, we can locate the corresponding positions in the original image. 2nd step: The color change was analyzed for each isolated channel. The digital quantification of a color involves the use of a model, such as the Hue Saturation Value (HSV) system. A colored image consists of a collection of pixels, with each pixel containing a set of three HSV coordinates. In our software, we used the hue value of the HSV model to quantify the channel color and, thus, based on its change, whether the sample was positive or negative. 3rd step: The assessment of the proposed hue value was carried out using samples of known bacterial concentrations, as will be shown in .
Clear plexiglass sheets, 4 mm thick, were purchased from IRPEN (Spain). Plasma treatment was performed in a parallel-plate reactive-ion etcher (from Nextral, France) operating at 13.56 MHz. PMMA samples were treated, typically for 12 min, in O 2 plasma under highly anisotropic etching conditions (O 2 flow: 100 sccm; plasma power: 400 W; pressure: 10 mTorr). After the plasma treatment, an annealing step was performed in an oven at 100 °C for 2 h (i.e., below the glass transition temperature, T g = 109 °C, of the polymer). The chip was sealed with lamination using pressure-sensitive adhesive tape (3M Advanced Polyolefin Microplate Sealing Tape 9795 from 3Μ Co, Olitecn Lab Equipment, Athens, Greece). Anti- Legionella pneumophila rabbit polyclonal antibody was purchased from VWR International, LLC (Radnor PA, USA). Lyophilization (freeze drying) was performed using the FDL-10N-50-TD system from MRC Lab (Harlow, UK). Certified reference material (CRM) of L. pneumophila (CRM12821M LENTICULE ® discs, Supelco (Bellefonte, PA, USA)) was purchased from LIFESCIENCE (Athens, Greece) and kept at −20 °C until used. Ringer’s solution was purchased from MERCK (Darmstadt, Germany). A Triton mix (Triton X-100), purchased from New England Biolabs (NEB, Ipswich, MA, USA), was used as lysis buffer. The WarmStart Colorimetric LAMP 2X Master Mix (M1800 or M1804S with UDG) LAMP amplification kit and DNAse- and RNAse-free water were purchased from New England Biolabs (NEB, Ipswich, MA, USA).
Once removed from the freezer, the LENTICULE ® discs were transferred to ambient temperature within 1 h. The CRMs were then rehydrated in 10 mL of 1/40 Ringer’s solution for approximately 10 min; appropriate serial dilutions were conducted in the same liquid media; then, 0.1 mL of the diluted CRMs was surface spread-plated over glycine vancomycin polymyxin B cycloheximide (GVPC) (VWR, Radnor, PA, USA) plates. Inoculated plates were incubated at 36 °C for 8–10 days and the number of colonies obtained was counted.
The chip comprised two straight microfluidic channels, each connected to an inlet and a waste well (see ). One microchannel was used as a control and the other as a sample microchannel (17 mm long). A superhydrophobic valve connected the microfluidic channel to the waste well to prevent sample loss or waste backflow. In detail, a constriction of the microchannel width down to 500 μm was fabricated by CNC milling and was subsequently coated by drop-casting a commercial hydrophobic thin film (Teflon ® AF 1600 from DuPont) to act as a superhydrophobic valve that facilitated the complete filling of the microchannel with a liquid volume equal to 25 and 12 μL for the sample and negative control line, respectively. To push liquid to the waste well, new liquid or air should be injected into the microchannel with pressure exceeding the valve operation threshold. The chip was fabricated on a polymeric poly(methyl methacrylate) (PMMA) substrate. Alternative polymers that have been tested include Cyclo Olefin (COP), PS, and PEEK. The process for the fabrication of the chips is shown schematically in . The microchannels and wells were fabricated by Computer Numerical Control (CNC) milling (step 1). After CNC, the chip was micro/nanotextured by treatment (through a stencil mask) with O 2 plasma (step 2; for processing details, see ). The exposed chip surface was etched and simultaneously nanotextured to acquire a high surface area . Subsequently, the microchannels and waste wells were sealed (step 3) with a pressure-sensitive adhesive film (in a laminator). Following sealing, the capturing antibodies (at a high concentration) were immobilized on the micro/nanotextured surface (step 4) to achieve bacteria capturing and isolation with an extremely high specificity and efficiency. It was proven convenient to lyophilize the antibody solution added into the microchannels (step 4) to allow the long-term usage of the chips (up to 18 months), facilitate their transport at ambient temperature, and simplify the process for the end-user.
To enable the easy storage of the chips, the antibodies were immobilized and the functionalized surface was subjected to freeze drying in the presence of BSA . The process followed is described in detail below: L. pneumophila antibodies (4 mg/mL stock initial solution) were immobilized onto the sealed chips through the injection of 25 μL of a 100 μg/mL solution in carbonate buffer (pH 9.2, 50 mM) and overnight incubation at 4–8 °C in a controlled-humidity chamber (75% humidity). Then, the washing of the chip surfaces with 25 μL of 10 mM Tris-HCl, pH 8.25, and 9 g/L NaCl (washing buffer) and blocking using 25 μL of the 10 g/L BSA solution in 0.1 M NaHCO 3 , pH 8.5, for 1 h at RT followed. After blocking, cooling down to around −50 °C took place by lowering the freeze-dryer shelf temperature for approximately 5 h to ensure uniform supercooling across the batch. The antibodies were incubated in the chip overnight (~18 h). Following freeze drying, the chips were stored inside Mylar foil pouches together with desiccant bags and kept at room temperature (RT) until use (for up to 18 months without losing their functionality).
Sample preconcentration is necessary in order to reduce the water sample volume from several hundreds of milliliters to a few hundred microliters. For this reason, a filtration-based sample preconcentration protocol was developed ( a). Two hundred milliliters of artificially contaminated water or another liquid sample was filtered using 0.2 μm pore size filters. The bacterial pellet that formed on the filter was resuspended in 200 μL of buffered solution. Subsequently, 100 μL of the resuspended pellet solution was injected into the functionalized chip to initiate antibody-based bacteria capture ( b). Following bacteria capture, a lysis buffer and a LAMP amplification cocktail were introduced. In detail, the chip operation sequence after bacteria capturing was as follows: PBS buffer injection to wash the unbound bacteria (total volume: 30 μL), LAMP reagent injection (total volume: 24 μL), chemical lysis with triton (included in the LAMP reaction) for 10 min, and chip heating on a plate at 65 °C for up to 60 min. DNA detection can be achieved visually by detecting the color change by the naked eye or by a camera. For increased capturing efficiency, the chip was operated in batch mode as follows: A sample volume equal to the microchannel volume (25 μL) was introduced and was allowed to react with the antibodies for an optimized time of 25 min. Then, the sample was pushed to the waste well and a second batch of equal volume was allowed to enter and react with the antibodies. Following bacteria capture on the microchannel surface from 4 batch samples, in situ lysis and specific DNA amplification were performed, followed by chromogenic naked-eye end-point detection to allow us to reach a decision on the potential presence of living bacteria with high sensitivity.
The sequences of the primers used in this work for DNA amplification via LAMP were as follows: F3_CGTTACCCACAGAAGAAGC; B3_ACCCTCTCCCATACTCGA; FIP_AGTAATTCCGATTAACGCTCGCAACCGGCTAACT CCGTGC; BIP_GGCGTAAAGGGTGCGTAGGTGACCAGTATTATCT GACCGTCC. The LAMP assay for Legionella detection is shown below: The LAMP reagent mix, with a total volume of 25 μL, contained the following: 12.5 μL of WarmStart Colorimetric Lamp 2X Master Mix (New England BioLabs, Ipswich, MA, USA), 2.5 μL of 10× primer mix (16 μM (each) of Forward Inner Primer (FIP) and Backward Inner Primer (BIP); 2 μM (each) of F3 and B3 primers), 8 μL of DNAse- and RNAse-free water, and 1 μL of Triton mix (10% Triton X-100 + 44 μL of DNAse-free water + 0.5 μL of 1 M Tris buffer pH 8.4 + 0.5 μL of 5% phenol red solution) for chemical lysis and 1 μL of any crude water sample, containing cells or not. After 5–10 min of waiting for the chemical lysis with Triton X-100 to take place, the chip was placed on a heating plate at 66 °C for up to 60 min. Amplification in the presence of target nucleic acids results in the production of protons that decrease the mixture’s pH, resulting in a color change in the phenol red from pink to yellow that is easily detectable by the naked eye. For validating amplification, the DNA amplicons were analyzed using electrophoresis on 2% agarose gel containing GelRed (Biotium, Lab Supplies, Athens, Greece) and visualized under UV light.
To facilitate the use of the amplification cocktail outside analytical laboratories, i.e., at the point of need or in resource-limited settings, the typical requirement for the storage of the reaction mixture components at −20 °C should be lifted. Thus, to allow room-temperature storage and create a stabilized form of the reaction mixture, lyophilization of the LAMP reagents for the detection of Legionella spp. was attempted. The addition of cryoprotective substances was investigated, as proposed by Klatser et al. and Agel et al. . In this work, it was found that the addition of the non-reducing sugar D-trehalose together with a small concentration of BSA stabilized the reagents during freeze drying. After lyophilization, the amplification cocktail was stored inside Mylar foil pouches together with desiccant bags. The Mylar foil pouches acted as an effective barrier against air moisture with very low water vapor transmission rates, oxygen, and light, resulting in a more stable dehydrated reagent mixture. The lyophilized reaction mixtures in the presence of D-trehalose/BSA, stored inside Mylar foil pouches, retained their activity for prolonged periods, i.e., more than 3 months under dry conditions at room temperature, without compromising the performance of the amplification cocktail.
In order to achieve a robust diagnosis for the presence of bacteria in the sample based on the color change in the microchannel, automated detection of the color change was attempted. To this end, a computational methodology was designed and realized in a software application, based on the analysis of camera photos of the chip, targeting an assessment of the color changes of the negative control and (sample) detection channels. The proposed methodology consisted of three steps following the acquisition of a photo of the chip after sample analysis. First, the automatic identification and isolation of the regions of interest of the chip photo, i.e., of the control and detection channels, were performed. Second, the color change in these areas (channels) was quantified through the measurement of a specific color index. Third, using standard samples of known bacterial concentrations, the proposed color index was assessed as a robust means on which the diagnosis result can be based. 1st step: To achieve the isolation of the detection and control channels, a machine learning model architecture, called U-Net , was trained using images from samples analyzed on chip. U-Net is a convolutional neural network architecture developed for image processing and object recognition applications. It consists of an “encoder” and a “decoder” that are combined with a connection bridge. The encoder detects high-level features in the image, while the decoder reconstructs an exact match of the original image. This architecture, shown schematically in , is highly efficient for tasks such as the requirement herein, namely the isolation of objects in an image (i.e., the detection and control channels). It must be emphasized that U-net produces a “segmentation map”; i.e., it produces an image with the same dimensions but with active pixels only in the areas of interest (control and detection channels). Therefore, by knowing the positions of the active pixels in the result, we can locate the corresponding positions in the original image. 2nd step: The color change was analyzed for each isolated channel. The digital quantification of a color involves the use of a model, such as the Hue Saturation Value (HSV) system. A colored image consists of a collection of pixels, with each pixel containing a set of three HSV coordinates. In our software, we used the hue value of the HSV model to quantify the channel color and, thus, based on its change, whether the sample was positive or negative. 3rd step: The assessment of the proposed hue value was carried out using samples of known bacterial concentrations, as will be shown in .
3.1. Sample Preconcentration Filtration-based sample preconcentration was implemented to reduce the sample volume from 200 mL to 100 μL, which is easily handled in the microfluidic channel. Filters from different materials were tested, all with a 0.2 μm pore size. To assess each filter’s capturing efficiency, the solution resuspended from the filter was cultured in plates, and the viable bacteria colonies were counted and compared to the original bacterial concentration spiked in the water sample. For comparison purposes, Salmonella bacteria were also used, in addition to L. pneumophila . In , the capturing efficiency for L. pneumophila and Salmonella is shown for different filters. a shows the capturing efficiency for L. pneumophila, upon the filtration of 5 × 10 3 CFU/mL of bacteria, while b shows the capturing efficiency for Salmonella when filtering a similar bacterial concentration on different filters. It can be observed that for L. pneumophila , the best working filters (with a capturing efficiency close to or higher than 50%) are poly(ether-sulfone) (PES) and nylon, while for Salmonella , the best working filters are nylon, cellulose acetate, and cellulose ester. 3.2. Bacteria Capturing Resuspended typical water sample volumes of 150–200 μL, equal to the filter volume, were used for increased detection sensitivity. However, flowing the water sample continuously through the chip may significantly reduce the bacteria capture efficiency. Therefore, the chip was operated in batch mode as follows: A sample volume equal to the microchannel volume (25 μL) was introduced and was allowed to react with the antibodies for a time of 25 min for enhanced bacteria capturing. Then, the sample was pushed to the waste well and a second batch of equal volume was allowed to enter and react with the antibodies. The sample introduction in four batches maintained the high efficiency of the chip while keeping the chip volume as low as possible and yet compatible with the volume of the amplification kit. The capturing efficiency of the chip was assessed for five different bacteria ( L. pneumophila , E-coli , Salmonella , Listeria , and B. cereus ), with the sample directly injected into the chip in four batches (as described above), after comparing the bacterial concentration exiting the micro/nanotextured microchannel that had been functionalized with the respective bacteria-specific antibodies (with an injected bacterial concentration of 200 CFU/100 μL). The results are shown in . The capture efficiency depended on the bacterial species and was between ~35% and 70% (the highest efficiency was observed for E. coli , Salmonella , and L. pneumophila ). 3.3. Optimization of Microchannel Depth for LAMP-Based DNA Detection In this work, among the different isothermal DNA amplification techniques, LAMP was chosen due to its utmost benefits (simple operation, robustness, and high sensitivity and specificity). Nucleotide incorporation by the DNA polymerase during amplification releases protons, changing the color of the pH-sensitive dye, phenol red, contained in the reaction mix. Since bacterial LAMP-based DNA detection is color-based, the microchannel depth should be optimized to achieve sufficient visible-light absorbance. To provide quantitative results, the absorbance at 430 nm and 560 nm was assessed by means of a UV–visible spectrophotometer (from θ-metrisis, Athens, Greece) equipped with a CCD detector, and the reaction performance was measured by the change in the ratio of light absorbed at 430 and 560 nm. End-point absorbance measurements of the LAMP-amplified DNA in the microchannels were performed for microchannels of increasing depth (from 100 to 140, 230, 310, 410, and 515 μm), and the absorbance ratio A430/A560 was calculated for all microchannel depths. The absorbance ratio A430/A560, which is proportional to the drop in pH occurring by the production of protons as the LAMP amplification progresses, is shown in as a function of the microchannel depth. As expected, the absorbance ratio A430/A560 increases as the microchannel depth increases. Thus, a microchannel depth of 500 μm was chosen as the depth sufficient to achieve a high absorbance ratio that can be easily manufactured by CNC in typical PMMA sheets. The chip design includes several innovations. As was shown in and herein, the microchannel’s geometrical characteristics were chosen to maximize the bacteria-capturing surface (microchannel length and roughness) and achieve high-contrast colorimetric detection (microchannel depth), while at the same time keeping the microchannel volume at 25 μL, for a reasonably low reagent volume and operation cost. 3.4. Validation of the LoC for Water Sample Analysis The proposed method for Legionella detection combines on-chip bacteria immunoaffinity capturing, chemical lysis, and DNA isothermal amplification, with all three steps accommodated in a single microfluidic chip, followed by simple color change detection in the presence of Legionella bacterial DNA in the sample. As explained in , after bacteria capturing on the microchannel wall, the on-chip lysis of captured cells (10 min) and on-chip LAMP-based DNA amplification (65 °C for up to 60 min) followed. The outcome of the amplification reaction based on colorimetric LAMP is observable by the naked eye for detecting the potential color change from pink (negative) to yellow (positive) in the case of bacterial presence in the water sample. To demonstrate that the LoC can operate with samples of different bacterial concentrations, absorbance measurements were performed for bacterial concentrations in the range of 0–10 4 CFU/100 mL in artificially contaminated water samples (from an L. pneumophila lenticule with standard concentration of 8.6 × 10 3 CFU), after on-chip capturing and LAMP amplification ( a). The absorbance, which is proportional to the pH reduction occurring due to the protons produced as the LAMP amplification progresses, increases significantly (by 50%) for bacterial concentrations between 0 (negative) and 100 CFU/100 mL, after which the absorbance increase is less prominent. The results were validated with gel electrophoresis ( b), indicating the sensitive detection of Legionella , even at concentrations as small as 100 CFU/100 mL (the Legionella concentration limit imposed by the new European legislation ). A similar trend was expected for the chip performance assessed by optical images of the chip at the end-point using artificially contaminated water with different bacterial concentrations, below and above the Legionella concentration limit imposed by the new EU legislation (100 CFU/100 mL). The presence of bacterial DNA caused the “sample” channel to turn yellow in color. A “negative control” channel was included, which always remains pink upon successful completion of the test. The images for the “negative” samples (below 100 CFU/100 mL) indicate a homogeneous pink color (<50 CFU/100 mL) or an inhomogeneous pink with orange areas (50 < CFU/100 mL < 100). The images for the “positive” samples show a color shift from faint yellow (for 100 < CFU/100 mL < 1000) to bright yellow (for >1000 CFU/100 mL) for increasing cell concentrations up to 10 4 CFU/100 mL. After 10 4 CFU, no significant color difference was observed, indicating saturation in absorbance, in agreement with a. The large color difference between the negative (<100 CFU/100 mL) and positive samples (>100 CFU/100 mL) ensures very sensitive naked-eye detection down to a few tens of cells. We note that the lowest detection limit for on-chip capturing and DNA amplification followed by color detection is between 50 and 100 CFU/100 mL, lower than the Legionella spp. concentration limit defined by the existing legislation for drinkable water (100 CFU/100 mL ). The analytical performance of the Legionella detection chip was assessed using two types of samples: water samples spiked with L. pneumophila lenticule CRM12821M (standard concentration: 8.6 × 10 3 CFU/lenticule) and real water samples provided from Athens Analysis Laboratories. For the spiked samples, a total of 130 samples were examined, with concentrations ranging from <50 CFU/100 mL to >1000 CFU/100 mL, created by dilutions of the aforementioned lenticule in 1000 mL. A direct comparison with ISO 11731:2017 showed 100% specificity (95% confidence interval, 99.1 to 100) and 99% sensitivity (95% confidence interval, 98.1 to 99.9). For the real water samples, a total of 111 samples were examined, collected from various hot- and cold-water distribution systems including water basins, taps (including potable water), showers, and water features/swimming pools. A direct comparison with ISO 11731:2017 showed 99.95% specificity (95% confidence interval, 98.07 to 99.93) and 97.9% sensitivity (95% confidence interval, 96.97 to 98.83). 3.5. Reproducibility To demonstrate the reproducibility of the on-chip bacteria capture directly from the water samples after preconcentration, L. pneumophila capturing was demonstrated directly from the spiked water samples injected into the chip. a depicts the capture efficiency (on freeze-dried antibodies) calculated after the plating of effluents coming out of the chip from experiments performed over a period of one year. The capture efficiency was stable over the testing period and equal to ~78% for injections corresponding to ~5 × 10 3 CFU/100 mL. To demonstrate the reproducibility of the chip regarding LAMP amplification and visualization, 10 batches of chips were fabricated over a period of 1 year and three chips from each batch were used each time. Representative images of the chip with positive and negative water samples are shown in b and indicate reproducible chip fabrication and DNA amplification, as demonstrated by the similar color change or no change for the positive and negative samples, respectively. 3.6. The Sensitivity of the Diagnostic Chip for L. pnemophilla Detection The performance of the method described herein is compared with the reference method ISO 11731:2017, which constitutes the gold-standard method used for the detection of Legionella. In detail, an experimental process was designed based on ISO 16140-4 factorial analysis. Specifically, three types of water samples (i.e., drinking water, non-drinking water, and drilling water) with known contamination concentrations, corresponding to three different levels of Legionella concentration, were used: a zero level (L 0 , do not contain Legionella), medium level (L 1 > 1–2 × 10 2 CFU/100 mL), and high level (L 2 > 2 × 10 3 CFU/100 mL). In each case, 15 repetitive analyses of each sample were performed. Representative images of the results are shown in . The zero-level (L 0 ) samples were tested to demonstrate that there were no false-positive results (cross-reactivity, e.g., with the water sample). The results from the chip-based method described herein and ISO 11731:2017 were in full agreement, demonstrating the good correlation of this method with the gold-standard method. The data were also used to calculate the sensitivity of the diagnostic chip, that is, the probability of a true-positive result concerning the detection performance of the chip, for these three types of water in three different concentrations. The sensitivity data with a 95% confidence interval are shown in . The data indicate the highest sensitivity (100%) for L. pneumophila detection in drinking water, lower sensitivity for drilling water, and even lower sensitivity (although higher than 67%) for non-drinking water. 3.7. Computational Image Analysis Computational color analysis of the camera images such as those in was performed to provide metrics for a robust quantification of the colors in the negative control and (sample) detection channels of the chip. A Hue Saturation Value (HSV)-based methodology was used to quantify the color of both channels (H detect and H control ) and visualize (as in ) the classified outcomes based on their color values and their corresponding bacterial concentration. Then, the result was compared with the actual labels to validate the success of the on-chip measurements. In , the hue values for the (sample) detection channels (positive and negative) as well as for the negative control channels are shown. The positive samples (>100 CFU/100 mL) contain hue values higher than 0.1, while the negative ones are between 0.9 and 1/0. The negative samples specifically show hue values close to the control hue values (with overlaps in the hue values close to ~0.93). Also, the negative samples with a concentration of 50 < CFU/100 mL < 100 are close to the negative ones with concentration values of <50 CFU/100 mL. Lastly, we see that the positive samples with concentrations of 100 < CFU/100 mL < 1000 and >1000 CFU/100 mL have a hue value difference larger than 0.05. Therefore, the proposed computational image analysis can provide robust semiquantitative sample analysis based on the hue value.
Filtration-based sample preconcentration was implemented to reduce the sample volume from 200 mL to 100 μL, which is easily handled in the microfluidic channel. Filters from different materials were tested, all with a 0.2 μm pore size. To assess each filter’s capturing efficiency, the solution resuspended from the filter was cultured in plates, and the viable bacteria colonies were counted and compared to the original bacterial concentration spiked in the water sample. For comparison purposes, Salmonella bacteria were also used, in addition to L. pneumophila . In , the capturing efficiency for L. pneumophila and Salmonella is shown for different filters. a shows the capturing efficiency for L. pneumophila, upon the filtration of 5 × 10 3 CFU/mL of bacteria, while b shows the capturing efficiency for Salmonella when filtering a similar bacterial concentration on different filters. It can be observed that for L. pneumophila , the best working filters (with a capturing efficiency close to or higher than 50%) are poly(ether-sulfone) (PES) and nylon, while for Salmonella , the best working filters are nylon, cellulose acetate, and cellulose ester.
Resuspended typical water sample volumes of 150–200 μL, equal to the filter volume, were used for increased detection sensitivity. However, flowing the water sample continuously through the chip may significantly reduce the bacteria capture efficiency. Therefore, the chip was operated in batch mode as follows: A sample volume equal to the microchannel volume (25 μL) was introduced and was allowed to react with the antibodies for a time of 25 min for enhanced bacteria capturing. Then, the sample was pushed to the waste well and a second batch of equal volume was allowed to enter and react with the antibodies. The sample introduction in four batches maintained the high efficiency of the chip while keeping the chip volume as low as possible and yet compatible with the volume of the amplification kit. The capturing efficiency of the chip was assessed for five different bacteria ( L. pneumophila , E-coli , Salmonella , Listeria , and B. cereus ), with the sample directly injected into the chip in four batches (as described above), after comparing the bacterial concentration exiting the micro/nanotextured microchannel that had been functionalized with the respective bacteria-specific antibodies (with an injected bacterial concentration of 200 CFU/100 μL). The results are shown in . The capture efficiency depended on the bacterial species and was between ~35% and 70% (the highest efficiency was observed for E. coli , Salmonella , and L. pneumophila ).
In this work, among the different isothermal DNA amplification techniques, LAMP was chosen due to its utmost benefits (simple operation, robustness, and high sensitivity and specificity). Nucleotide incorporation by the DNA polymerase during amplification releases protons, changing the color of the pH-sensitive dye, phenol red, contained in the reaction mix. Since bacterial LAMP-based DNA detection is color-based, the microchannel depth should be optimized to achieve sufficient visible-light absorbance. To provide quantitative results, the absorbance at 430 nm and 560 nm was assessed by means of a UV–visible spectrophotometer (from θ-metrisis, Athens, Greece) equipped with a CCD detector, and the reaction performance was measured by the change in the ratio of light absorbed at 430 and 560 nm. End-point absorbance measurements of the LAMP-amplified DNA in the microchannels were performed for microchannels of increasing depth (from 100 to 140, 230, 310, 410, and 515 μm), and the absorbance ratio A430/A560 was calculated for all microchannel depths. The absorbance ratio A430/A560, which is proportional to the drop in pH occurring by the production of protons as the LAMP amplification progresses, is shown in as a function of the microchannel depth. As expected, the absorbance ratio A430/A560 increases as the microchannel depth increases. Thus, a microchannel depth of 500 μm was chosen as the depth sufficient to achieve a high absorbance ratio that can be easily manufactured by CNC in typical PMMA sheets. The chip design includes several innovations. As was shown in and herein, the microchannel’s geometrical characteristics were chosen to maximize the bacteria-capturing surface (microchannel length and roughness) and achieve high-contrast colorimetric detection (microchannel depth), while at the same time keeping the microchannel volume at 25 μL, for a reasonably low reagent volume and operation cost.
The proposed method for Legionella detection combines on-chip bacteria immunoaffinity capturing, chemical lysis, and DNA isothermal amplification, with all three steps accommodated in a single microfluidic chip, followed by simple color change detection in the presence of Legionella bacterial DNA in the sample. As explained in , after bacteria capturing on the microchannel wall, the on-chip lysis of captured cells (10 min) and on-chip LAMP-based DNA amplification (65 °C for up to 60 min) followed. The outcome of the amplification reaction based on colorimetric LAMP is observable by the naked eye for detecting the potential color change from pink (negative) to yellow (positive) in the case of bacterial presence in the water sample. To demonstrate that the LoC can operate with samples of different bacterial concentrations, absorbance measurements were performed for bacterial concentrations in the range of 0–10 4 CFU/100 mL in artificially contaminated water samples (from an L. pneumophila lenticule with standard concentration of 8.6 × 10 3 CFU), after on-chip capturing and LAMP amplification ( a). The absorbance, which is proportional to the pH reduction occurring due to the protons produced as the LAMP amplification progresses, increases significantly (by 50%) for bacterial concentrations between 0 (negative) and 100 CFU/100 mL, after which the absorbance increase is less prominent. The results were validated with gel electrophoresis ( b), indicating the sensitive detection of Legionella , even at concentrations as small as 100 CFU/100 mL (the Legionella concentration limit imposed by the new European legislation ). A similar trend was expected for the chip performance assessed by optical images of the chip at the end-point using artificially contaminated water with different bacterial concentrations, below and above the Legionella concentration limit imposed by the new EU legislation (100 CFU/100 mL). The presence of bacterial DNA caused the “sample” channel to turn yellow in color. A “negative control” channel was included, which always remains pink upon successful completion of the test. The images for the “negative” samples (below 100 CFU/100 mL) indicate a homogeneous pink color (<50 CFU/100 mL) or an inhomogeneous pink with orange areas (50 < CFU/100 mL < 100). The images for the “positive” samples show a color shift from faint yellow (for 100 < CFU/100 mL < 1000) to bright yellow (for >1000 CFU/100 mL) for increasing cell concentrations up to 10 4 CFU/100 mL. After 10 4 CFU, no significant color difference was observed, indicating saturation in absorbance, in agreement with a. The large color difference between the negative (<100 CFU/100 mL) and positive samples (>100 CFU/100 mL) ensures very sensitive naked-eye detection down to a few tens of cells. We note that the lowest detection limit for on-chip capturing and DNA amplification followed by color detection is between 50 and 100 CFU/100 mL, lower than the Legionella spp. concentration limit defined by the existing legislation for drinkable water (100 CFU/100 mL ). The analytical performance of the Legionella detection chip was assessed using two types of samples: water samples spiked with L. pneumophila lenticule CRM12821M (standard concentration: 8.6 × 10 3 CFU/lenticule) and real water samples provided from Athens Analysis Laboratories. For the spiked samples, a total of 130 samples were examined, with concentrations ranging from <50 CFU/100 mL to >1000 CFU/100 mL, created by dilutions of the aforementioned lenticule in 1000 mL. A direct comparison with ISO 11731:2017 showed 100% specificity (95% confidence interval, 99.1 to 100) and 99% sensitivity (95% confidence interval, 98.1 to 99.9). For the real water samples, a total of 111 samples were examined, collected from various hot- and cold-water distribution systems including water basins, taps (including potable water), showers, and water features/swimming pools. A direct comparison with ISO 11731:2017 showed 99.95% specificity (95% confidence interval, 98.07 to 99.93) and 97.9% sensitivity (95% confidence interval, 96.97 to 98.83).
To demonstrate the reproducibility of the on-chip bacteria capture directly from the water samples after preconcentration, L. pneumophila capturing was demonstrated directly from the spiked water samples injected into the chip. a depicts the capture efficiency (on freeze-dried antibodies) calculated after the plating of effluents coming out of the chip from experiments performed over a period of one year. The capture efficiency was stable over the testing period and equal to ~78% for injections corresponding to ~5 × 10 3 CFU/100 mL. To demonstrate the reproducibility of the chip regarding LAMP amplification and visualization, 10 batches of chips were fabricated over a period of 1 year and three chips from each batch were used each time. Representative images of the chip with positive and negative water samples are shown in b and indicate reproducible chip fabrication and DNA amplification, as demonstrated by the similar color change or no change for the positive and negative samples, respectively.
The performance of the method described herein is compared with the reference method ISO 11731:2017, which constitutes the gold-standard method used for the detection of Legionella. In detail, an experimental process was designed based on ISO 16140-4 factorial analysis. Specifically, three types of water samples (i.e., drinking water, non-drinking water, and drilling water) with known contamination concentrations, corresponding to three different levels of Legionella concentration, were used: a zero level (L 0 , do not contain Legionella), medium level (L 1 > 1–2 × 10 2 CFU/100 mL), and high level (L 2 > 2 × 10 3 CFU/100 mL). In each case, 15 repetitive analyses of each sample were performed. Representative images of the results are shown in . The zero-level (L 0 ) samples were tested to demonstrate that there were no false-positive results (cross-reactivity, e.g., with the water sample). The results from the chip-based method described herein and ISO 11731:2017 were in full agreement, demonstrating the good correlation of this method with the gold-standard method. The data were also used to calculate the sensitivity of the diagnostic chip, that is, the probability of a true-positive result concerning the detection performance of the chip, for these three types of water in three different concentrations. The sensitivity data with a 95% confidence interval are shown in . The data indicate the highest sensitivity (100%) for L. pneumophila detection in drinking water, lower sensitivity for drilling water, and even lower sensitivity (although higher than 67%) for non-drinking water.
Computational color analysis of the camera images such as those in was performed to provide metrics for a robust quantification of the colors in the negative control and (sample) detection channels of the chip. A Hue Saturation Value (HSV)-based methodology was used to quantify the color of both channels (H detect and H control ) and visualize (as in ) the classified outcomes based on their color values and their corresponding bacterial concentration. Then, the result was compared with the actual labels to validate the success of the on-chip measurements. In , the hue values for the (sample) detection channels (positive and negative) as well as for the negative control channels are shown. The positive samples (>100 CFU/100 mL) contain hue values higher than 0.1, while the negative ones are between 0.9 and 1/0. The negative samples specifically show hue values close to the control hue values (with overlaps in the hue values close to ~0.93). Also, the negative samples with a concentration of 50 < CFU/100 mL < 100 are close to the negative ones with concentration values of <50 CFU/100 mL. Lastly, we see that the positive samples with concentrations of 100 < CFU/100 mL < 1000 and >1000 CFU/100 mL have a hue value difference larger than 0.05. Therefore, the proposed computational image analysis can provide robust semiquantitative sample analysis based on the hue value.
In the present study, an easy-to-use, compact chip integrating bacteria capturing, lysis, and DNA amplification, all in one microchannel, was developed for the colorimetric, naked-eye, semiquantitative (POS/NEG, based on a color code) detection of Legionella present in water samples. A rapid preconcentration (filtering) step was also developed and the optimum commercial filter material (PES) was decided for Legionella , demonstrating a capturing efficiency of 70%. Following this off-chip step, a concentrated sample volume (100 μL) was injected into the chip in four batches to accommodate the small chip volume (25 μL). Each sample batch was incubated for 25 min in the chip for the optimum capturing efficiency of Legionella (78%) on the functionalized plasma micro/nanotextured PMMA walls of the microchannel. To allow the convenient long-term storage of the chips (up to 18 months), facilitate their transport at ambient temperature, and simplify the process for the end-user, the antibodies immobilized onto the microchannel were lyophilized, thus providing stable-over-time functionality. Following bacteria capturing, the LAMP amplification mixture, containing a lysis buffer, was injected and incubated in the microchannel for 10 min before the chip was placed on a hot plate (heated to 66 °C) for a maximum time of 60 min to complete bacterial DNA amplification and naked-eye detection, thanks to a color change from pink to yellow in the case of positive Legionella -containing water samples. In addition, the lyophilization of the LAMP reagents was achieved in the presence of D-trehalose/BSA to allow the prolonged (more than 3 months) storage of these reagents at room temperature and their use outside analytical laboratories, i.e., at the point of need or in resource-limited settings. Despite the fact that the filtering step and the reagent injections are performed manually, they are indeed easy to be performed by non-skilled personnel, allowing, in addition to lyophilization, ease of usage for the proposed chip at the point of need. Furthermore, the method presented herein for the detection of Legionella was validated with the reference method ISO 11731:2017 providing the same POS/NEG results and was found to be in full agreement for all the water types and concentration levels tested. It showed the highest sensitivity for drinking water (100%) and the lowest for non-drinking water (66.7%). The recommended time for the LAMP reaction is 60 min as it is appropriate for both negative and positive samples, even at concentrations around the legislation limit. In addition to the color code for semiquantitative results based on end-point color detection, a computational image analysis was also developed. From the HSV-based methodology, we conclude the following: (1) the samples with different concentrations are numerically separated based on the hue color value; (2) the positive and negative samples are separated clearly; (3) the positive samples with concentrations in the range of 100–1000 CFU/100 mL and greater than 1000 CFU/100 mL are well separated; and (4) the negative samples show hue color values very close to the negative control ones. The chip operation as it was presented in this work is intended for semiquantitative (POS/NEG, based on a color code and computational image analysis) water sample analysis. However, quantitative analysis is also possible through the same computational methodology, considering the quantification of the color change as a function of time during the amplification of different L. pneumophila sample concentrations and of the negative control channels. Such computational work is currently in progress. The presented chip design includes several innovations. As was shown, the microchannel depth was maximized to maximize absorption and, thus, provide high-contrast colorimetric detection, while at the same time keeping the capturing efficiency of the chip (~80%) reasonably high and the microchannel volume at 25 μL, requiring a reasonably low reagent volume and operation cost. In addition, the chip footprint allows the integration of several microchannels on a 5 × 8 cm 2 PMMA plate, thus allowing the parallel testing of up to five different water samples on the same plate, further reducing the analysis cost. It is worth highlighting that the chip presented herein is highly specific for the detection of viable L. pneumophila bacteria due to their antibody immunocapturing and specific DNA amplification. Further, the chip was tested for the detection of Legionella spp., which was successfully demonstrated. In the future, the chip can be used to detect and discriminate between various Legionella serotypes, e.g., L. pneumophila serogroups 1–13, L. longbeachae , and L. bozemanii . In addition, with the appropriate selection of specific antibodies and primers, the chip can be used for the detection of other bacteria contaminating water samples. In this work, in addition to Legionella , the on-chip preconcentration and capturing of Salmonella bacteria were demonstrated.
A simple, compact chip was presented for the rapid detection of Legionella pneumophila in water samples, integrating for the first time on-chip sample preparation (by bacteria capture, lysis, and LAMP-based amplification of bacterial DNA) with naked-eye or image analysis-based semiquantitative end-point detection, reducing the analysis time required by the gold-standard method by 99%. Compared to PCR, the LAMP-based method presented herein demonstrates many advantages. Although the reduction in analysis time is not as impressive as that of the gold standard method, the LAMP assay exhibits higher sensitivity and specificity than PCR and higher robustness in the presence of certain environmental compounds that often cause the inhibition of PCR, and it enables performing the test in the absence of any sophisticated equipment, allowing its use at the point of need. In fact, the chip features double specificity, combining bacteria capturing on specific antibodies and LAMP-specific primers. In addition, it enables naked-eye qualitative or image analysis-based semiquantitative end-point detection thanks to an optimized microchannel depth, and exhibits high sensitivity and reproducibility, user-friendliness, and long-term storage (18 months) thanks to the lyophilization of antibodies immobilized on the plasma-enhanced surface area of the microfluidic channel. Its analytical performance was validated with the reference method ISO 11731:2017 (gold standard), and it was found to be highly sensitive (100–86.6%) for the analysis of drinking and drilling water, as well as sensitive enough (100–66.7%) for the analysis of non-drinking water (depending on the bacterial contamination level of water). Given the detection limit achieved (50–100 CFU/100 mL), the short analysis time, and the aforementioned advantages of the chip, the proposed platform can have a huge impact in tackling Legionella outbreaks and generally find widespread application for fast and reliable waterborne-bacteria detection.
A patent entitled “Diagnostic chip for analyzing the presence of bacteria in a sample”, Patent No GR1010186, has been awarded by the Greek Industrial Property Organization (OBI).
|
Human limits in machine learning: prediction of potato yield and disease using soil microbiome data | aaeaa5db-6809-4ebd-bd59-d70699c28fe9 | 11600749 | Microbiology[mh] | Machine learning (ML) has transformed how scientific research is conducted in recent years. Among the many tasks performed by ML models in our daily lives, researchers have relied on ML to assist in clinical diagnoses , identification of bacterial phenotypes such as antimicrobial resistance , and even identification of objects in space . Recently, the vast evidence of the prediction power of ML models on a wide range of applications has launched the adoption of these models in other domains such as sustainable agriculture where soil health – characterized by a wide range of biological, chemical, and physical properties —is explored as an important driver to predict plant phenotypes, such as disease susceptibility or yield. The question of whether the improvement of soil health could result in superior crop yield and disease resistance remains open, as researchers have not been able to identify a set of indicators to accurately and robustly predict plant outcomes from soil information. Among all the candidate indicators for soil health, soil microbiome is one that continues to be understudied in its predictive potential in the productivity and resilience of agricultural ecosystems . Amplicon sequencing of highly preserved, phylogenetically informative marker genes, such as the 16 S ribosomal RNA for bacteria and the internal transcribed spacer (ITS) for fungi, has enabled extensive studies on the complexity and diversity of soil microbial communities over the last decade. However, little is still understood about how changes in these microbial communities directly impact plant growth and health. Although it is well-recognized that the soil microbiome is closely related to plant health and productivity , most current research focuses on how microbial communities change in response to agricultural management. Fortunately, ML models can bridge this gap by handling complex, high-dimensional data for predictions without requiring prior knowledge of variable interactions. Since we have little or no knowledge of most species among the thousands contained in the soil microbial communities, it would be beneficial to exploit the power of ML models on microbiome data as they are able to explore the unknown interactions between the microbial communities and plant phenotypes. Analyzing microbiome datasets with ML has three main challenges: (1) The data is compositional, meaning raw counts are normalized by the total number of reads per sample. Therefore, microbial abundances are not independent, and the use of traditional statistical techniques (such as correlation) might lead to increased false discovery rates . (2) Data are highly sparse, which means that datasets include a large number of operational taxonomic units (OTUs) that are present in a small proportion of samples (or in none at all) . (3) Data are high-dimensional, which means that the number of OTUs are larger than the number of samples, especially in more specific taxonomic orders like Order, Family, and Genus. Beyond microbiome data challenges, prediction accuracy decreases with imbalanced or inaccurate labels. That is, when we consider the task of prediction, each microbial sample should be labeled by a plant outcome value (say, high yield or low yield). It is common that labels in these datasets are imbalanced with one class representing the majority of observed samples which is denoted as an imbalanced binary classification problem. Furthermore, decisions on the class labels are many times not straight-forward. While diseased plant or non-diseased plant tends to be an indisputable classification, determining what constitutes high versus low yield is up for debate and depends not only on the season and other environmental factors, but also on the specific crop variety. Here, we explore the potential of ML models in the prediction of plant phenotypes of yield and disease from soil data. We utilize a dataset from potato fields in Wisconsin and Minnesota and focus on the performance of the models while facing data challenges related to binarization, imbalance, compositionality, sparsity and high dimensionality. Furthermore, we test the impact of specific data preprocessing steps such as (1) different normalization and zero replacement strategies to overcome the compositionality and sparsity, (2) different feature selection strategies to overcome the high dimensionality, and (3) data augmentation techniques to overcome the imbalance of the data. In addition, we are also interested in answering the question of whether soil microbiome data alone has enough predictive power to predict the disease and yield phenotypes or whether other information on the soil, such as chemical composition, is necessary for accurate prediction. The answer to this question will inform farmers on the best data collection strategies given that soil microbiome data is expensive. That is, if the soil microbiome data does not provide enough predictive power compared to other (cheaper) soil measurements, then the collection of microbiome data for prediction purposes would be futile. For our investigation, we have chosen two distinct machine learning models. First, we employ random forests (RF), which have consistently demonstrated superior predictive performance across various domains . Further details about this method can be found in section “ ”. Second, we utilize a Bayesian Neural Network (Bayesian NN), known for its inherent protection against overfitting (See section “ ”). Given their capacity to capture complex interactions among features, neural networks are valuable, especially when dealing with unknown relationships. However, considering the small sample size and large feature space in our study, traditional back-propagation-based neural network models may exhibit a substantial bias or overfitting , hence the need to explore the Bayesian version. We initially explored a variety of machine learning methods to assess precision across different responses. However, given our goal of finding a generalized model that could reliably predict multiple response types, we selected Random Forest (RF). This decision was supported by the model’s consistently comparable results to the H 2 O AutoML package in Python , which considers different machine learning algorithms and selects the best-performing ones (See Fig. ). Additionally, we employed Bayesian Neural Networks (Bayesian NNs) because of their capacity to model parameter uncertainty and their potential to avoid overfitting, particularly useful when working with small datasets. Initially, we worked with continuous yield values, but due to the models’ inability to predict yield reliably (See Fig. ), we binarized the response to provide more robust and reliable methods. Among the main findings, we can highlight that microbiome data alone indeed has predictive power for disease outcomes, especially for pitted scab disease, but not to predict yield. We also find that the most powerful prediction is achieved by combination of environmental information and microbiome data. Among the data preprocessing strategies that we explored, we find that normalization and zero replacement strategies have a huge impact on the prediction power of the models, yet there are strong interaction effects with taxonomic levels, and thus, it is impossible to identify one strategy that outperforms the others. In terms of the model, RF outperforms other supervised classification machine learning models, including Bayesian NN which are more computationally expensive and may not be suitable for datasets with a large number of predictors. We conclude our investigation with a full model selection (FMS) decision tree approach to identify optimal combinations of normalization, zero replacement, feature selection, and model choices that maximize prediction accuracy for microbiome data analysis. Recommendations include using data augmentation and more specific taxonomic levels like Family and Genus for RF, but more general levels like Phylum without augmentation for Bayesian NN, while also identifying specific normalization methods suitable for both models. All the technical terms used in this study are defined in Table . Figure shows a graphical representation of our pipeline with three major steps: (1) data preparation, (2) feature selection, and (3) classification based on random forest (RF) and Bayesian NN models with various types of predictors, including microbiome data (OTUs), environmental, and a combination of both. We describe each step in the pipeline in the next subsections. Data description and data processing Data description In this study, we focus on the soil microbiome (matrix of abundances) in a variety of taxonomic orders, including Phylum, Class, Order, Family, and Genus as well as other environmental information from soil samples acquired from potato fields in Wisconsin and Minnesota. The dataset consists of measurements related to soil health, potato yield and soil quality information. The soil health data were collected in the fall of 2019 from pre-planting commercial potato fields and include soil physicochemical properties, soil microbiome composition, soil microbiome diversity, and soil pathogen abundance. Soils were collected from 36 Minnesota fields and 66 Wisconsin fields, with three bulk soil samples randomly selected from each field. The potato yield and quality information at each sampling location was measured at the end of the growing season (September of 2020) including tuber yield and disease severity. Overall, we have 256 samples, 108 of which are taken from fields in Minnesota, and 148 of which are taken from fields in Wisconsin. While this provides a strong foundation for modeling the upper Midwest potato-growing region, we acknowledge that the use of data from a single growing season and geographic region limits the generalizability of our findings. Future data collection from multiple growing seasons and regions is underway to enhance the robustness and applicability of our models. We list all measurements in Table in the Supplementary Material. Soil physicochemical properties Fresh field soils were measured for a variety of physicochemical properties in the Agvise soil testing lab (Benson, Minnesota). Measurements of soil pH, organic matter content, carbon fractions, organic nitrogen, macro and micronutrients are described in . Soil texture was measured by quantifying the relative amount of sand, silt, and clay using a hydrometer. Soil cation/anion exchange capacity was calculated from the nutrient test results mentioned above, reported as milliequivalents per 100 gs of soil. Soil microbial community composition and population abundance Soil microbial community was characterized by high-throughput sequencing of the bacterial 16 S rRNA gene and fungal ITS2 regions. A subsample of 0.25 g of frozen field soils were extracted with the DNeasy PowerSoil Pro DNA isolation kit (Qiagen, CA). Extracted DNA was used in a two-step PCR reaction , with the V3-V4 region of bacterial 16 S rRNA and the eukaryotic ITS2 region amplified using the primer set V3F and 806R, and 5.8S and ITS4, respectively . The final PCR product was normalized, pooled and cleaned-up before sequenced on a Illumina MiSeq platform at the University of Minnesota Genomics Center. Sequences were analyzed using Qiime2. Cutadapt was first used to remove the forward and reverse primers of the ITS reads. Trimmed ITS reads and the raw 16 S reads were then truncated, filtered, denoised, pair-end merged, and chimeras removed using the DADA2 pipeline. Taxonomy was assigned to the feature table of amplicon sequence variant (ASV) using Qiime2’s feature-classifier plugin, using the RDP Naïve Bayesian Classifier fit to the SILVA 138 database for 16 S reads and UNITE database for ITS reads. Bacterial and fungal ASV tables were merged at Phylum, Class, Order, Family, and Genus levels using phyloseq in R. Alpha diversity measured as Chao1, Abundance-based coverage estimator, Shannon, Simpson, and Inverse Simpson index were calculated after rarefying the samples to the minimum sample depth. Alpha diversity was calculated for each taxonomic level using the vegan package . The population abundance of bacteria, fungi, Verticillium dahliae , and Pathogenic Streptomyces were measured with quantitative polymerase chain reaction as described in . Yield and disease Potatoes were harvested by hand from a one-meter hill (usually 3-4 plants) at each sampling location at the end of the growing season. One plant was used for tuber disease assessment, and the rest plants were used for yield estimation. Tubers were visually assessed for common scab, silver scurf, and black scurf, and then cut-open to evaluate Verticillium dahliae infection (dark vascular ring), and hollow heart. Tuber yield was estimated as the fresh weight of cleaned tubers. Binarization of response variables We have six phenotypes (response variables) of interest, four of them correspond to diseases and two of them correspond to yield (Table ). All six responses in the dataset are continuous, so we need to binarize them to fit the classification models. For the disease-related responses, we simply make the binary label 0 if there is no presence of disease, and 1 if there was detection of disease (that is, if the continuous response is greater than 0.0). Binarizing the yield response variables is harder as there is no universal standard to classify potato yield to be low or high. Furthermore, yield values are highly dependent on the type of potato variety. We assign a label of 0 (low yield) to samples with a yield less than the variety-specific median. Similarly, we assign a label of 1 (high yield) to samples with a yield greater than the variety-specific median. We illustrate this approach in Fig. in the Supplementary Material. After binarization, we note that pitted scab disease (denoted Scabpit in the figures), superficial scab disease (denoted Scabsuper in the figures), and both yield responses are balanced, whereas other responses are highly imbalanced: scab disease (denoted Scab in the figures) has 80% of samples labeled as 1, and black scurf disease (denoted Black_Scurf in the figures) has only six samples labeled as 1. We use these imbalanced cases to assess the performance of the methods under imbalance settings and data augmentation strategies. Data filtering, normalization, and zero replacement The input data is a matrix with non-negative read counts that were generated by a sequencing procedure. Let [12pt]{minimal} $$w^{(k)}= [w_1^{(k)},...,w_p^{(k)}]$$ w ( k ) = [ w 1 ( k ) , . . . , w p ( k ) ] be the total read counts of sample k containing p OTUs, where [12pt]{minimal} $$w^{(k)}$$ w ( k ) is a composition that adds up to a fixed value of [12pt]{minimal} $$m^{(k)}= _{i=1}^p w_i^{(k)}$$ m ( k ) = ∑ i = 1 p w i ( k ) . This value [12pt]{minimal} $$m^{(k)}$$ m ( k ) is the sequencing depth, which varies across samples and is predetermined by technical factors resulting in highly sparse data. It is reasonable to filter out a certain set of OTUs as the first data preparation step. For filtering, we only include OTUs that appear in at least 15 samples. Table displays the number of features (OTUs) before and after filtering for different taxonomic levels. As mentioned, the input data is compositional and highly sparse. It is known that ML methods do not perform well with unnormalized data and with sparse data . Therefore, we explore the effect of four zero replacement strategies (to overcome sparsity) and five normalization strategies (to overcome compositionality). All strategies are implemented in the NetCoMi R package . In particular, we consider the four zero replacement strategies: (1) the original dataset which included zeros (denoted none in the figures), (2) pseudo-zero replacement which replaces zero counts by a predefined pseudo count (denoted pseudo in the figures), (3) multiplicative zero replacement which imputes left-censored compositional values by a given fraction and applies a multiplicative adjustment to preserve the multivariate compositional properties of the samples (denoted multRepl in the figures) , and (4) Bayesian-multiplicative treatment which imputes zero counts by posterior estimates of the multinomial probabilities generating the counts, assuming a Dirichlet prior distribution (denoted bayesMult in the figures) . Next, we use five normalization methods: (1) Total sum scaling which simply converts counts to appropriately scaled ratios (denoted TSS in the figures) , (2) Cumulative sum scaling which rescales the samples based on a subset (quartile) of lower abundant taxa, thereby excluding the impact of highly abundant taxa (denoted CSS in the figures) , (3) Common sum scaling in which counts are scaled to the minimum depth of each sample (denoted COM in the figures) , (4) Rarefying which random samples without replacement after a minimum count threshold has been applied (denoted rarefy in the figures) , and (5) Centered Log-ratio which transforms the data using the geometric mean as the reference (denoted clr in the figures) . With four zero replacement methods and five normalization methods, we create 20 datasets by the combination of zero replacement and normalization techniques. Our goal is to study the effect of the zero replacement and normalization choice in the performance of the deep learning methods. Namely, we have the following 20 combinations, [12pt]{minimal} $$ {NM}_1$$ NM 1 : TSS+none, [12pt]{minimal} $$ {NM}_2$$ NM 2 : TSS+pseudo, [12pt]{minimal} $$ {NM}_3$$ NM 3 : TSS+multRepl, [12pt]{minimal} $$ {NM}_4$$ NM 4 : TSS+bayesMult, [12pt]{minimal} $$ {NM}_5$$ NM 5 : CSS+none, [12pt]{minimal} $$ {NM}_6$$ NM 6 : CSS+pseudo, [12pt]{minimal} $$ {NM}_7$$ NM 7 : CSS+multRepl, [12pt]{minimal} $$ {NM}_8$$ NM 8 : CSS+bayesMult, [12pt]{minimal} $$ {NM}_9$$ NM 9 : COM+none, [12pt]{minimal} $$ {NM}_{10}$$ NM 10 : COM+pseudo, [12pt]{minimal} $$ {NM}_{11}$$ NM 11 : COM+multRepl, [12pt]{minimal} $$ {NM}_{12}$$ NM 12 : COM+bayesMult, [12pt]{minimal} $$ {NM}_{13}$$ NM 13 : rarefy+none, [12pt]{minimal} $$ {NM}_{14}$$ NM 14 : rarefy+pseudo, [12pt]{minimal} $$ {NM}_{15}$$ NM 15 : rarefy+multRepl, [12pt]{minimal} $$ {NM}_{16}$$ NM 16 : rarefy+bayesMult, [12pt]{minimal} $$ {NM}_{17}$$ NM 17 : clr+none, [12pt]{minimal} $$ {NM}_{18}$$ NM 18 : clr+pseudo,, [12pt]{minimal} $$ {NM}_{19}$$ NM 19 : clr+multRepl, and [12pt]{minimal} $$ {NM}_{20}$$ NM 20 : clr+bayesMult. For convenience, we use the notation [12pt]{minimal} $$ {NM}_i$$ NM i (Normalization Method) for [12pt]{minimal} $$i=1, ,20$$ i = 1 , ⋯ , 20 in the Full Model Selection section (See section “ ”). For the environmental predictors of soil chemistry and microbial population density in the soil, we apply six scaling methods: (1) standardize features by subtracting the mean and scaling to unit variance ; (2) scale each feature to a [0, 1] range; (3) scale each feature by its maximum absolute value; (4) scale features by subtracting the median and scaling to the interquartile range ; (5) transform the features to follow a normal distribution ; (6) normalize samples individually to the unit norm. After normalization, the datasets are split into training, validation, and testing sets with 10-fold cross-validation. We used 80% of samples for training and validation, and 20% for testing. Data augmentation There are three main goals that we wish to achieve with data augmentation: (1) improve the model’s prediction performance with more artificial samples; (2) balance the number of labels with artificial samples, and (3) make the model more robust and avoid overfitting with unseen (artificial) data. We note that augmenting the whole dataset and then splitting it into training and testing sets would result in data leakage . For example, when the original sample is in the testing set and the augmented sample from this sample is in the training set, the model is essentially training and testing on the same sample since the normalized values of OTUs are very close. Thus, we split the data into training and testing sets first and only augment the training set. This strategy also allows us to have a fair performance comparison for augmented and non-augmented sets with the same testing data. Regarding the data augmentation procedure, instead of simply adding a randomly generated noise to the original sample, we subset the data by variety and label, compute the mean (and standard deviation) abundance value for this subset, and create a new sample that includes the original data plus a Gaussian error with mean [12pt]{minimal} $$ /100$$ μ / 100 and standard deviation [12pt]{minimal} $$ /100$$ σ / 100 where [12pt]{minimal} $$ , $$ μ , σ are the subset-specific mean and standard deviation, respectively. This approach is illustrated in Fig. in the Supplementary Material. By the end of this procedure we would have a balanced augmented training set with 400 samples per label for each of the five taxonomic levels (Phylum, Class, Order, Family, and Genus), the number of samples is shown in Table . Feature selection Feature selection involves the identification of important features (or covariates) that have high predictive power. Given the high-dimensionality of the data (e.g. 256 original samples for 485 OTUs at the Genus level), feature selection is necessary, especially for Bayesian NN models that are computationally expensive. We pursue two approaches for feature selection: (1) using ML models to assess variable importance, and (2) using network analyses. To focus exclusively on the effect of feature selection, we only consider one type of normalization and zero replacement strategy in this investigation, namely, total sum scaling normalization without zero replacement ( [12pt]{minimal} $$ {NM}_1$$ NM 1 : TSS+none). Using ML models for feature selection To identify important OTUs, we use six ML strategies implemented in scikit-learn : (1) “SelectKBest” method selects features based on the k highest analysis of variance F-value scores, (2) select the top k features based on the mutual information statistic, (3) recursive feature elimination (RFE) with logistic regression, (4) RFE with decision tree, (5) RFE with gradient boosting, and (6) RFE with RF. In addition to the six ML strategies, we consider a 7th strategy which consists in including OTUs in the model if their maximum value is within the top 30%. After running all seven strategies, we assign a value (“TOTAL”) to each OTU based on the number of times the OTU is selected as an important feature under the seven criteria. That is, an OTU that is selected as important by all seven strategies will have a value of 7. The OTUs are sorted based on “TOTAL” column and the top 30% of them are selected as important features. Thus, 30, 36, 75, 85, and 162 OTUs are selected for Phylum, Class, Order, Family, and Genus levels, respectively. For example, Table in the Supplementary Material shows the top 30 OTUs at the Phylum level and by which strategies they are identified as important features to predict the pitted scab response. Using network comparison for feature selection Next, we identify important OTUs by comparing their interactions in microbial networks when the network is constructed with samples from one class (say, low yield) versus with samples from the other class (say, high yield) . Indeed, evidence shows that microbial interactions can help differentiate between crop disease states . We identify the OTUs that interact differently on the samples corresponding to one class versus another class as relevant OTUs that contain information about the outcome. We use the SPRING method (semi-parametric rank-based approach for inference in the graphical model) in the NetComi package to construct two microbial networks: one corresponding to label 0 and one corresponding to label 1. To build these networks, the estimated partial correlations are transformed into dissimilarities via the signed distance metric, and the corresponding similarities are then used as edge weights. We then compare these networks in order to identify OTUs that have the greatest difference based on degree values. Figure shows an example of two microbial networks with different graphical structures, and Table in the Supplementary Material lists OTUs ranked by the difference in degree in the two microbial networks for diseased and non-diseased classes of pitted scab response. The OTUs are sorted based on the values of the degree difference column and similar to ML strategy top 30% of them are selected as important features. Combination of ML and network strategies in feature selection We define a scoring value for each OTU based on whether they are identified as important by ML strategy ( [12pt]{minimal} $$score=1$$ s c o r e = 1 ), by the network comparison ( [12pt]{minimal} $$score=2$$ s c o r e = 2 ), or both ( [12pt]{minimal} $$score=3$$ s c o r e = 3 ). If the OTU is not identified as important by any strategy, it is denoted [12pt]{minimal} $$score=0$$ s c o r e = 0 . Figure displays the scoring values of all OTUs at the Phylum level for all responses. Model descriptions: random forest and Bayesian neural network We apply two types of classification models to predict potato disease and yield: RF and Bayesian NN. We consider separately three types of predictors: OTU abundances (5 taxonomic levels and 20 normalization/zero replacement strategies described in section “ ”), environmental predictors such as soil characteristics and microbial population density, and a combination of both types. Table lists all the models we consider in this study. Table shows the number of predictors included in each model as well as the data choices related to taxonomic level or normalization and zero replacement strategies. For example, using all OTUs (first row), we have five taxonomic levels and 20 normalization+zero replacement strategies each, so in total, we have 100 (20 normalization strategies times 5 taxonomic levels) different OTU datasets. For each of these 100 datasets, the number of predictors (i.e., OTUs) would depend on the taxonomic level being analyzed. For example, at the Phylum level, there might be 42 predictors, while at the Genus level, there could be up to 485 predictors. Random forest model The RF classifier is a powerful ML technique that has gained significant popularity in the last two decades because of its accuracy and speed. RF randomly creates an ensemble of decision trees. Each tree picks a random set of samples (bagging) from the data and models the samples independently from other trees. Instead of relying on a single learning model, RF builds a collection of decision models, and the final decision is based on the output of all the trees in the model. The bagging approach promotes the generation of uncorrelated trees which reduces the risk of overfitting. Each decision tree is generated individually without any pruning and each node is split using a user-defined number of features. By expanding the forest to a user-specified size, the technique generates trees with a high variance and low bias. The final classification choice is determined by summing the class-assignment probability obtained by each tree. A new unlabeled data in testing set input is thus compared to all decision trees formed in the ensemble, with each tree voting for class membership and the membership category with the most votes will be picked. The RF has several hyperparameters to be determined by the user, such as the number of decision trees to be generated, the number of variables to be selected and tested for the best split when growing the trees, the maximum depth of the tree, the minimum number of samples required to split an internal node, among others. Generally, a grid search is combined with K-fold cross validation to select the best hyperparameters . GridsearchCV is a well-known search method which is available in scikit-learn and it evaluates all possible parameter combinations to determine optimal values. Here, we set different values for parameters (see Table in the Supplementary Material) and tune them using GridsearchCV to find the optimal values for the RF classifier. GridSearchCV uses a “score” method for evaluating the performance of the cross-validated model on the test set. For evaluating our results, we employ the weighted F1 score due to the presence of unbalanced data. The weighted F1 score provides a comprehensive evaluation metric that considers both precision and recall across multiple classes, taking into account the class imbalance. It is calculated using this formula: [12pt]{minimal} $$ {weighted F1 scores} = ^{N} w_i {F1 score}_i}{ _{i=1}^{N} w_i}$$ weighted F1 scores = ∑ i = 1 N w i · F1 score i ∑ i = 1 N w i where [12pt]{minimal} $$N$$ N is the number of classes, here is 2 and [12pt]{minimal} $$w_i$$ w i is the weight assigned to class [12pt]{minimal} $$i$$ i which is determined based on the size of each class, and [12pt]{minimal} $$ {F1 score}_i$$ F1 score i is the F1 score for class [12pt]{minimal} $$i$$ i . A weighted F1 score of 1 indicates the best possible result, while a score of 0 indicates the worst possible result. Finally, when the parameters for the RF model are tuned, we use 20% of samples to report the performance of the final model. We use the weighted average F1 score to evaluate the performance of the models, which is computed by averaging all the per-class F1 scores while accounting for the number of samples in each class. Bayesian neural network model The size of the samples used in our study is too small for traditional Deep Learning approach. Bayesian NN are suitable for small sample sizes as they provide natural protection against overfitting by considering distributions for the model parameters. This is due to the fact that distributions are considered for the parameters in the model which allow us to marginalize them so that the prediction is based on data points alone . In the first paragraph of section “ ” of the Supplementary Material, we provide the mathematical details of Bayesian NN models. We structure our Bayesian NN models based on the datasets and we set the prior distributions and hyperparameters following the scheme described in . The detailed mathematical representation of the parameters and structure information of our model could be found in the second paragraph of section “ ” of the Supplementary Material. The training and the approximation of the posterior distribution are done via a Hamiltonian (Hybrid) Monte Carlo (HMC) implemented in . The choice of leap frog lengths and step sizes could be found in the last paragraph of section “ ” of the Supplementary Material. Given that HMC does not scale well for high dimensional parameter spaces and large datasets , we did not fit the Bayesian NN on all the models described in Table . In particular, we do not fit a Bayesian NN model on the Genus level for the ALL-OTUs as this model would involve over 9 million weights in the network with 485 input neurons. Full model selection FMS involves the process of listing all data preprocessing steps, model options and selection of predictors, and using a decision tree model to identify the choices that yield the highest measure of performance. Here, we fit a FMS strategy with the following options: 1) type of normalization, 2) type of zero replacement, 3) taxonomic level, and 4) data augmentation. We combine the type of normalization and type of zero replacement strategy into one variable (denoted [12pt]{minimal} $$ {NM}_i$$ NM i for [12pt]{minimal} $$i=1, ,20$$ i = 1 , ⋯ , 20 ). We focus on the weighted F1 score as measure of performance, and we include all OTU predictors (that is, we do not consider feature selection as one of the options to compare). We build the regression decision tree by using the DecisionTreeRegressor which is available in scikit-learn . We use the default parameters in the DecisionTreeRegressor such as “squared error” as the criterion to measure the quality of a split, a minimum number of 2 for the samples required to split an internal node, and a minimum number of 1 sample required to be at a leaf node. In order to create an informative decision tree that can be interpreted, we use a maximum depth of 4. Data description In this study, we focus on the soil microbiome (matrix of abundances) in a variety of taxonomic orders, including Phylum, Class, Order, Family, and Genus as well as other environmental information from soil samples acquired from potato fields in Wisconsin and Minnesota. The dataset consists of measurements related to soil health, potato yield and soil quality information. The soil health data were collected in the fall of 2019 from pre-planting commercial potato fields and include soil physicochemical properties, soil microbiome composition, soil microbiome diversity, and soil pathogen abundance. Soils were collected from 36 Minnesota fields and 66 Wisconsin fields, with three bulk soil samples randomly selected from each field. The potato yield and quality information at each sampling location was measured at the end of the growing season (September of 2020) including tuber yield and disease severity. Overall, we have 256 samples, 108 of which are taken from fields in Minnesota, and 148 of which are taken from fields in Wisconsin. While this provides a strong foundation for modeling the upper Midwest potato-growing region, we acknowledge that the use of data from a single growing season and geographic region limits the generalizability of our findings. Future data collection from multiple growing seasons and regions is underway to enhance the robustness and applicability of our models. We list all measurements in Table in the Supplementary Material. Soil physicochemical properties Fresh field soils were measured for a variety of physicochemical properties in the Agvise soil testing lab (Benson, Minnesota). Measurements of soil pH, organic matter content, carbon fractions, organic nitrogen, macro and micronutrients are described in . Soil texture was measured by quantifying the relative amount of sand, silt, and clay using a hydrometer. Soil cation/anion exchange capacity was calculated from the nutrient test results mentioned above, reported as milliequivalents per 100 gs of soil. Soil microbial community composition and population abundance Soil microbial community was characterized by high-throughput sequencing of the bacterial 16 S rRNA gene and fungal ITS2 regions. A subsample of 0.25 g of frozen field soils were extracted with the DNeasy PowerSoil Pro DNA isolation kit (Qiagen, CA). Extracted DNA was used in a two-step PCR reaction , with the V3-V4 region of bacterial 16 S rRNA and the eukaryotic ITS2 region amplified using the primer set V3F and 806R, and 5.8S and ITS4, respectively . The final PCR product was normalized, pooled and cleaned-up before sequenced on a Illumina MiSeq platform at the University of Minnesota Genomics Center. Sequences were analyzed using Qiime2. Cutadapt was first used to remove the forward and reverse primers of the ITS reads. Trimmed ITS reads and the raw 16 S reads were then truncated, filtered, denoised, pair-end merged, and chimeras removed using the DADA2 pipeline. Taxonomy was assigned to the feature table of amplicon sequence variant (ASV) using Qiime2’s feature-classifier plugin, using the RDP Naïve Bayesian Classifier fit to the SILVA 138 database for 16 S reads and UNITE database for ITS reads. Bacterial and fungal ASV tables were merged at Phylum, Class, Order, Family, and Genus levels using phyloseq in R. Alpha diversity measured as Chao1, Abundance-based coverage estimator, Shannon, Simpson, and Inverse Simpson index were calculated after rarefying the samples to the minimum sample depth. Alpha diversity was calculated for each taxonomic level using the vegan package . The population abundance of bacteria, fungi, Verticillium dahliae , and Pathogenic Streptomyces were measured with quantitative polymerase chain reaction as described in . Yield and disease Potatoes were harvested by hand from a one-meter hill (usually 3-4 plants) at each sampling location at the end of the growing season. One plant was used for tuber disease assessment, and the rest plants were used for yield estimation. Tubers were visually assessed for common scab, silver scurf, and black scurf, and then cut-open to evaluate Verticillium dahliae infection (dark vascular ring), and hollow heart. Tuber yield was estimated as the fresh weight of cleaned tubers. Binarization of response variables We have six phenotypes (response variables) of interest, four of them correspond to diseases and two of them correspond to yield (Table ). All six responses in the dataset are continuous, so we need to binarize them to fit the classification models. For the disease-related responses, we simply make the binary label 0 if there is no presence of disease, and 1 if there was detection of disease (that is, if the continuous response is greater than 0.0). Binarizing the yield response variables is harder as there is no universal standard to classify potato yield to be low or high. Furthermore, yield values are highly dependent on the type of potato variety. We assign a label of 0 (low yield) to samples with a yield less than the variety-specific median. Similarly, we assign a label of 1 (high yield) to samples with a yield greater than the variety-specific median. We illustrate this approach in Fig. in the Supplementary Material. After binarization, we note that pitted scab disease (denoted Scabpit in the figures), superficial scab disease (denoted Scabsuper in the figures), and both yield responses are balanced, whereas other responses are highly imbalanced: scab disease (denoted Scab in the figures) has 80% of samples labeled as 1, and black scurf disease (denoted Black_Scurf in the figures) has only six samples labeled as 1. We use these imbalanced cases to assess the performance of the methods under imbalance settings and data augmentation strategies. Data filtering, normalization, and zero replacement The input data is a matrix with non-negative read counts that were generated by a sequencing procedure. Let [12pt]{minimal} $$w^{(k)}= [w_1^{(k)},...,w_p^{(k)}]$$ w ( k ) = [ w 1 ( k ) , . . . , w p ( k ) ] be the total read counts of sample k containing p OTUs, where [12pt]{minimal} $$w^{(k)}$$ w ( k ) is a composition that adds up to a fixed value of [12pt]{minimal} $$m^{(k)}= _{i=1}^p w_i^{(k)}$$ m ( k ) = ∑ i = 1 p w i ( k ) . This value [12pt]{minimal} $$m^{(k)}$$ m ( k ) is the sequencing depth, which varies across samples and is predetermined by technical factors resulting in highly sparse data. It is reasonable to filter out a certain set of OTUs as the first data preparation step. For filtering, we only include OTUs that appear in at least 15 samples. Table displays the number of features (OTUs) before and after filtering for different taxonomic levels. As mentioned, the input data is compositional and highly sparse. It is known that ML methods do not perform well with unnormalized data and with sparse data . Therefore, we explore the effect of four zero replacement strategies (to overcome sparsity) and five normalization strategies (to overcome compositionality). All strategies are implemented in the NetCoMi R package . In particular, we consider the four zero replacement strategies: (1) the original dataset which included zeros (denoted none in the figures), (2) pseudo-zero replacement which replaces zero counts by a predefined pseudo count (denoted pseudo in the figures), (3) multiplicative zero replacement which imputes left-censored compositional values by a given fraction and applies a multiplicative adjustment to preserve the multivariate compositional properties of the samples (denoted multRepl in the figures) , and (4) Bayesian-multiplicative treatment which imputes zero counts by posterior estimates of the multinomial probabilities generating the counts, assuming a Dirichlet prior distribution (denoted bayesMult in the figures) . Next, we use five normalization methods: (1) Total sum scaling which simply converts counts to appropriately scaled ratios (denoted TSS in the figures) , (2) Cumulative sum scaling which rescales the samples based on a subset (quartile) of lower abundant taxa, thereby excluding the impact of highly abundant taxa (denoted CSS in the figures) , (3) Common sum scaling in which counts are scaled to the minimum depth of each sample (denoted COM in the figures) , (4) Rarefying which random samples without replacement after a minimum count threshold has been applied (denoted rarefy in the figures) , and (5) Centered Log-ratio which transforms the data using the geometric mean as the reference (denoted clr in the figures) . With four zero replacement methods and five normalization methods, we create 20 datasets by the combination of zero replacement and normalization techniques. Our goal is to study the effect of the zero replacement and normalization choice in the performance of the deep learning methods. Namely, we have the following 20 combinations, [12pt]{minimal} $$ {NM}_1$$ NM 1 : TSS+none, [12pt]{minimal} $$ {NM}_2$$ NM 2 : TSS+pseudo, [12pt]{minimal} $$ {NM}_3$$ NM 3 : TSS+multRepl, [12pt]{minimal} $$ {NM}_4$$ NM 4 : TSS+bayesMult, [12pt]{minimal} $$ {NM}_5$$ NM 5 : CSS+none, [12pt]{minimal} $$ {NM}_6$$ NM 6 : CSS+pseudo, [12pt]{minimal} $$ {NM}_7$$ NM 7 : CSS+multRepl, [12pt]{minimal} $$ {NM}_8$$ NM 8 : CSS+bayesMult, [12pt]{minimal} $$ {NM}_9$$ NM 9 : COM+none, [12pt]{minimal} $$ {NM}_{10}$$ NM 10 : COM+pseudo, [12pt]{minimal} $$ {NM}_{11}$$ NM 11 : COM+multRepl, [12pt]{minimal} $$ {NM}_{12}$$ NM 12 : COM+bayesMult, [12pt]{minimal} $$ {NM}_{13}$$ NM 13 : rarefy+none, [12pt]{minimal} $$ {NM}_{14}$$ NM 14 : rarefy+pseudo, [12pt]{minimal} $$ {NM}_{15}$$ NM 15 : rarefy+multRepl, [12pt]{minimal} $$ {NM}_{16}$$ NM 16 : rarefy+bayesMult, [12pt]{minimal} $$ {NM}_{17}$$ NM 17 : clr+none, [12pt]{minimal} $$ {NM}_{18}$$ NM 18 : clr+pseudo,, [12pt]{minimal} $$ {NM}_{19}$$ NM 19 : clr+multRepl, and [12pt]{minimal} $$ {NM}_{20}$$ NM 20 : clr+bayesMult. For convenience, we use the notation [12pt]{minimal} $$ {NM}_i$$ NM i (Normalization Method) for [12pt]{minimal} $$i=1, ,20$$ i = 1 , ⋯ , 20 in the Full Model Selection section (See section “ ”). For the environmental predictors of soil chemistry and microbial population density in the soil, we apply six scaling methods: (1) standardize features by subtracting the mean and scaling to unit variance ; (2) scale each feature to a [0, 1] range; (3) scale each feature by its maximum absolute value; (4) scale features by subtracting the median and scaling to the interquartile range ; (5) transform the features to follow a normal distribution ; (6) normalize samples individually to the unit norm. After normalization, the datasets are split into training, validation, and testing sets with 10-fold cross-validation. We used 80% of samples for training and validation, and 20% for testing. Data augmentation There are three main goals that we wish to achieve with data augmentation: (1) improve the model’s prediction performance with more artificial samples; (2) balance the number of labels with artificial samples, and (3) make the model more robust and avoid overfitting with unseen (artificial) data. We note that augmenting the whole dataset and then splitting it into training and testing sets would result in data leakage . For example, when the original sample is in the testing set and the augmented sample from this sample is in the training set, the model is essentially training and testing on the same sample since the normalized values of OTUs are very close. Thus, we split the data into training and testing sets first and only augment the training set. This strategy also allows us to have a fair performance comparison for augmented and non-augmented sets with the same testing data. Regarding the data augmentation procedure, instead of simply adding a randomly generated noise to the original sample, we subset the data by variety and label, compute the mean (and standard deviation) abundance value for this subset, and create a new sample that includes the original data plus a Gaussian error with mean [12pt]{minimal} $$ /100$$ μ / 100 and standard deviation [12pt]{minimal} $$ /100$$ σ / 100 where [12pt]{minimal} $$ , $$ μ , σ are the subset-specific mean and standard deviation, respectively. This approach is illustrated in Fig. in the Supplementary Material. By the end of this procedure we would have a balanced augmented training set with 400 samples per label for each of the five taxonomic levels (Phylum, Class, Order, Family, and Genus), the number of samples is shown in Table . In this study, we focus on the soil microbiome (matrix of abundances) in a variety of taxonomic orders, including Phylum, Class, Order, Family, and Genus as well as other environmental information from soil samples acquired from potato fields in Wisconsin and Minnesota. The dataset consists of measurements related to soil health, potato yield and soil quality information. The soil health data were collected in the fall of 2019 from pre-planting commercial potato fields and include soil physicochemical properties, soil microbiome composition, soil microbiome diversity, and soil pathogen abundance. Soils were collected from 36 Minnesota fields and 66 Wisconsin fields, with three bulk soil samples randomly selected from each field. The potato yield and quality information at each sampling location was measured at the end of the growing season (September of 2020) including tuber yield and disease severity. Overall, we have 256 samples, 108 of which are taken from fields in Minnesota, and 148 of which are taken from fields in Wisconsin. While this provides a strong foundation for modeling the upper Midwest potato-growing region, we acknowledge that the use of data from a single growing season and geographic region limits the generalizability of our findings. Future data collection from multiple growing seasons and regions is underway to enhance the robustness and applicability of our models. We list all measurements in Table in the Supplementary Material. Soil physicochemical properties Fresh field soils were measured for a variety of physicochemical properties in the Agvise soil testing lab (Benson, Minnesota). Measurements of soil pH, organic matter content, carbon fractions, organic nitrogen, macro and micronutrients are described in . Soil texture was measured by quantifying the relative amount of sand, silt, and clay using a hydrometer. Soil cation/anion exchange capacity was calculated from the nutrient test results mentioned above, reported as milliequivalents per 100 gs of soil. Soil microbial community composition and population abundance Soil microbial community was characterized by high-throughput sequencing of the bacterial 16 S rRNA gene and fungal ITS2 regions. A subsample of 0.25 g of frozen field soils were extracted with the DNeasy PowerSoil Pro DNA isolation kit (Qiagen, CA). Extracted DNA was used in a two-step PCR reaction , with the V3-V4 region of bacterial 16 S rRNA and the eukaryotic ITS2 region amplified using the primer set V3F and 806R, and 5.8S and ITS4, respectively . The final PCR product was normalized, pooled and cleaned-up before sequenced on a Illumina MiSeq platform at the University of Minnesota Genomics Center. Sequences were analyzed using Qiime2. Cutadapt was first used to remove the forward and reverse primers of the ITS reads. Trimmed ITS reads and the raw 16 S reads were then truncated, filtered, denoised, pair-end merged, and chimeras removed using the DADA2 pipeline. Taxonomy was assigned to the feature table of amplicon sequence variant (ASV) using Qiime2’s feature-classifier plugin, using the RDP Naïve Bayesian Classifier fit to the SILVA 138 database for 16 S reads and UNITE database for ITS reads. Bacterial and fungal ASV tables were merged at Phylum, Class, Order, Family, and Genus levels using phyloseq in R. Alpha diversity measured as Chao1, Abundance-based coverage estimator, Shannon, Simpson, and Inverse Simpson index were calculated after rarefying the samples to the minimum sample depth. Alpha diversity was calculated for each taxonomic level using the vegan package . The population abundance of bacteria, fungi, Verticillium dahliae , and Pathogenic Streptomyces were measured with quantitative polymerase chain reaction as described in . Yield and disease Potatoes were harvested by hand from a one-meter hill (usually 3-4 plants) at each sampling location at the end of the growing season. One plant was used for tuber disease assessment, and the rest plants were used for yield estimation. Tubers were visually assessed for common scab, silver scurf, and black scurf, and then cut-open to evaluate Verticillium dahliae infection (dark vascular ring), and hollow heart. Tuber yield was estimated as the fresh weight of cleaned tubers. We have six phenotypes (response variables) of interest, four of them correspond to diseases and two of them correspond to yield (Table ). All six responses in the dataset are continuous, so we need to binarize them to fit the classification models. For the disease-related responses, we simply make the binary label 0 if there is no presence of disease, and 1 if there was detection of disease (that is, if the continuous response is greater than 0.0). Binarizing the yield response variables is harder as there is no universal standard to classify potato yield to be low or high. Furthermore, yield values are highly dependent on the type of potato variety. We assign a label of 0 (low yield) to samples with a yield less than the variety-specific median. Similarly, we assign a label of 1 (high yield) to samples with a yield greater than the variety-specific median. We illustrate this approach in Fig. in the Supplementary Material. After binarization, we note that pitted scab disease (denoted Scabpit in the figures), superficial scab disease (denoted Scabsuper in the figures), and both yield responses are balanced, whereas other responses are highly imbalanced: scab disease (denoted Scab in the figures) has 80% of samples labeled as 1, and black scurf disease (denoted Black_Scurf in the figures) has only six samples labeled as 1. We use these imbalanced cases to assess the performance of the methods under imbalance settings and data augmentation strategies. The input data is a matrix with non-negative read counts that were generated by a sequencing procedure. Let [12pt]{minimal} $$w^{(k)}= [w_1^{(k)},...,w_p^{(k)}]$$ w ( k ) = [ w 1 ( k ) , . . . , w p ( k ) ] be the total read counts of sample k containing p OTUs, where [12pt]{minimal} $$w^{(k)}$$ w ( k ) is a composition that adds up to a fixed value of [12pt]{minimal} $$m^{(k)}= _{i=1}^p w_i^{(k)}$$ m ( k ) = ∑ i = 1 p w i ( k ) . This value [12pt]{minimal} $$m^{(k)}$$ m ( k ) is the sequencing depth, which varies across samples and is predetermined by technical factors resulting in highly sparse data. It is reasonable to filter out a certain set of OTUs as the first data preparation step. For filtering, we only include OTUs that appear in at least 15 samples. Table displays the number of features (OTUs) before and after filtering for different taxonomic levels. As mentioned, the input data is compositional and highly sparse. It is known that ML methods do not perform well with unnormalized data and with sparse data . Therefore, we explore the effect of four zero replacement strategies (to overcome sparsity) and five normalization strategies (to overcome compositionality). All strategies are implemented in the NetCoMi R package . In particular, we consider the four zero replacement strategies: (1) the original dataset which included zeros (denoted none in the figures), (2) pseudo-zero replacement which replaces zero counts by a predefined pseudo count (denoted pseudo in the figures), (3) multiplicative zero replacement which imputes left-censored compositional values by a given fraction and applies a multiplicative adjustment to preserve the multivariate compositional properties of the samples (denoted multRepl in the figures) , and (4) Bayesian-multiplicative treatment which imputes zero counts by posterior estimates of the multinomial probabilities generating the counts, assuming a Dirichlet prior distribution (denoted bayesMult in the figures) . Next, we use five normalization methods: (1) Total sum scaling which simply converts counts to appropriately scaled ratios (denoted TSS in the figures) , (2) Cumulative sum scaling which rescales the samples based on a subset (quartile) of lower abundant taxa, thereby excluding the impact of highly abundant taxa (denoted CSS in the figures) , (3) Common sum scaling in which counts are scaled to the minimum depth of each sample (denoted COM in the figures) , (4) Rarefying which random samples without replacement after a minimum count threshold has been applied (denoted rarefy in the figures) , and (5) Centered Log-ratio which transforms the data using the geometric mean as the reference (denoted clr in the figures) . With four zero replacement methods and five normalization methods, we create 20 datasets by the combination of zero replacement and normalization techniques. Our goal is to study the effect of the zero replacement and normalization choice in the performance of the deep learning methods. Namely, we have the following 20 combinations, [12pt]{minimal} $$ {NM}_1$$ NM 1 : TSS+none, [12pt]{minimal} $$ {NM}_2$$ NM 2 : TSS+pseudo, [12pt]{minimal} $$ {NM}_3$$ NM 3 : TSS+multRepl, [12pt]{minimal} $$ {NM}_4$$ NM 4 : TSS+bayesMult, [12pt]{minimal} $$ {NM}_5$$ NM 5 : CSS+none, [12pt]{minimal} $$ {NM}_6$$ NM 6 : CSS+pseudo, [12pt]{minimal} $$ {NM}_7$$ NM 7 : CSS+multRepl, [12pt]{minimal} $$ {NM}_8$$ NM 8 : CSS+bayesMult, [12pt]{minimal} $$ {NM}_9$$ NM 9 : COM+none, [12pt]{minimal} $$ {NM}_{10}$$ NM 10 : COM+pseudo, [12pt]{minimal} $$ {NM}_{11}$$ NM 11 : COM+multRepl, [12pt]{minimal} $$ {NM}_{12}$$ NM 12 : COM+bayesMult, [12pt]{minimal} $$ {NM}_{13}$$ NM 13 : rarefy+none, [12pt]{minimal} $$ {NM}_{14}$$ NM 14 : rarefy+pseudo, [12pt]{minimal} $$ {NM}_{15}$$ NM 15 : rarefy+multRepl, [12pt]{minimal} $$ {NM}_{16}$$ NM 16 : rarefy+bayesMult, [12pt]{minimal} $$ {NM}_{17}$$ NM 17 : clr+none, [12pt]{minimal} $$ {NM}_{18}$$ NM 18 : clr+pseudo,, [12pt]{minimal} $$ {NM}_{19}$$ NM 19 : clr+multRepl, and [12pt]{minimal} $$ {NM}_{20}$$ NM 20 : clr+bayesMult. For convenience, we use the notation [12pt]{minimal} $$ {NM}_i$$ NM i (Normalization Method) for [12pt]{minimal} $$i=1, ,20$$ i = 1 , ⋯ , 20 in the Full Model Selection section (See section “ ”). For the environmental predictors of soil chemistry and microbial population density in the soil, we apply six scaling methods: (1) standardize features by subtracting the mean and scaling to unit variance ; (2) scale each feature to a [0, 1] range; (3) scale each feature by its maximum absolute value; (4) scale features by subtracting the median and scaling to the interquartile range ; (5) transform the features to follow a normal distribution ; (6) normalize samples individually to the unit norm. After normalization, the datasets are split into training, validation, and testing sets with 10-fold cross-validation. We used 80% of samples for training and validation, and 20% for testing. There are three main goals that we wish to achieve with data augmentation: (1) improve the model’s prediction performance with more artificial samples; (2) balance the number of labels with artificial samples, and (3) make the model more robust and avoid overfitting with unseen (artificial) data. We note that augmenting the whole dataset and then splitting it into training and testing sets would result in data leakage . For example, when the original sample is in the testing set and the augmented sample from this sample is in the training set, the model is essentially training and testing on the same sample since the normalized values of OTUs are very close. Thus, we split the data into training and testing sets first and only augment the training set. This strategy also allows us to have a fair performance comparison for augmented and non-augmented sets with the same testing data. Regarding the data augmentation procedure, instead of simply adding a randomly generated noise to the original sample, we subset the data by variety and label, compute the mean (and standard deviation) abundance value for this subset, and create a new sample that includes the original data plus a Gaussian error with mean [12pt]{minimal} $$ /100$$ μ / 100 and standard deviation [12pt]{minimal} $$ /100$$ σ / 100 where [12pt]{minimal} $$ , $$ μ , σ are the subset-specific mean and standard deviation, respectively. This approach is illustrated in Fig. in the Supplementary Material. By the end of this procedure we would have a balanced augmented training set with 400 samples per label for each of the five taxonomic levels (Phylum, Class, Order, Family, and Genus), the number of samples is shown in Table . Feature selection involves the identification of important features (or covariates) that have high predictive power. Given the high-dimensionality of the data (e.g. 256 original samples for 485 OTUs at the Genus level), feature selection is necessary, especially for Bayesian NN models that are computationally expensive. We pursue two approaches for feature selection: (1) using ML models to assess variable importance, and (2) using network analyses. To focus exclusively on the effect of feature selection, we only consider one type of normalization and zero replacement strategy in this investigation, namely, total sum scaling normalization without zero replacement ( [12pt]{minimal} $$ {NM}_1$$ NM 1 : TSS+none). Using ML models for feature selection To identify important OTUs, we use six ML strategies implemented in scikit-learn : (1) “SelectKBest” method selects features based on the k highest analysis of variance F-value scores, (2) select the top k features based on the mutual information statistic, (3) recursive feature elimination (RFE) with logistic regression, (4) RFE with decision tree, (5) RFE with gradient boosting, and (6) RFE with RF. In addition to the six ML strategies, we consider a 7th strategy which consists in including OTUs in the model if their maximum value is within the top 30%. After running all seven strategies, we assign a value (“TOTAL”) to each OTU based on the number of times the OTU is selected as an important feature under the seven criteria. That is, an OTU that is selected as important by all seven strategies will have a value of 7. The OTUs are sorted based on “TOTAL” column and the top 30% of them are selected as important features. Thus, 30, 36, 75, 85, and 162 OTUs are selected for Phylum, Class, Order, Family, and Genus levels, respectively. For example, Table in the Supplementary Material shows the top 30 OTUs at the Phylum level and by which strategies they are identified as important features to predict the pitted scab response. Using network comparison for feature selection Next, we identify important OTUs by comparing their interactions in microbial networks when the network is constructed with samples from one class (say, low yield) versus with samples from the other class (say, high yield) . Indeed, evidence shows that microbial interactions can help differentiate between crop disease states . We identify the OTUs that interact differently on the samples corresponding to one class versus another class as relevant OTUs that contain information about the outcome. We use the SPRING method (semi-parametric rank-based approach for inference in the graphical model) in the NetComi package to construct two microbial networks: one corresponding to label 0 and one corresponding to label 1. To build these networks, the estimated partial correlations are transformed into dissimilarities via the signed distance metric, and the corresponding similarities are then used as edge weights. We then compare these networks in order to identify OTUs that have the greatest difference based on degree values. Figure shows an example of two microbial networks with different graphical structures, and Table in the Supplementary Material lists OTUs ranked by the difference in degree in the two microbial networks for diseased and non-diseased classes of pitted scab response. The OTUs are sorted based on the values of the degree difference column and similar to ML strategy top 30% of them are selected as important features. Combination of ML and network strategies in feature selection We define a scoring value for each OTU based on whether they are identified as important by ML strategy ( [12pt]{minimal} $$score=1$$ s c o r e = 1 ), by the network comparison ( [12pt]{minimal} $$score=2$$ s c o r e = 2 ), or both ( [12pt]{minimal} $$score=3$$ s c o r e = 3 ). If the OTU is not identified as important by any strategy, it is denoted [12pt]{minimal} $$score=0$$ s c o r e = 0 . Figure displays the scoring values of all OTUs at the Phylum level for all responses. To identify important OTUs, we use six ML strategies implemented in scikit-learn : (1) “SelectKBest” method selects features based on the k highest analysis of variance F-value scores, (2) select the top k features based on the mutual information statistic, (3) recursive feature elimination (RFE) with logistic regression, (4) RFE with decision tree, (5) RFE with gradient boosting, and (6) RFE with RF. In addition to the six ML strategies, we consider a 7th strategy which consists in including OTUs in the model if their maximum value is within the top 30%. After running all seven strategies, we assign a value (“TOTAL”) to each OTU based on the number of times the OTU is selected as an important feature under the seven criteria. That is, an OTU that is selected as important by all seven strategies will have a value of 7. The OTUs are sorted based on “TOTAL” column and the top 30% of them are selected as important features. Thus, 30, 36, 75, 85, and 162 OTUs are selected for Phylum, Class, Order, Family, and Genus levels, respectively. For example, Table in the Supplementary Material shows the top 30 OTUs at the Phylum level and by which strategies they are identified as important features to predict the pitted scab response. Next, we identify important OTUs by comparing their interactions in microbial networks when the network is constructed with samples from one class (say, low yield) versus with samples from the other class (say, high yield) . Indeed, evidence shows that microbial interactions can help differentiate between crop disease states . We identify the OTUs that interact differently on the samples corresponding to one class versus another class as relevant OTUs that contain information about the outcome. We use the SPRING method (semi-parametric rank-based approach for inference in the graphical model) in the NetComi package to construct two microbial networks: one corresponding to label 0 and one corresponding to label 1. To build these networks, the estimated partial correlations are transformed into dissimilarities via the signed distance metric, and the corresponding similarities are then used as edge weights. We then compare these networks in order to identify OTUs that have the greatest difference based on degree values. Figure shows an example of two microbial networks with different graphical structures, and Table in the Supplementary Material lists OTUs ranked by the difference in degree in the two microbial networks for diseased and non-diseased classes of pitted scab response. The OTUs are sorted based on the values of the degree difference column and similar to ML strategy top 30% of them are selected as important features. We define a scoring value for each OTU based on whether they are identified as important by ML strategy ( [12pt]{minimal} $$score=1$$ s c o r e = 1 ), by the network comparison ( [12pt]{minimal} $$score=2$$ s c o r e = 2 ), or both ( [12pt]{minimal} $$score=3$$ s c o r e = 3 ). If the OTU is not identified as important by any strategy, it is denoted [12pt]{minimal} $$score=0$$ s c o r e = 0 . Figure displays the scoring values of all OTUs at the Phylum level for all responses. We apply two types of classification models to predict potato disease and yield: RF and Bayesian NN. We consider separately three types of predictors: OTU abundances (5 taxonomic levels and 20 normalization/zero replacement strategies described in section “ ”), environmental predictors such as soil characteristics and microbial population density, and a combination of both types. Table lists all the models we consider in this study. Table shows the number of predictors included in each model as well as the data choices related to taxonomic level or normalization and zero replacement strategies. For example, using all OTUs (first row), we have five taxonomic levels and 20 normalization+zero replacement strategies each, so in total, we have 100 (20 normalization strategies times 5 taxonomic levels) different OTU datasets. For each of these 100 datasets, the number of predictors (i.e., OTUs) would depend on the taxonomic level being analyzed. For example, at the Phylum level, there might be 42 predictors, while at the Genus level, there could be up to 485 predictors. Random forest model The RF classifier is a powerful ML technique that has gained significant popularity in the last two decades because of its accuracy and speed. RF randomly creates an ensemble of decision trees. Each tree picks a random set of samples (bagging) from the data and models the samples independently from other trees. Instead of relying on a single learning model, RF builds a collection of decision models, and the final decision is based on the output of all the trees in the model. The bagging approach promotes the generation of uncorrelated trees which reduces the risk of overfitting. Each decision tree is generated individually without any pruning and each node is split using a user-defined number of features. By expanding the forest to a user-specified size, the technique generates trees with a high variance and low bias. The final classification choice is determined by summing the class-assignment probability obtained by each tree. A new unlabeled data in testing set input is thus compared to all decision trees formed in the ensemble, with each tree voting for class membership and the membership category with the most votes will be picked. The RF has several hyperparameters to be determined by the user, such as the number of decision trees to be generated, the number of variables to be selected and tested for the best split when growing the trees, the maximum depth of the tree, the minimum number of samples required to split an internal node, among others. Generally, a grid search is combined with K-fold cross validation to select the best hyperparameters . GridsearchCV is a well-known search method which is available in scikit-learn and it evaluates all possible parameter combinations to determine optimal values. Here, we set different values for parameters (see Table in the Supplementary Material) and tune them using GridsearchCV to find the optimal values for the RF classifier. GridSearchCV uses a “score” method for evaluating the performance of the cross-validated model on the test set. For evaluating our results, we employ the weighted F1 score due to the presence of unbalanced data. The weighted F1 score provides a comprehensive evaluation metric that considers both precision and recall across multiple classes, taking into account the class imbalance. It is calculated using this formula: [12pt]{minimal} $$ {weighted F1 scores} = ^{N} w_i {F1 score}_i}{ _{i=1}^{N} w_i}$$ weighted F1 scores = ∑ i = 1 N w i · F1 score i ∑ i = 1 N w i where [12pt]{minimal} $$N$$ N is the number of classes, here is 2 and [12pt]{minimal} $$w_i$$ w i is the weight assigned to class [12pt]{minimal} $$i$$ i which is determined based on the size of each class, and [12pt]{minimal} $$ {F1 score}_i$$ F1 score i is the F1 score for class [12pt]{minimal} $$i$$ i . A weighted F1 score of 1 indicates the best possible result, while a score of 0 indicates the worst possible result. Finally, when the parameters for the RF model are tuned, we use 20% of samples to report the performance of the final model. We use the weighted average F1 score to evaluate the performance of the models, which is computed by averaging all the per-class F1 scores while accounting for the number of samples in each class. Bayesian neural network model The size of the samples used in our study is too small for traditional Deep Learning approach. Bayesian NN are suitable for small sample sizes as they provide natural protection against overfitting by considering distributions for the model parameters. This is due to the fact that distributions are considered for the parameters in the model which allow us to marginalize them so that the prediction is based on data points alone . In the first paragraph of section “ ” of the Supplementary Material, we provide the mathematical details of Bayesian NN models. We structure our Bayesian NN models based on the datasets and we set the prior distributions and hyperparameters following the scheme described in . The detailed mathematical representation of the parameters and structure information of our model could be found in the second paragraph of section “ ” of the Supplementary Material. The training and the approximation of the posterior distribution are done via a Hamiltonian (Hybrid) Monte Carlo (HMC) implemented in . The choice of leap frog lengths and step sizes could be found in the last paragraph of section “ ” of the Supplementary Material. Given that HMC does not scale well for high dimensional parameter spaces and large datasets , we did not fit the Bayesian NN on all the models described in Table . In particular, we do not fit a Bayesian NN model on the Genus level for the ALL-OTUs as this model would involve over 9 million weights in the network with 485 input neurons. The RF classifier is a powerful ML technique that has gained significant popularity in the last two decades because of its accuracy and speed. RF randomly creates an ensemble of decision trees. Each tree picks a random set of samples (bagging) from the data and models the samples independently from other trees. Instead of relying on a single learning model, RF builds a collection of decision models, and the final decision is based on the output of all the trees in the model. The bagging approach promotes the generation of uncorrelated trees which reduces the risk of overfitting. Each decision tree is generated individually without any pruning and each node is split using a user-defined number of features. By expanding the forest to a user-specified size, the technique generates trees with a high variance and low bias. The final classification choice is determined by summing the class-assignment probability obtained by each tree. A new unlabeled data in testing set input is thus compared to all decision trees formed in the ensemble, with each tree voting for class membership and the membership category with the most votes will be picked. The RF has several hyperparameters to be determined by the user, such as the number of decision trees to be generated, the number of variables to be selected and tested for the best split when growing the trees, the maximum depth of the tree, the minimum number of samples required to split an internal node, among others. Generally, a grid search is combined with K-fold cross validation to select the best hyperparameters . GridsearchCV is a well-known search method which is available in scikit-learn and it evaluates all possible parameter combinations to determine optimal values. Here, we set different values for parameters (see Table in the Supplementary Material) and tune them using GridsearchCV to find the optimal values for the RF classifier. GridSearchCV uses a “score” method for evaluating the performance of the cross-validated model on the test set. For evaluating our results, we employ the weighted F1 score due to the presence of unbalanced data. The weighted F1 score provides a comprehensive evaluation metric that considers both precision and recall across multiple classes, taking into account the class imbalance. It is calculated using this formula: [12pt]{minimal} $$ {weighted F1 scores} = ^{N} w_i {F1 score}_i}{ _{i=1}^{N} w_i}$$ weighted F1 scores = ∑ i = 1 N w i · F1 score i ∑ i = 1 N w i where [12pt]{minimal} $$N$$ N is the number of classes, here is 2 and [12pt]{minimal} $$w_i$$ w i is the weight assigned to class [12pt]{minimal} $$i$$ i which is determined based on the size of each class, and [12pt]{minimal} $$ {F1 score}_i$$ F1 score i is the F1 score for class [12pt]{minimal} $$i$$ i . A weighted F1 score of 1 indicates the best possible result, while a score of 0 indicates the worst possible result. Finally, when the parameters for the RF model are tuned, we use 20% of samples to report the performance of the final model. We use the weighted average F1 score to evaluate the performance of the models, which is computed by averaging all the per-class F1 scores while accounting for the number of samples in each class. The size of the samples used in our study is too small for traditional Deep Learning approach. Bayesian NN are suitable for small sample sizes as they provide natural protection against overfitting by considering distributions for the model parameters. This is due to the fact that distributions are considered for the parameters in the model which allow us to marginalize them so that the prediction is based on data points alone . In the first paragraph of section “ ” of the Supplementary Material, we provide the mathematical details of Bayesian NN models. We structure our Bayesian NN models based on the datasets and we set the prior distributions and hyperparameters following the scheme described in . The detailed mathematical representation of the parameters and structure information of our model could be found in the second paragraph of section “ ” of the Supplementary Material. The training and the approximation of the posterior distribution are done via a Hamiltonian (Hybrid) Monte Carlo (HMC) implemented in . The choice of leap frog lengths and step sizes could be found in the last paragraph of section “ ” of the Supplementary Material. Given that HMC does not scale well for high dimensional parameter spaces and large datasets , we did not fit the Bayesian NN on all the models described in Table . In particular, we do not fit a Bayesian NN model on the Genus level for the ALL-OTUs as this model would involve over 9 million weights in the network with 485 input neurons. FMS involves the process of listing all data preprocessing steps, model options and selection of predictors, and using a decision tree model to identify the choices that yield the highest measure of performance. Here, we fit a FMS strategy with the following options: 1) type of normalization, 2) type of zero replacement, 3) taxonomic level, and 4) data augmentation. We combine the type of normalization and type of zero replacement strategy into one variable (denoted [12pt]{minimal} $$ {NM}_i$$ NM i for [12pt]{minimal} $$i=1, ,20$$ i = 1 , ⋯ , 20 ). We focus on the weighted F1 score as measure of performance, and we include all OTU predictors (that is, we do not consider feature selection as one of the options to compare). We build the regression decision tree by using the DecisionTreeRegressor which is available in scikit-learn . We use the default parameters in the DecisionTreeRegressor such as “squared error” as the criterion to measure the quality of a split, a minimum number of 2 for the samples required to split an internal node, and a minimum number of 1 sample required to be at a leaf node. In order to create an informative decision tree that can be interpreted, we use a maximum depth of 4. Performance evaluation of predictive models for yield responses We implemented the H 2 O AutoML package in Python , an open-source package designed for automated ML, which trains multiple models such as RF, Gradient Boosting Machines, and Deep Learning models. H 2 O AutoML automates model selection and hyperparameter tuning, providing a comprehensive comparison of different ML methods. The best-performing model is selected based on Root Mean Square Error (RMSE), a standard metric for regression tasks. This process allows for a more robust evaluation of model performance in predicting continuous yield, avoiding the biases introduced by arbitrary binarization. However, to maintain consistency with the rest of the study, we also report results from RF models. We report Mean Absolute Percentage Error (MAPE) as the evaluation metric. MAPE calculates the average absolute difference between predicted and actual values as a percentage of the actual values. The formula for MAPE is defined as: [12pt]{minimal} $$ {MAPE} = _{i=1}^{n} | | 100 $$ MAPE = 1 n ∑ i = 1 n A i - F i A i × 100 where [12pt]{minimal} $$A_i$$ A i is the actual value, [12pt]{minimal} $$F_i$$ F i is the predicted value, and n is the number of observations. MAPE, evaluated on a scale from 0 to 1, is commonly used for regression tasks because it is particularly valuable as it is scale-independent, making it easy to compare performance across different models and datasets . A lower MAPE indicates better model accuracy, with 0% representing perfect prediction. A MAPE of, for example, 10% means that, on average, the model’s predictions are 10% off from the actual values. The results of MAPE values are visualized using box plots, which clearly represent the variability and performance of the models across different normalization methods and taxonomic levels in Fig. . The left part of the figure displays the results from RF models. For a more comprehensive analysis, we also show the best result obtained from the H 2 O AutoML model as shown in the right part of Fig. . We acknowledge that binarization does lead to information loss. The continuous modeling results demonstrate that predicting Yield_Plant is particularly challenging, likely due to the biological variability between individual plants. However, we observed better performance for Yield_Meter, which is a more stable measure due to its aggregation over a larger area. To balance accuracy and complexity in this study, we applied binarization to the continuous response in cases where it improved performance, although we recognize that this is not an ideal long-term solution. In future work, we plan to explore larger datasets and incorporate additional environmental variables to enhance model accuracy without relying on binarization. Comparing the performance of random forest and H 2 O autoML for disease and binarized yield prediction Figure presents the weighted F1 scores for RF (left panel) and the best H 2 O AutoML models (right panel) across various responses, including both yield and disease outcomes (binary responses). Each boxplot represents different taxonomic levels (Phylum, Class, Order, Family, Genus) to evaluate model performance. The results show that RF performs comparable to the best H 2 O AutoML models, particularly excelling in Scabpit response. This demonstrates RF’s reliability, as its performance is consistently close to or equal to the more complex models selected by AutoML. We also choose Deep Learning Models as they are known to excel in exploring deep relationships between predictors. We believe it is essential to include a Deep Learning model as this is the cutting-edge method that resulted in most success in ML applications in the last decade. However, while deep learning approaches were considered, we determined that our dataset, with approximately 200 samples, is too small to effectively apply deep learning models. Deep learning typically requires larger datasets to avoid overfitting and produce generalizable results, and thus, it was not a viable option for our study. While there are many computational efficient models, none would be properly trained with 200 samples without underfitting or overfitting. Despite the computational challenges, Bayesian NNs are known to be informative for small sample sizes, offering protection against overfitting by modeling parameter uncertainty. Even though the Markov Chain Monte Carlo (MCMC) process for parameter estimation is very computationally inefficient in the context of Deep Learning, more research has gone into alternate estimation methods such as Variational Inference methods in the past few years which might drastically improve the computational efficiency in the near future. The Deep Learning result also validates the result of the RF model, a completely different approach that is more computationally efficient. Therefore, we will retain the Bayesian NN results in the paper alongside the other models, as they provide valuable insights despite the computational demands. Regarding black scurf disease, we found that it is a very imbalanced dataset, with only 6 samples exhibiting the disease. Due to the extremely limited number of cases, the results from machine learning methods cannot be considered reliable. Consequently, we decided not to focus on this disease in our current analysis. Instead, we plan to use data augmentation methods to improve predictions for black scurf. Additionally, we are collecting more datasets, after which we will apply ML methods to the original data to obtain more reliable results. Although the H 2 O AutoML framework identifies slightly better-performing models in certain cases, RF maintains a strong balance between predictive performance and computational efficiency. In yield-related predictions like Yield_Plant and Yield_Meter, RF produces F1 scores that are very close to those of the best AutoML models, making it a dependable choice. Given the practical constraints of running models across 600 configurations (5 taxonomic levels [12pt]{minimal} $$$$ × 20 normalization methods [12pt]{minimal} $$$$ × 6 responses) and the inclusion of environmental predictors, RF’s lower computational demand provides an efficient solution without compromising accuracy. This comparison supports our choice of RF as a reliable and efficient model, reinforcing the robustness of our study’s findings. Furthermore, we aimed to use one robust model that works effectively across all response types, and RF consistently met this criterion, demonstrating reliable performance across various conditions. Given these results, RF was selected as the main model due to its interpretability, robustness, and lower computational demands. While H 2 O AutoML occasionally yielded slightly higher scores, RF’s performance remained consistently close, underscoring its suitability as the primary model for microbiome studies with limited computational resources. Overall performance of predictive models: manual binarization causes inaccurate prediction of yield First, we identify the outcomes (disease or binarized yield) that are accurately predicted across models (and are thus robust to prediction regardless of model choices), as well as the models that accurately predict across outcomes (and are thus the most powerful model alternatives). To do so, we aggregate the weighted F1 scores on data preprocessing choices such as normalization, zero replacement, and taxonomic levels for every model and every outcome. We employed RF and Bayesian NN models across various predictive scenarios (14 different models). The predictive capabilities of these models are illustrated in Figs. (RF) and (Bayesian NN), providing a comprehensive analysis of all 14 models outlined in Table . This detailed comparison illustrates how each feature selection strategy impacts model performance across taxonomic levels and response types. Columns correspond to the six responses: four diseases and two yield outcomes. For a given panel (model in row and response in column), the boxplot corresponds to the different weighted F1 scores for every combination of normalizations/zero replacement strategies as well as different taxonomic levels (Table ). For example, the boxplots for the ALL-OTU model (first row) include weighted F1 scores of the model fit on 20 normalization/zero replacement strategies, and 5 taxonomic levels (100 different weighted F1 scores per outcome). The performance of these models was assessed using the weighted F1 score metric, which accounts for the imbalance in classes, making it particularly suited for this dataset. The dashed line in each panel corresponds to the average weighted F1 score of the model when fit with all random datasets (see section “ ” and Figs. and in Supplementary file). This line allows us to assess whether the real data has more predictive power than random data . For Random Forest models, feature selection performance ranks in this order: OTU-S3 > OTU-S1 > OTU-S2 > OTU-S0. Combining OTU-S3 with environmental information further enhances performance. Figure follows the same structure for Bayesian NN models. Similar to Random Forest, OTU-S3 shows better performance compared to OTU-S1, OTU-S2, and OTU-S0. The best results are achieved by combining OTU-S3 with soil information. The comparison against random datasets provides a baseline, helping to ensure that the model’s performance is not due to chance but reflects real patterns in the data. While these plots do not allow us to distinguish differences by taxonomic level or normalization/zero replacement strategy (more on that in the next subsections), we can identify outcomes (columns) that can be more accurately predicted across models (rows). Additionally, we can identify models that are capable of accurately predicting more outcomes. It is readily evident, for example, that the yield outcomes cannot be accurately predicted by any model as all the weighted F1 scores fall consistently below the dashed line. It is notable that the yield responses are those for which there is not a clear binarization strategy. Since we are artificially separating samples into the two classes (low and high yield) based on whether they are above or below the variety-specific median, samples on the boundary will in fact be very similar to each other, and thus, difficult to classify. Furthermore, the poor prediction of yield is not restricted to one data type (microbiome vs environmental) which also suggests that the prediction challenges arise from the binarization process rather than the model or set of predictors. Disease outcomes, on the contrary, display higher weighted F1 scores overall, and in particular, pitted scab displays weighted F1 scores that are consistently above the random prediction dashed line across different models. For the case of black scurf disease, even when the weighted F1 scores are very high, this is a deceiving result, as this disease outcome is highly imbalanced. This means that a naive model predicting all samples to belong to the majority class will have high prediction accuracy (see dashed line above 0.8 for random data). We investigate the prediction of black scurf disease more carefully with the augmented data that balances the proportion of both classes (Section “ ”). Figure showcases the best performance of RF and Bayesian NN models. In the RF model, the integration of alpha diversity and soil chemistry data (referred to as Alpha+Soil) yields the most accurate predictions across all outcomes. Conversely, optimal performance for the Bayesian NN models is achieved by OTUs identified as significant by both ML and network comparison strategies (more details on feature selection are provided in section “ ”), alongside soil chemistry data (denoted as OTU-S3+Soil). These findings highlight the importance of integrating environmental information with microbiome data to enhance predictive power for disease outcomes. Finally, given the poor performance on yield, we focus on the fine-grained description of results for the disease outcomes only for the remaining of the manuscript. Prediction of potato disease from microbiome data Normalization and zero replacement strategy has been proven to have impact on predictive power One of the goals of our study is to identify ideal data preprocessing steps that are guaranteed to maximize predictive power on the ML models. Figure shows the weighted F1 scores for pitted scab for different combinations of normalization and zero replacement strategies (x-axis) for the two types of models (RF and Bayesian NN). These analyses include all OTUs under the five taxonomic levels (different colors). Similar plots for other diseases are presented in Figs. , , , and in the Supplementary Material. In fact, there is considerable interaction between the normalization/zero replacement method and the taxonomic level. For example, for the RF model, the best result is achieved with Phylum level and cumulative sum scaling normalization with pseudo-zero replacement strategy (CSS+pseudo) or common sum scaling normalization without any zero replacement strategy (COM+none). Additionally, the rarefy+none and rarefy+multRepl strategies demonstrate good performance. For Bayesian NN, however, the best results are achieved with the common sum scaling normalization with the multiplicative zero replacement (COM+multRepl) for the Phylum level (See Fig. ). For a given normalization/zero replacement strategy (x-axis), the variability in the scatterplot points indicates that taxonomic levels have an impact on the predictive power of the model. When we compare the range of weighted F1 scores across normalization and zero replacement strategies, we see that the effect of the strategy is not negligible. For example, at the Phylum level, the lowest weighted F1 score is around 0.75 for centered log-ratio normalization with pseudo-zero replacement strategy (clr+pseudo) to around 0.9 for cumulative sum scaling normalization with pseudo-zero replacement strategy (CSS+pseudo). This implies that for a given taxonomic level, the resulted weighted F1 score will be highly influenced by the normalization and zero replacement strategy. Traditionally, microbiome researchers use the total sum scaling normalization without any zero replacement strategy (TSS+none) on their data which has a range of 0.80–0.90 weighted F1 scores for the RF model (0.8–0.85 for the Bayesian NN model) depending on the taxonomic level. The strong interaction effects of taxonomic level, normalization, and zero replacement strategy prevent us from making recommendations about the best data preprocessing practices that can be generalizable to other datasets. We conclude by suggesting data practitioners to consider trying a variety of appropriate normalization and zero replacement strategies instead of relying solely on one approach, but see section “ ” for more recommendations. Effective preservation of predictive signal with different feature selection strategies One of the standard steps in the ML pipeline is feature selection, especially for cases of high-dimensional data. We compare the ability to retain predictive signal of three feature selection strategies: standard importance score from ML methods, comparison of microbial network topologies, and combination of both. More details on the feature selection strategies can be found in Methods. Figure shows the weighted F1 scores for the two types of models (RF and Bayesian NN) on pitted scab disease (Scapbit) under different subsets of predictors: (1) all OTUs (ALL-OTU), (2) only OTUs that were identified as important by the ML strategy (OTU-S1), (3) only OTUs that were identified as important by the network comparison strategy (OTU-S2), (4) OTUs that were identified as important by both strategies (OTU-S3), or (5) OTUs that were not identified as important by neither strategy (OTU-S0). For fair comparison, we include the same number of predictors in OTU-S0 as in OTU-S3. Similar figures for other responses are shown in Figs. , , , and in the Supplementary Material. Again, we perceive a strong interaction between taxonomic level and feature selection strategy. For the RF model, the highest weighted F1 score is achieved when including all OTUs (ALL-OTU) at the Order level whereas for the Bayesian NN model, the highest weighted F1 score is achieved when including OTUs identified by the ML strategy (OTU-S1) at the Genus level. RF models on all OTUs (ALL-OTU) have a weighted F1 score above 0.8 in all taxonomic levels which suggests that this model could be a better alternative compared to Bayesian NN which is more computationally intensive. There are also smaller differences in RF models when comparing the performance on OTU-S3 (important OTUs) and ALL-OTU (all OTUs) which suggests that the feature selection strategy is sufficient to preserve the predictive signal in the data while reducing the number of predictors in the model. This is relevant for computationally intensive models such as Bayesian NN that do not allow the inclusion of all OTUs for certain taxonomic levels. To provide more interpretability, we compiled a comprehensive table that lists the key taxa across different taxonomic levels and responses. Each taxon is assigned a score based on its selection by ML and network-based feature selection methods: 0: OTUs not selected by either ML-based or network-based feature selection. 1: OTUs selected by ML-based feature selection. 2: OTUs selected by network-based feature selection. 3: OTUs selected by both ML-based and network-based approaches. This scoring system identifies the microbial taxa with the highest predictive importance for disease suppression or yield outcomes. End-users can access the corresponding table on our GitHub repository (link: https://github.com/solislemuslab/soil-microbiome-nn/blob/master/python-code/important_features_score.xlsx ) to determine which taxa are most relevant for practical interventions or microbiome management strategies in their fields. We focused on the top five taxa in each taxonomic level to examine the literature for evidence of their importance in soil microbiome studies. In each level, we found support for their key roles, which aligns with our findings, indicating that our methods for feature selection and combining the two approaches to identify reliable taxa were successful. Other important taxa, such as the top 10 percent, can be considered for further studies as significant candidates. For instance, our models showed that taxa from abundant phyla such as Proteobacteria and Chloroflexi , as well as taxa from less abundant phyla including Myxococcota , Spirochaetota , and NB1.j , were significant predictors (see Tables and in the Supplementary Materials). This suggests that both dominant and rare microbial community members play a crucial role in ecosystem functions related to crop health, such as nutrient cycling, growth promotion, and disease suppression . In addition, some of the key taxa identified have well-established roles in agricultural systems. For example, taxa from the Class level: Alphaproteobacteria and Gammaproteobacteria are frequently studied for their roles in soil health and disease suppression . Furthermore, Paenibacillaceae and, Syntrophaceae at the Family level have been recognized for their plant-growth-promoting properties and biocontrol capabilities. However, some taxa from the Genus level we identified, such as Acidothermus , Myxococcota , and Haliangium are relatively novel in this context and could represent promising candidates for further research. Moreover, Pseudonocardiales and Frankiales were detected as important taxa at the Order level. Robust prediction in imbalanced datasets with data augmentation High prediction power in imbalanced datasets is misleading as a naive predictor that classifies all samples as the majority class will have high accuracy. In our data, black scurf disease is highly imbalanced, and thus, the high prediction accuracy is unreliable. We confirmed, however, that after data augmentation which balanced the data, accurate prediction persisted. To illustrate this, Fig. depicts the weighted F1 scores on original and augmented datasets for all yield and disease outcomes (x-axis) and both models (RF and Bayesian NN). The range of each box plot depicts the weighted F1 scores for 20 normalized datasets at each taxonomic level. We observe that black scurf and pitted scab can be reliably predicted across taxonomic levels as their median weighted F1 scores for all taxonomic orders are around 0.8 when the models are fitted on the original datasets. As mentioned before, however, black scurf is highly imbalanced, so the results on the original data are not reliable. Fortunately, the median weighted F1 scores on augmented data (which is perfectly balanced by design) increase for both diseases, such that they are around 0.9 for all taxonomic levels. These results suggest that data augmentation, especially in cases of highly imbalanced data, is an appropriate strategy that improves the robustness of the model and, in some cases even increases the accuracy. One has to be careful, however, in that augmented data can yield certain models prohibited. For example, the Bayesian NN model could not be fit on the augmented datasets for Order, Family, or Genus levels due to computational limitations. Identifying general practices to predict potato disease from microbiome data using full model selection models As evidenced by our analyses, every single data and model choice has an impact on the predictive performance of our methods. The effects of different data preprocessing steps appear to strongly interact, and thus, we could not identify clear patterns on strategies to maximize prediction power. With a FMS strategy, however, we are able to identify the choices that yield the highest measure of performance. More details on the FMS models can be found in Methods. Figures and show the FMS decision trees for the RF and Bayesian NN models on pitted scab disease, respectively. A FMS decision tree shows the different data preprocessing steps that yield different weighted F1 scores, so that practitioners can select the options that result in the highest predictive power. Here, we have five taxonomic levels, 20 normalization+zero replacement strategies, and 2 data augmentation options: no data augmentation (Aug=0) and data augmentation (Aug=1). Thus, in total, we have 200 data preprocessing options (20 normalization strategies times 5 taxonomic levels times 2 data augmentation). To interpret a FMS decision tree, each node corresponds to a specific step in the data preprocessing pipeline, for example, whether to perform data augmentation or not. If the condition is true, we follow the branch to the left; if the condition is false, we follow the branch to the right. At the top of the decision tree, we have the root which represents the data preprocessing step that has the greatest effect on model accuracy. At the bottom of the decision tree, we have the leaves with the average weighted F1 score of the model fitted on the data that satisfies all conditions towards the root. Each node also displays the percentage of data preprocessing options included in the node. For example, in Fig. , the root node covers 100% of the options with average weighted F1 scores 0.865. The condition at the root node ( [12pt]{minimal} $$ {Aug}=0$$ Aug = 0 ) represents the case of “no data augmentation”. Thus, “true” (left of the root) means “no data augmentation”, and “false” (right of the root) means “data augmentation”. For simplicity, we denote the 20 normalization/zero replacement strategies as [12pt]{minimal} $$ {NM}_i$$ NM i for [12pt]{minimal} $$i=1, ,20$$ i = 1 , ⋯ , 20 . See section “ ” for a description on each normalization/zero replacement strategy. For the FMS decision tree for the RF model (Fig. ), the highest weighted F1 score (0.934 with 0.5% of the data) is achieved with data augmentation, normalization/zero replacement strategy #6 (CSS+pseudo), and Order level. Another path of the decision tree follows data augmentation and any normalization/zero replacement strategy except #6 (CSS+pseudo), #14 (rarefy+pseudo), and #18 (clr+pseudo) which yields an average weighted F1 scores of 0.892 for 42.5% of the data preprocessing options. If data augmentation is not an option (left of the root), the highest weighted F1 score available is 0.868 with Phylum level, and any normalization/zero replacement strategy except #18 (clr+pseudo) or #20 (clr+bayesMult). For the FMS decision trees on the other responses, see Figs. , , , and in the Supplementary Material. Similarly, in Fig. for the Bayesian NN model, the highest weighted F1 score (0.896 with 15% of the data preprocessing options) is achieved when we do data augmentation, we use any taxonomic level except Phylum, and we use any normalization/zero replacement strategy except #10 (COM+pseudo) and #18 (clr+pseudo). See Figs. , , , and in the Supplementary Material for other responses. While the specific recommendations on normalization, zero replacement and taxonomic level are model-specific, both models perform better with data augmentation. In terms of taxonomic level, we note that the Bayesian NN was only run on Phylum, Class, and Family levels, and thus, the highest accuracy is obtained with Class level (when Phylum [12pt]{minimal} $$=0$$ = 0 is true). This does not contradict the result from the RF that identified Order level as the one yielding higher accuracy. We cannot rule out that the Bayesian NN would also have higher accuracy with Order compared to Class. The results from the RF, though, seem to suggest that there is a peak at Order, and more granularity in Family and Genus does not seem to provide more predictive power. Table presents a summary of the best FMS decision tree results from the RF model for all responses (Figs. , , , , and ). We focus on diseases that have reasonable outcomes. The most critical decisions across all diseases involve first utilizing the augmentation method and then selecting the appropriate taxonomic level-either Family or Order-while avoiding Phylum (due to its lower information content) and Genus (which can lead to overfitting due to small sample sizes). The final key factor influencing Random Forest results is the choice of the normalization method. Our analysis suggests the best results are achieved with NM6: CSS+pseudo and NM14: rarefy+pseudo, while the least effective methods were NM18: clr+pseudo, NM2: TSS+pseudo, and NM20: clr+bayesMult. We strongly recommend that future studies employ the FMS decision tree approach when a gold standard is available for evaluation. In cases where there is no way to find the optimal normalization method, we suggest applying multiple normalization strategies (as outlined in this paper) and reporting consensus results based on the outcomes of different normalized datasets. This approach can help yield more robust and reliable results. For the Bayesian NN model, the focus was similarly on diseases with reasonable outcomes (Table ). Unlike the RF model, we did not observe a consistent pattern regarding the importance of augmentation or taxonomic level (Figs. , , , , and ). However, the results indicate that the selection of augmentation, taxonomic level, and normalization methods at the first node of the FMS tree significantly influences the model’s performance. This variability could be attributed to computational limitations that prevented us from running the model on all taxonomic levels. The best-performing normalization methods for the BNN model were NM1: TSS+none, NM2: TSS+pseudo, NM4: TSS+bayesMult, NM7: CSS+multRepl, NM12: COM+bayesMult, NM13: rarefy+none, and NM16: rarefy+bayesMult. In overall, with a deep investigation of the FMS results for all responses, we can recommend some normalization methods and taxonomic levels for further study. For RF model, it is recommended to use the augmenting method since RF can have better performance with more samples. For RF, normalization methods have a lower impact on the results, and in general, performing data augmentation and using a more specific taxonomic level like Class and Family are more important and located higher in the decision tree depth. This agrees with the fact that RF is known for being tolerant to high dimensional data, non-normal data, and missing values . In summary, four normalization methods consistently performed well across both disease and yield tasks: NM1: TSS+none, NM4: TSS+bayesMult, NM13: rarefy+none, and NM16: rarefy+bayesMult. There is no evidence from the FMS analysis to suggest that using these methods decreases performance, making them strong candidates for future studies. Prediction of potato disease from environmental data One of the questions to address in our work is whether prediction accuracy is improved by the inclusion of microbiome data, or if environmental factors (usually cheaper to collect) provide enough signal to classify potatoes in diseased or non-diseased groups. We found that environmental factors indeed provide sufficient signals to predict pitted scab disease as illustrated in Fig. which shows the weighted F1 scores by RF and Bayesian NN models based on environmental (soil characteristics) data for pitted scab. The range of each boxplot corresponds to the six scaling methods described in section “ ”. In contrast with the normalization methods in microbiome data, we observe here that the scaling methods do not seem to have an effect on prediction as evidenced by narrow boxplots, and that weighted F1 scores are all higher than 0.75, and therefore, comparable to the models fitted on microbiome data alone. These results suggest that environmental factors alone are powerful to predict the incidence of pitted scab in the tubers. As microbiome data is more expensive than environmental data, we suggest to prefer environmental predictors under restricted monetary budget. See Figs. , , , and in the Supplementary Material for other responses. Leveraging microbiome and environmental data in the prediction of potato disease As expected, prediction accuracy improves when both microbiome and environmental data are included. Fig. shows the weighted F1 scores by RF and Bayesian NN models based on combined datasets with environmental and microbial predictors for pitted scab. We only focus on the most accurate models identified in section “ ”. First, we note that a model that uses OTU abundances outperforms a model that uses alpha diversity as a predictor (comparison of Alpha with OTU-S3) for both types of models (RF and Bayesian NN). This suggests that we lose information by transforming abundances into diversity measures. Second, models including only OTU abundances (OTU-S3) perform comparably to models that include both types of predictors (OTU-S3+Soil+DS) which suggests that the microbial data indeed has substantial predictive power on its own, but adding microbiome to soil predictors may not provide much benefit for high predictive power, with the only exception of Phylum level OTU-S3+Soil+DS in a RF model (Fig. ). Generally, the model with only soil information (shown as a blue dashed line) performs just as accurately. Third, contrary to prior expectations that microbial communities at finer resolution would be a better choice for predicting pitted scab or other diseases, our study does not find any evidence that the prediction power increases when moving up from Phylum to Genus level. Particularly, the prediction power of OTU-S3 in RF model increases from Class to Genus, and this pattern is not preserved when diversity is used instead of OTU abundances. For example, a model with only alpha diversity as the predictor (Alpha) shows decreasing weighted F1 score as we move from Phylum to Genus level. Both models (RF and Bayesian NN) when including all types of predictors (OTU-S3+Soil+DS) result in the similar weighted F1 score regardless of taxonomic level. See Figs. , , , and in the Supplementary Material for other responses. According to Figs. , , , , and in the Supplementary Material, OTU-S3 performs well and is considered for comparison with other models in Figs. , , , , and in the Supplementary Material. By comparing results based on a few selected features (OTU-S3), we observe reliable performance for both disease and yield prediction. This underscores the robustness of OTU-S3 as an effective feature selection strategy. Running time The RF model was implemented and tested on a MacBook Pro with an Apple M1 Pro chip and 16 GB of RAM. The running time for the RF model is provided in Table in the supplementary materials. Depending on the number of trees and features selected, the RF model typically requires a few minutes per model on a dataset of our size (~200 samples). For the Bayesian NN model, the computational demands are significantly higher due to the need for probabilistic inference. We ran the Bayesian NN model on the Center for High Throughput Computation (CHTC) platform at UW-Madison with RTX2080ti graph cards. Despite using more computation resources, the Bayesian NN model would take between 24-72 h to compute the result for the models on all but Phylum level. Thus, it will be infeasible to run the Bayesian NN model on this particular problem on any personal devices. The runtime log of Bayesian NN models has been lost, unfortunately, as the CHTC platform only keeps the log files for 6 months, while the model was run more than 2 years ago. In summary, while the RF model can be efficiently run on a standard laptop such as a MacBook Pro, the Bayesian NN model requires significantly more computational resources. We recommend high-performance computing resources for readers planning to implement Bayesian NNs, particularly for larger datasets or more complex models. We implemented the H 2 O AutoML package in Python , an open-source package designed for automated ML, which trains multiple models such as RF, Gradient Boosting Machines, and Deep Learning models. H 2 O AutoML automates model selection and hyperparameter tuning, providing a comprehensive comparison of different ML methods. The best-performing model is selected based on Root Mean Square Error (RMSE), a standard metric for regression tasks. This process allows for a more robust evaluation of model performance in predicting continuous yield, avoiding the biases introduced by arbitrary binarization. However, to maintain consistency with the rest of the study, we also report results from RF models. We report Mean Absolute Percentage Error (MAPE) as the evaluation metric. MAPE calculates the average absolute difference between predicted and actual values as a percentage of the actual values. The formula for MAPE is defined as: [12pt]{minimal} $$ {MAPE} = _{i=1}^{n} | | 100 $$ MAPE = 1 n ∑ i = 1 n A i - F i A i × 100 where [12pt]{minimal} $$A_i$$ A i is the actual value, [12pt]{minimal} $$F_i$$ F i is the predicted value, and n is the number of observations. MAPE, evaluated on a scale from 0 to 1, is commonly used for regression tasks because it is particularly valuable as it is scale-independent, making it easy to compare performance across different models and datasets . A lower MAPE indicates better model accuracy, with 0% representing perfect prediction. A MAPE of, for example, 10% means that, on average, the model’s predictions are 10% off from the actual values. The results of MAPE values are visualized using box plots, which clearly represent the variability and performance of the models across different normalization methods and taxonomic levels in Fig. . The left part of the figure displays the results from RF models. For a more comprehensive analysis, we also show the best result obtained from the H 2 O AutoML model as shown in the right part of Fig. . We acknowledge that binarization does lead to information loss. The continuous modeling results demonstrate that predicting Yield_Plant is particularly challenging, likely due to the biological variability between individual plants. However, we observed better performance for Yield_Meter, which is a more stable measure due to its aggregation over a larger area. To balance accuracy and complexity in this study, we applied binarization to the continuous response in cases where it improved performance, although we recognize that this is not an ideal long-term solution. In future work, we plan to explore larger datasets and incorporate additional environmental variables to enhance model accuracy without relying on binarization. 2 O autoML for disease and binarized yield prediction Figure presents the weighted F1 scores for RF (left panel) and the best H 2 O AutoML models (right panel) across various responses, including both yield and disease outcomes (binary responses). Each boxplot represents different taxonomic levels (Phylum, Class, Order, Family, Genus) to evaluate model performance. The results show that RF performs comparable to the best H 2 O AutoML models, particularly excelling in Scabpit response. This demonstrates RF’s reliability, as its performance is consistently close to or equal to the more complex models selected by AutoML. We also choose Deep Learning Models as they are known to excel in exploring deep relationships between predictors. We believe it is essential to include a Deep Learning model as this is the cutting-edge method that resulted in most success in ML applications in the last decade. However, while deep learning approaches were considered, we determined that our dataset, with approximately 200 samples, is too small to effectively apply deep learning models. Deep learning typically requires larger datasets to avoid overfitting and produce generalizable results, and thus, it was not a viable option for our study. While there are many computational efficient models, none would be properly trained with 200 samples without underfitting or overfitting. Despite the computational challenges, Bayesian NNs are known to be informative for small sample sizes, offering protection against overfitting by modeling parameter uncertainty. Even though the Markov Chain Monte Carlo (MCMC) process for parameter estimation is very computationally inefficient in the context of Deep Learning, more research has gone into alternate estimation methods such as Variational Inference methods in the past few years which might drastically improve the computational efficiency in the near future. The Deep Learning result also validates the result of the RF model, a completely different approach that is more computationally efficient. Therefore, we will retain the Bayesian NN results in the paper alongside the other models, as they provide valuable insights despite the computational demands. Regarding black scurf disease, we found that it is a very imbalanced dataset, with only 6 samples exhibiting the disease. Due to the extremely limited number of cases, the results from machine learning methods cannot be considered reliable. Consequently, we decided not to focus on this disease in our current analysis. Instead, we plan to use data augmentation methods to improve predictions for black scurf. Additionally, we are collecting more datasets, after which we will apply ML methods to the original data to obtain more reliable results. Although the H 2 O AutoML framework identifies slightly better-performing models in certain cases, RF maintains a strong balance between predictive performance and computational efficiency. In yield-related predictions like Yield_Plant and Yield_Meter, RF produces F1 scores that are very close to those of the best AutoML models, making it a dependable choice. Given the practical constraints of running models across 600 configurations (5 taxonomic levels [12pt]{minimal} $$$$ × 20 normalization methods [12pt]{minimal} $$$$ × 6 responses) and the inclusion of environmental predictors, RF’s lower computational demand provides an efficient solution without compromising accuracy. This comparison supports our choice of RF as a reliable and efficient model, reinforcing the robustness of our study’s findings. Furthermore, we aimed to use one robust model that works effectively across all response types, and RF consistently met this criterion, demonstrating reliable performance across various conditions. Given these results, RF was selected as the main model due to its interpretability, robustness, and lower computational demands. While H 2 O AutoML occasionally yielded slightly higher scores, RF’s performance remained consistently close, underscoring its suitability as the primary model for microbiome studies with limited computational resources. First, we identify the outcomes (disease or binarized yield) that are accurately predicted across models (and are thus robust to prediction regardless of model choices), as well as the models that accurately predict across outcomes (and are thus the most powerful model alternatives). To do so, we aggregate the weighted F1 scores on data preprocessing choices such as normalization, zero replacement, and taxonomic levels for every model and every outcome. We employed RF and Bayesian NN models across various predictive scenarios (14 different models). The predictive capabilities of these models are illustrated in Figs. (RF) and (Bayesian NN), providing a comprehensive analysis of all 14 models outlined in Table . This detailed comparison illustrates how each feature selection strategy impacts model performance across taxonomic levels and response types. Columns correspond to the six responses: four diseases and two yield outcomes. For a given panel (model in row and response in column), the boxplot corresponds to the different weighted F1 scores for every combination of normalizations/zero replacement strategies as well as different taxonomic levels (Table ). For example, the boxplots for the ALL-OTU model (first row) include weighted F1 scores of the model fit on 20 normalization/zero replacement strategies, and 5 taxonomic levels (100 different weighted F1 scores per outcome). The performance of these models was assessed using the weighted F1 score metric, which accounts for the imbalance in classes, making it particularly suited for this dataset. The dashed line in each panel corresponds to the average weighted F1 score of the model when fit with all random datasets (see section “ ” and Figs. and in Supplementary file). This line allows us to assess whether the real data has more predictive power than random data . For Random Forest models, feature selection performance ranks in this order: OTU-S3 > OTU-S1 > OTU-S2 > OTU-S0. Combining OTU-S3 with environmental information further enhances performance. Figure follows the same structure for Bayesian NN models. Similar to Random Forest, OTU-S3 shows better performance compared to OTU-S1, OTU-S2, and OTU-S0. The best results are achieved by combining OTU-S3 with soil information. The comparison against random datasets provides a baseline, helping to ensure that the model’s performance is not due to chance but reflects real patterns in the data. While these plots do not allow us to distinguish differences by taxonomic level or normalization/zero replacement strategy (more on that in the next subsections), we can identify outcomes (columns) that can be more accurately predicted across models (rows). Additionally, we can identify models that are capable of accurately predicting more outcomes. It is readily evident, for example, that the yield outcomes cannot be accurately predicted by any model as all the weighted F1 scores fall consistently below the dashed line. It is notable that the yield responses are those for which there is not a clear binarization strategy. Since we are artificially separating samples into the two classes (low and high yield) based on whether they are above or below the variety-specific median, samples on the boundary will in fact be very similar to each other, and thus, difficult to classify. Furthermore, the poor prediction of yield is not restricted to one data type (microbiome vs environmental) which also suggests that the prediction challenges arise from the binarization process rather than the model or set of predictors. Disease outcomes, on the contrary, display higher weighted F1 scores overall, and in particular, pitted scab displays weighted F1 scores that are consistently above the random prediction dashed line across different models. For the case of black scurf disease, even when the weighted F1 scores are very high, this is a deceiving result, as this disease outcome is highly imbalanced. This means that a naive model predicting all samples to belong to the majority class will have high prediction accuracy (see dashed line above 0.8 for random data). We investigate the prediction of black scurf disease more carefully with the augmented data that balances the proportion of both classes (Section “ ”). Figure showcases the best performance of RF and Bayesian NN models. In the RF model, the integration of alpha diversity and soil chemistry data (referred to as Alpha+Soil) yields the most accurate predictions across all outcomes. Conversely, optimal performance for the Bayesian NN models is achieved by OTUs identified as significant by both ML and network comparison strategies (more details on feature selection are provided in section “ ”), alongside soil chemistry data (denoted as OTU-S3+Soil). These findings highlight the importance of integrating environmental information with microbiome data to enhance predictive power for disease outcomes. Finally, given the poor performance on yield, we focus on the fine-grained description of results for the disease outcomes only for the remaining of the manuscript. Normalization and zero replacement strategy has been proven to have impact on predictive power One of the goals of our study is to identify ideal data preprocessing steps that are guaranteed to maximize predictive power on the ML models. Figure shows the weighted F1 scores for pitted scab for different combinations of normalization and zero replacement strategies (x-axis) for the two types of models (RF and Bayesian NN). These analyses include all OTUs under the five taxonomic levels (different colors). Similar plots for other diseases are presented in Figs. , , , and in the Supplementary Material. In fact, there is considerable interaction between the normalization/zero replacement method and the taxonomic level. For example, for the RF model, the best result is achieved with Phylum level and cumulative sum scaling normalization with pseudo-zero replacement strategy (CSS+pseudo) or common sum scaling normalization without any zero replacement strategy (COM+none). Additionally, the rarefy+none and rarefy+multRepl strategies demonstrate good performance. For Bayesian NN, however, the best results are achieved with the common sum scaling normalization with the multiplicative zero replacement (COM+multRepl) for the Phylum level (See Fig. ). For a given normalization/zero replacement strategy (x-axis), the variability in the scatterplot points indicates that taxonomic levels have an impact on the predictive power of the model. When we compare the range of weighted F1 scores across normalization and zero replacement strategies, we see that the effect of the strategy is not negligible. For example, at the Phylum level, the lowest weighted F1 score is around 0.75 for centered log-ratio normalization with pseudo-zero replacement strategy (clr+pseudo) to around 0.9 for cumulative sum scaling normalization with pseudo-zero replacement strategy (CSS+pseudo). This implies that for a given taxonomic level, the resulted weighted F1 score will be highly influenced by the normalization and zero replacement strategy. Traditionally, microbiome researchers use the total sum scaling normalization without any zero replacement strategy (TSS+none) on their data which has a range of 0.80–0.90 weighted F1 scores for the RF model (0.8–0.85 for the Bayesian NN model) depending on the taxonomic level. The strong interaction effects of taxonomic level, normalization, and zero replacement strategy prevent us from making recommendations about the best data preprocessing practices that can be generalizable to other datasets. We conclude by suggesting data practitioners to consider trying a variety of appropriate normalization and zero replacement strategies instead of relying solely on one approach, but see section “ ” for more recommendations. Effective preservation of predictive signal with different feature selection strategies One of the standard steps in the ML pipeline is feature selection, especially for cases of high-dimensional data. We compare the ability to retain predictive signal of three feature selection strategies: standard importance score from ML methods, comparison of microbial network topologies, and combination of both. More details on the feature selection strategies can be found in Methods. Figure shows the weighted F1 scores for the two types of models (RF and Bayesian NN) on pitted scab disease (Scapbit) under different subsets of predictors: (1) all OTUs (ALL-OTU), (2) only OTUs that were identified as important by the ML strategy (OTU-S1), (3) only OTUs that were identified as important by the network comparison strategy (OTU-S2), (4) OTUs that were identified as important by both strategies (OTU-S3), or (5) OTUs that were not identified as important by neither strategy (OTU-S0). For fair comparison, we include the same number of predictors in OTU-S0 as in OTU-S3. Similar figures for other responses are shown in Figs. , , , and in the Supplementary Material. Again, we perceive a strong interaction between taxonomic level and feature selection strategy. For the RF model, the highest weighted F1 score is achieved when including all OTUs (ALL-OTU) at the Order level whereas for the Bayesian NN model, the highest weighted F1 score is achieved when including OTUs identified by the ML strategy (OTU-S1) at the Genus level. RF models on all OTUs (ALL-OTU) have a weighted F1 score above 0.8 in all taxonomic levels which suggests that this model could be a better alternative compared to Bayesian NN which is more computationally intensive. There are also smaller differences in RF models when comparing the performance on OTU-S3 (important OTUs) and ALL-OTU (all OTUs) which suggests that the feature selection strategy is sufficient to preserve the predictive signal in the data while reducing the number of predictors in the model. This is relevant for computationally intensive models such as Bayesian NN that do not allow the inclusion of all OTUs for certain taxonomic levels. To provide more interpretability, we compiled a comprehensive table that lists the key taxa across different taxonomic levels and responses. Each taxon is assigned a score based on its selection by ML and network-based feature selection methods: 0: OTUs not selected by either ML-based or network-based feature selection. 1: OTUs selected by ML-based feature selection. 2: OTUs selected by network-based feature selection. 3: OTUs selected by both ML-based and network-based approaches. This scoring system identifies the microbial taxa with the highest predictive importance for disease suppression or yield outcomes. End-users can access the corresponding table on our GitHub repository (link: https://github.com/solislemuslab/soil-microbiome-nn/blob/master/python-code/important_features_score.xlsx ) to determine which taxa are most relevant for practical interventions or microbiome management strategies in their fields. We focused on the top five taxa in each taxonomic level to examine the literature for evidence of their importance in soil microbiome studies. In each level, we found support for their key roles, which aligns with our findings, indicating that our methods for feature selection and combining the two approaches to identify reliable taxa were successful. Other important taxa, such as the top 10 percent, can be considered for further studies as significant candidates. For instance, our models showed that taxa from abundant phyla such as Proteobacteria and Chloroflexi , as well as taxa from less abundant phyla including Myxococcota , Spirochaetota , and NB1.j , were significant predictors (see Tables and in the Supplementary Materials). This suggests that both dominant and rare microbial community members play a crucial role in ecosystem functions related to crop health, such as nutrient cycling, growth promotion, and disease suppression . In addition, some of the key taxa identified have well-established roles in agricultural systems. For example, taxa from the Class level: Alphaproteobacteria and Gammaproteobacteria are frequently studied for their roles in soil health and disease suppression . Furthermore, Paenibacillaceae and, Syntrophaceae at the Family level have been recognized for their plant-growth-promoting properties and biocontrol capabilities. However, some taxa from the Genus level we identified, such as Acidothermus , Myxococcota , and Haliangium are relatively novel in this context and could represent promising candidates for further research. Moreover, Pseudonocardiales and Frankiales were detected as important taxa at the Order level. Robust prediction in imbalanced datasets with data augmentation High prediction power in imbalanced datasets is misleading as a naive predictor that classifies all samples as the majority class will have high accuracy. In our data, black scurf disease is highly imbalanced, and thus, the high prediction accuracy is unreliable. We confirmed, however, that after data augmentation which balanced the data, accurate prediction persisted. To illustrate this, Fig. depicts the weighted F1 scores on original and augmented datasets for all yield and disease outcomes (x-axis) and both models (RF and Bayesian NN). The range of each box plot depicts the weighted F1 scores for 20 normalized datasets at each taxonomic level. We observe that black scurf and pitted scab can be reliably predicted across taxonomic levels as their median weighted F1 scores for all taxonomic orders are around 0.8 when the models are fitted on the original datasets. As mentioned before, however, black scurf is highly imbalanced, so the results on the original data are not reliable. Fortunately, the median weighted F1 scores on augmented data (which is perfectly balanced by design) increase for both diseases, such that they are around 0.9 for all taxonomic levels. These results suggest that data augmentation, especially in cases of highly imbalanced data, is an appropriate strategy that improves the robustness of the model and, in some cases even increases the accuracy. One has to be careful, however, in that augmented data can yield certain models prohibited. For example, the Bayesian NN model could not be fit on the augmented datasets for Order, Family, or Genus levels due to computational limitations. Identifying general practices to predict potato disease from microbiome data using full model selection models As evidenced by our analyses, every single data and model choice has an impact on the predictive performance of our methods. The effects of different data preprocessing steps appear to strongly interact, and thus, we could not identify clear patterns on strategies to maximize prediction power. With a FMS strategy, however, we are able to identify the choices that yield the highest measure of performance. More details on the FMS models can be found in Methods. Figures and show the FMS decision trees for the RF and Bayesian NN models on pitted scab disease, respectively. A FMS decision tree shows the different data preprocessing steps that yield different weighted F1 scores, so that practitioners can select the options that result in the highest predictive power. Here, we have five taxonomic levels, 20 normalization+zero replacement strategies, and 2 data augmentation options: no data augmentation (Aug=0) and data augmentation (Aug=1). Thus, in total, we have 200 data preprocessing options (20 normalization strategies times 5 taxonomic levels times 2 data augmentation). To interpret a FMS decision tree, each node corresponds to a specific step in the data preprocessing pipeline, for example, whether to perform data augmentation or not. If the condition is true, we follow the branch to the left; if the condition is false, we follow the branch to the right. At the top of the decision tree, we have the root which represents the data preprocessing step that has the greatest effect on model accuracy. At the bottom of the decision tree, we have the leaves with the average weighted F1 score of the model fitted on the data that satisfies all conditions towards the root. Each node also displays the percentage of data preprocessing options included in the node. For example, in Fig. , the root node covers 100% of the options with average weighted F1 scores 0.865. The condition at the root node ( [12pt]{minimal} $$ {Aug}=0$$ Aug = 0 ) represents the case of “no data augmentation”. Thus, “true” (left of the root) means “no data augmentation”, and “false” (right of the root) means “data augmentation”. For simplicity, we denote the 20 normalization/zero replacement strategies as [12pt]{minimal} $$ {NM}_i$$ NM i for [12pt]{minimal} $$i=1, ,20$$ i = 1 , ⋯ , 20 . See section “ ” for a description on each normalization/zero replacement strategy. For the FMS decision tree for the RF model (Fig. ), the highest weighted F1 score (0.934 with 0.5% of the data) is achieved with data augmentation, normalization/zero replacement strategy #6 (CSS+pseudo), and Order level. Another path of the decision tree follows data augmentation and any normalization/zero replacement strategy except #6 (CSS+pseudo), #14 (rarefy+pseudo), and #18 (clr+pseudo) which yields an average weighted F1 scores of 0.892 for 42.5% of the data preprocessing options. If data augmentation is not an option (left of the root), the highest weighted F1 score available is 0.868 with Phylum level, and any normalization/zero replacement strategy except #18 (clr+pseudo) or #20 (clr+bayesMult). For the FMS decision trees on the other responses, see Figs. , , , and in the Supplementary Material. Similarly, in Fig. for the Bayesian NN model, the highest weighted F1 score (0.896 with 15% of the data preprocessing options) is achieved when we do data augmentation, we use any taxonomic level except Phylum, and we use any normalization/zero replacement strategy except #10 (COM+pseudo) and #18 (clr+pseudo). See Figs. , , , and in the Supplementary Material for other responses. While the specific recommendations on normalization, zero replacement and taxonomic level are model-specific, both models perform better with data augmentation. In terms of taxonomic level, we note that the Bayesian NN was only run on Phylum, Class, and Family levels, and thus, the highest accuracy is obtained with Class level (when Phylum [12pt]{minimal} $$=0$$ = 0 is true). This does not contradict the result from the RF that identified Order level as the one yielding higher accuracy. We cannot rule out that the Bayesian NN would also have higher accuracy with Order compared to Class. The results from the RF, though, seem to suggest that there is a peak at Order, and more granularity in Family and Genus does not seem to provide more predictive power. Table presents a summary of the best FMS decision tree results from the RF model for all responses (Figs. , , , , and ). We focus on diseases that have reasonable outcomes. The most critical decisions across all diseases involve first utilizing the augmentation method and then selecting the appropriate taxonomic level-either Family or Order-while avoiding Phylum (due to its lower information content) and Genus (which can lead to overfitting due to small sample sizes). The final key factor influencing Random Forest results is the choice of the normalization method. Our analysis suggests the best results are achieved with NM6: CSS+pseudo and NM14: rarefy+pseudo, while the least effective methods were NM18: clr+pseudo, NM2: TSS+pseudo, and NM20: clr+bayesMult. We strongly recommend that future studies employ the FMS decision tree approach when a gold standard is available for evaluation. In cases where there is no way to find the optimal normalization method, we suggest applying multiple normalization strategies (as outlined in this paper) and reporting consensus results based on the outcomes of different normalized datasets. This approach can help yield more robust and reliable results. For the Bayesian NN model, the focus was similarly on diseases with reasonable outcomes (Table ). Unlike the RF model, we did not observe a consistent pattern regarding the importance of augmentation or taxonomic level (Figs. , , , , and ). However, the results indicate that the selection of augmentation, taxonomic level, and normalization methods at the first node of the FMS tree significantly influences the model’s performance. This variability could be attributed to computational limitations that prevented us from running the model on all taxonomic levels. The best-performing normalization methods for the BNN model were NM1: TSS+none, NM2: TSS+pseudo, NM4: TSS+bayesMult, NM7: CSS+multRepl, NM12: COM+bayesMult, NM13: rarefy+none, and NM16: rarefy+bayesMult. In overall, with a deep investigation of the FMS results for all responses, we can recommend some normalization methods and taxonomic levels for further study. For RF model, it is recommended to use the augmenting method since RF can have better performance with more samples. For RF, normalization methods have a lower impact on the results, and in general, performing data augmentation and using a more specific taxonomic level like Class and Family are more important and located higher in the decision tree depth. This agrees with the fact that RF is known for being tolerant to high dimensional data, non-normal data, and missing values . In summary, four normalization methods consistently performed well across both disease and yield tasks: NM1: TSS+none, NM4: TSS+bayesMult, NM13: rarefy+none, and NM16: rarefy+bayesMult. There is no evidence from the FMS analysis to suggest that using these methods decreases performance, making them strong candidates for future studies. One of the goals of our study is to identify ideal data preprocessing steps that are guaranteed to maximize predictive power on the ML models. Figure shows the weighted F1 scores for pitted scab for different combinations of normalization and zero replacement strategies (x-axis) for the two types of models (RF and Bayesian NN). These analyses include all OTUs under the five taxonomic levels (different colors). Similar plots for other diseases are presented in Figs. , , , and in the Supplementary Material. In fact, there is considerable interaction between the normalization/zero replacement method and the taxonomic level. For example, for the RF model, the best result is achieved with Phylum level and cumulative sum scaling normalization with pseudo-zero replacement strategy (CSS+pseudo) or common sum scaling normalization without any zero replacement strategy (COM+none). Additionally, the rarefy+none and rarefy+multRepl strategies demonstrate good performance. For Bayesian NN, however, the best results are achieved with the common sum scaling normalization with the multiplicative zero replacement (COM+multRepl) for the Phylum level (See Fig. ). For a given normalization/zero replacement strategy (x-axis), the variability in the scatterplot points indicates that taxonomic levels have an impact on the predictive power of the model. When we compare the range of weighted F1 scores across normalization and zero replacement strategies, we see that the effect of the strategy is not negligible. For example, at the Phylum level, the lowest weighted F1 score is around 0.75 for centered log-ratio normalization with pseudo-zero replacement strategy (clr+pseudo) to around 0.9 for cumulative sum scaling normalization with pseudo-zero replacement strategy (CSS+pseudo). This implies that for a given taxonomic level, the resulted weighted F1 score will be highly influenced by the normalization and zero replacement strategy. Traditionally, microbiome researchers use the total sum scaling normalization without any zero replacement strategy (TSS+none) on their data which has a range of 0.80–0.90 weighted F1 scores for the RF model (0.8–0.85 for the Bayesian NN model) depending on the taxonomic level. The strong interaction effects of taxonomic level, normalization, and zero replacement strategy prevent us from making recommendations about the best data preprocessing practices that can be generalizable to other datasets. We conclude by suggesting data practitioners to consider trying a variety of appropriate normalization and zero replacement strategies instead of relying solely on one approach, but see section “ ” for more recommendations. One of the standard steps in the ML pipeline is feature selection, especially for cases of high-dimensional data. We compare the ability to retain predictive signal of three feature selection strategies: standard importance score from ML methods, comparison of microbial network topologies, and combination of both. More details on the feature selection strategies can be found in Methods. Figure shows the weighted F1 scores for the two types of models (RF and Bayesian NN) on pitted scab disease (Scapbit) under different subsets of predictors: (1) all OTUs (ALL-OTU), (2) only OTUs that were identified as important by the ML strategy (OTU-S1), (3) only OTUs that were identified as important by the network comparison strategy (OTU-S2), (4) OTUs that were identified as important by both strategies (OTU-S3), or (5) OTUs that were not identified as important by neither strategy (OTU-S0). For fair comparison, we include the same number of predictors in OTU-S0 as in OTU-S3. Similar figures for other responses are shown in Figs. , , , and in the Supplementary Material. Again, we perceive a strong interaction between taxonomic level and feature selection strategy. For the RF model, the highest weighted F1 score is achieved when including all OTUs (ALL-OTU) at the Order level whereas for the Bayesian NN model, the highest weighted F1 score is achieved when including OTUs identified by the ML strategy (OTU-S1) at the Genus level. RF models on all OTUs (ALL-OTU) have a weighted F1 score above 0.8 in all taxonomic levels which suggests that this model could be a better alternative compared to Bayesian NN which is more computationally intensive. There are also smaller differences in RF models when comparing the performance on OTU-S3 (important OTUs) and ALL-OTU (all OTUs) which suggests that the feature selection strategy is sufficient to preserve the predictive signal in the data while reducing the number of predictors in the model. This is relevant for computationally intensive models such as Bayesian NN that do not allow the inclusion of all OTUs for certain taxonomic levels. To provide more interpretability, we compiled a comprehensive table that lists the key taxa across different taxonomic levels and responses. Each taxon is assigned a score based on its selection by ML and network-based feature selection methods: 0: OTUs not selected by either ML-based or network-based feature selection. 1: OTUs selected by ML-based feature selection. 2: OTUs selected by network-based feature selection. 3: OTUs selected by both ML-based and network-based approaches. This scoring system identifies the microbial taxa with the highest predictive importance for disease suppression or yield outcomes. End-users can access the corresponding table on our GitHub repository (link: https://github.com/solislemuslab/soil-microbiome-nn/blob/master/python-code/important_features_score.xlsx ) to determine which taxa are most relevant for practical interventions or microbiome management strategies in their fields. We focused on the top five taxa in each taxonomic level to examine the literature for evidence of their importance in soil microbiome studies. In each level, we found support for their key roles, which aligns with our findings, indicating that our methods for feature selection and combining the two approaches to identify reliable taxa were successful. Other important taxa, such as the top 10 percent, can be considered for further studies as significant candidates. For instance, our models showed that taxa from abundant phyla such as Proteobacteria and Chloroflexi , as well as taxa from less abundant phyla including Myxococcota , Spirochaetota , and NB1.j , were significant predictors (see Tables and in the Supplementary Materials). This suggests that both dominant and rare microbial community members play a crucial role in ecosystem functions related to crop health, such as nutrient cycling, growth promotion, and disease suppression . In addition, some of the key taxa identified have well-established roles in agricultural systems. For example, taxa from the Class level: Alphaproteobacteria and Gammaproteobacteria are frequently studied for their roles in soil health and disease suppression . Furthermore, Paenibacillaceae and, Syntrophaceae at the Family level have been recognized for their plant-growth-promoting properties and biocontrol capabilities. However, some taxa from the Genus level we identified, such as Acidothermus , Myxococcota , and Haliangium are relatively novel in this context and could represent promising candidates for further research. Moreover, Pseudonocardiales and Frankiales were detected as important taxa at the Order level. High prediction power in imbalanced datasets is misleading as a naive predictor that classifies all samples as the majority class will have high accuracy. In our data, black scurf disease is highly imbalanced, and thus, the high prediction accuracy is unreliable. We confirmed, however, that after data augmentation which balanced the data, accurate prediction persisted. To illustrate this, Fig. depicts the weighted F1 scores on original and augmented datasets for all yield and disease outcomes (x-axis) and both models (RF and Bayesian NN). The range of each box plot depicts the weighted F1 scores for 20 normalized datasets at each taxonomic level. We observe that black scurf and pitted scab can be reliably predicted across taxonomic levels as their median weighted F1 scores for all taxonomic orders are around 0.8 when the models are fitted on the original datasets. As mentioned before, however, black scurf is highly imbalanced, so the results on the original data are not reliable. Fortunately, the median weighted F1 scores on augmented data (which is perfectly balanced by design) increase for both diseases, such that they are around 0.9 for all taxonomic levels. These results suggest that data augmentation, especially in cases of highly imbalanced data, is an appropriate strategy that improves the robustness of the model and, in some cases even increases the accuracy. One has to be careful, however, in that augmented data can yield certain models prohibited. For example, the Bayesian NN model could not be fit on the augmented datasets for Order, Family, or Genus levels due to computational limitations. As evidenced by our analyses, every single data and model choice has an impact on the predictive performance of our methods. The effects of different data preprocessing steps appear to strongly interact, and thus, we could not identify clear patterns on strategies to maximize prediction power. With a FMS strategy, however, we are able to identify the choices that yield the highest measure of performance. More details on the FMS models can be found in Methods. Figures and show the FMS decision trees for the RF and Bayesian NN models on pitted scab disease, respectively. A FMS decision tree shows the different data preprocessing steps that yield different weighted F1 scores, so that practitioners can select the options that result in the highest predictive power. Here, we have five taxonomic levels, 20 normalization+zero replacement strategies, and 2 data augmentation options: no data augmentation (Aug=0) and data augmentation (Aug=1). Thus, in total, we have 200 data preprocessing options (20 normalization strategies times 5 taxonomic levels times 2 data augmentation). To interpret a FMS decision tree, each node corresponds to a specific step in the data preprocessing pipeline, for example, whether to perform data augmentation or not. If the condition is true, we follow the branch to the left; if the condition is false, we follow the branch to the right. At the top of the decision tree, we have the root which represents the data preprocessing step that has the greatest effect on model accuracy. At the bottom of the decision tree, we have the leaves with the average weighted F1 score of the model fitted on the data that satisfies all conditions towards the root. Each node also displays the percentage of data preprocessing options included in the node. For example, in Fig. , the root node covers 100% of the options with average weighted F1 scores 0.865. The condition at the root node ( [12pt]{minimal} $$ {Aug}=0$$ Aug = 0 ) represents the case of “no data augmentation”. Thus, “true” (left of the root) means “no data augmentation”, and “false” (right of the root) means “data augmentation”. For simplicity, we denote the 20 normalization/zero replacement strategies as [12pt]{minimal} $$ {NM}_i$$ NM i for [12pt]{minimal} $$i=1, ,20$$ i = 1 , ⋯ , 20 . See section “ ” for a description on each normalization/zero replacement strategy. For the FMS decision tree for the RF model (Fig. ), the highest weighted F1 score (0.934 with 0.5% of the data) is achieved with data augmentation, normalization/zero replacement strategy #6 (CSS+pseudo), and Order level. Another path of the decision tree follows data augmentation and any normalization/zero replacement strategy except #6 (CSS+pseudo), #14 (rarefy+pseudo), and #18 (clr+pseudo) which yields an average weighted F1 scores of 0.892 for 42.5% of the data preprocessing options. If data augmentation is not an option (left of the root), the highest weighted F1 score available is 0.868 with Phylum level, and any normalization/zero replacement strategy except #18 (clr+pseudo) or #20 (clr+bayesMult). For the FMS decision trees on the other responses, see Figs. , , , and in the Supplementary Material. Similarly, in Fig. for the Bayesian NN model, the highest weighted F1 score (0.896 with 15% of the data preprocessing options) is achieved when we do data augmentation, we use any taxonomic level except Phylum, and we use any normalization/zero replacement strategy except #10 (COM+pseudo) and #18 (clr+pseudo). See Figs. , , , and in the Supplementary Material for other responses. While the specific recommendations on normalization, zero replacement and taxonomic level are model-specific, both models perform better with data augmentation. In terms of taxonomic level, we note that the Bayesian NN was only run on Phylum, Class, and Family levels, and thus, the highest accuracy is obtained with Class level (when Phylum [12pt]{minimal} $$=0$$ = 0 is true). This does not contradict the result from the RF that identified Order level as the one yielding higher accuracy. We cannot rule out that the Bayesian NN would also have higher accuracy with Order compared to Class. The results from the RF, though, seem to suggest that there is a peak at Order, and more granularity in Family and Genus does not seem to provide more predictive power. Table presents a summary of the best FMS decision tree results from the RF model for all responses (Figs. , , , , and ). We focus on diseases that have reasonable outcomes. The most critical decisions across all diseases involve first utilizing the augmentation method and then selecting the appropriate taxonomic level-either Family or Order-while avoiding Phylum (due to its lower information content) and Genus (which can lead to overfitting due to small sample sizes). The final key factor influencing Random Forest results is the choice of the normalization method. Our analysis suggests the best results are achieved with NM6: CSS+pseudo and NM14: rarefy+pseudo, while the least effective methods were NM18: clr+pseudo, NM2: TSS+pseudo, and NM20: clr+bayesMult. We strongly recommend that future studies employ the FMS decision tree approach when a gold standard is available for evaluation. In cases where there is no way to find the optimal normalization method, we suggest applying multiple normalization strategies (as outlined in this paper) and reporting consensus results based on the outcomes of different normalized datasets. This approach can help yield more robust and reliable results. For the Bayesian NN model, the focus was similarly on diseases with reasonable outcomes (Table ). Unlike the RF model, we did not observe a consistent pattern regarding the importance of augmentation or taxonomic level (Figs. , , , , and ). However, the results indicate that the selection of augmentation, taxonomic level, and normalization methods at the first node of the FMS tree significantly influences the model’s performance. This variability could be attributed to computational limitations that prevented us from running the model on all taxonomic levels. The best-performing normalization methods for the BNN model were NM1: TSS+none, NM2: TSS+pseudo, NM4: TSS+bayesMult, NM7: CSS+multRepl, NM12: COM+bayesMult, NM13: rarefy+none, and NM16: rarefy+bayesMult. In overall, with a deep investigation of the FMS results for all responses, we can recommend some normalization methods and taxonomic levels for further study. For RF model, it is recommended to use the augmenting method since RF can have better performance with more samples. For RF, normalization methods have a lower impact on the results, and in general, performing data augmentation and using a more specific taxonomic level like Class and Family are more important and located higher in the decision tree depth. This agrees with the fact that RF is known for being tolerant to high dimensional data, non-normal data, and missing values . In summary, four normalization methods consistently performed well across both disease and yield tasks: NM1: TSS+none, NM4: TSS+bayesMult, NM13: rarefy+none, and NM16: rarefy+bayesMult. There is no evidence from the FMS analysis to suggest that using these methods decreases performance, making them strong candidates for future studies. One of the questions to address in our work is whether prediction accuracy is improved by the inclusion of microbiome data, or if environmental factors (usually cheaper to collect) provide enough signal to classify potatoes in diseased or non-diseased groups. We found that environmental factors indeed provide sufficient signals to predict pitted scab disease as illustrated in Fig. which shows the weighted F1 scores by RF and Bayesian NN models based on environmental (soil characteristics) data for pitted scab. The range of each boxplot corresponds to the six scaling methods described in section “ ”. In contrast with the normalization methods in microbiome data, we observe here that the scaling methods do not seem to have an effect on prediction as evidenced by narrow boxplots, and that weighted F1 scores are all higher than 0.75, and therefore, comparable to the models fitted on microbiome data alone. These results suggest that environmental factors alone are powerful to predict the incidence of pitted scab in the tubers. As microbiome data is more expensive than environmental data, we suggest to prefer environmental predictors under restricted monetary budget. See Figs. , , , and in the Supplementary Material for other responses. As expected, prediction accuracy improves when both microbiome and environmental data are included. Fig. shows the weighted F1 scores by RF and Bayesian NN models based on combined datasets with environmental and microbial predictors for pitted scab. We only focus on the most accurate models identified in section “ ”. First, we note that a model that uses OTU abundances outperforms a model that uses alpha diversity as a predictor (comparison of Alpha with OTU-S3) for both types of models (RF and Bayesian NN). This suggests that we lose information by transforming abundances into diversity measures. Second, models including only OTU abundances (OTU-S3) perform comparably to models that include both types of predictors (OTU-S3+Soil+DS) which suggests that the microbial data indeed has substantial predictive power on its own, but adding microbiome to soil predictors may not provide much benefit for high predictive power, with the only exception of Phylum level OTU-S3+Soil+DS in a RF model (Fig. ). Generally, the model with only soil information (shown as a blue dashed line) performs just as accurately. Third, contrary to prior expectations that microbial communities at finer resolution would be a better choice for predicting pitted scab or other diseases, our study does not find any evidence that the prediction power increases when moving up from Phylum to Genus level. Particularly, the prediction power of OTU-S3 in RF model increases from Class to Genus, and this pattern is not preserved when diversity is used instead of OTU abundances. For example, a model with only alpha diversity as the predictor (Alpha) shows decreasing weighted F1 score as we move from Phylum to Genus level. Both models (RF and Bayesian NN) when including all types of predictors (OTU-S3+Soil+DS) result in the similar weighted F1 score regardless of taxonomic level. See Figs. , , , and in the Supplementary Material for other responses. According to Figs. , , , , and in the Supplementary Material, OTU-S3 performs well and is considered for comparison with other models in Figs. , , , , and in the Supplementary Material. By comparing results based on a few selected features (OTU-S3), we observe reliable performance for both disease and yield prediction. This underscores the robustness of OTU-S3 as an effective feature selection strategy. The RF model was implemented and tested on a MacBook Pro with an Apple M1 Pro chip and 16 GB of RAM. The running time for the RF model is provided in Table in the supplementary materials. Depending on the number of trees and features selected, the RF model typically requires a few minutes per model on a dataset of our size (~200 samples). For the Bayesian NN model, the computational demands are significantly higher due to the need for probabilistic inference. We ran the Bayesian NN model on the Center for High Throughput Computation (CHTC) platform at UW-Madison with RTX2080ti graph cards. Despite using more computation resources, the Bayesian NN model would take between 24-72 h to compute the result for the models on all but Phylum level. Thus, it will be infeasible to run the Bayesian NN model on this particular problem on any personal devices. The runtime log of Bayesian NN models has been lost, unfortunately, as the CHTC platform only keeps the log files for 6 months, while the model was run more than 2 years ago. In summary, while the RF model can be efficiently run on a standard laptop such as a MacBook Pro, the Bayesian NN model requires significantly more computational resources. We recommend high-performance computing resources for readers planning to implement Bayesian NNs, particularly for larger datasets or more complex models. The following points summarize the main findings and their significance in advancing our understanding of soil microbiome and plant health: Can machines classify what humans cannot? The importance of accurate labels. The prediction power of microbiome data varies depending on the outcome we want to predict. For example, among all models that predict diseases, models for the pitted scab disease receive very high weighted F1 scores compared to other diseases. We further confirm this predictive power by comparing the performance of models trained on the real microbiome data to models trained on randomly generated data (see Figs. and ). Given that the prediction of the pitted scab disease is far from random, we can confidently conclude that this disease can be accurately predicted from microbiome data. It is noteworthy, however, that pitted scab disease is precisely one of the diseases that are easier to be visually detected, and thus, there is a reliable separation among the two classes (diseased and non-diseased) which is aiding in prediction by ML models. Other diseases, and more so yield, do not have such clear distinction between classes which results in lower predictive power. That is, we believe that the lack of prediction accuracy in yield is not driven by a lack of a biological connection between soil microbiome and yield, but on the lack of accurate labels that distinguish the two classes (e.g. low and high yield). If humans cannot distinguish what is low yield vs high yield, then that ambivalence will propagate into the ML classification. This conclusion seems to be confirmed when we notice that yield cannot be accurately predicted by any of 14 models in consideration (Table ). We conclude that one of the main challenges when applying ML methods in biological applications is the artificial binarization of phenotypes. We acknowledge that binarization leads to information loss; however, in this study, we found that it improved model performance in some cases. Continuous modeling results indicated that predicting Yield_Plant is challenging due to individual plant variability, whereas Yield_Meter showed better performance as a more stable measure. In future work, we aim to explore larger datasets and additional environmental variables to enhance model accuracy without relying on binarization. More work is needed to improve the performance of regression models that can predict continuous phenotypes when faced with limited sample sizes that are common in biological domains. Human analytical choices: Data preprocessing has a substantial impact in prediction performance. We demonstrate that the choice of normalization methods for microbiome datasets profoundly impacts prediction outcomes. Regrettably, our analysis did not reveal a discernible pattern indicating the superiority of one normalization method over others. We recommend domain scientists to explore various normalization methods for their data before utilizing them for prediction purposes. For further performance comparisons on normalization types, see also . Upon comprehensive analysis of the FMS results, we offer recommendations for normalization methods for future investigations. For RF model, normalization methods have a lesser impact on results overall, with greater emphasis placed on other factors such as data augmentation and taxonomic levels. However, certain normalization methods are recommended against, as indicated in Table . In contrast, recommendations for the Bayesian NN model differ. Please refer to Table for specific suggestions on normalization methods. In summary, four normalization methods consistently demonstrated strong performance across both disease and yield tasks for both RF and Bayesian NN: NM1 (TSS+none), NM4 (TSS+bayesMult), NM13 (rarefy+none), and NM16 (rarefy+bayesMult). Feature selection effectively preserves the predictive signal on lower dimensions. In terms of feature selection, we considered different strategies to select important features (ML, network comparison, and intersection of both). There is a significant overlap between important OTUs by two methods and the inclusion of this subset of predictors allowed us to build less complex models with comparable good performance (see, for example, Fig. for pitted scab disease). We did not observe one feature selection strategy that outperformed the others. For example, the weighted F1 scores obtained when using the OTUs identified as important by the ML methods are comparable to the weighted F1 scores when including OTUs identified as important by network comparison. It is worth mentioning, however, that the performance is also comparable to that obtained when including all OTUs which shows that the feature selection strategies work at preserving the OTUs that have predictive signals while simultaneously allowing computationally expensive models (like Bayesian NN) to be applied. In addition to showing that both RF and Bayesian NN methods perform well in feature selection, we provide a list of important taxa identified through two different strategies in https://github.com/solislemuslab/soil-microbiome-nn/blob/master/python-code/important_features_score.xlsx . Our analysis shows that some of the microbial taxa might be important, such as Paenibacillaceae , Moraxellaceae , and Syntrophaceae , as they consistently proved to be key predictors in model performance. Those taxa may also be related to nutrient cycling, plant growth, and disease suppression . Important taxa with strong prediction power for yield and disease can play a significant role in sustainable agriculture, where maintaining healthy soil microbiomes is a key objective. These taxa can also inform future research on biofertilizer development and microbial inoculants in soil microbiome engineering. Our study provided a robust framework for identifying key microbial taxa using multiple classification methods, including RF and Bayesian NN. Finer taxonomic levels provide higher prediction power for Random Forest and not much for Bayesian NN. While it is intuitive to expect that finer taxonomic levels would provide more predictive power, our analysis reveals this to be the case only in specific instances, such as the OTU-S3 model in RF and the OTU-S1 model in Bayesian NN, where the Genus level outperforms others (Fig. ). However, this expectation does not hold true across all scenarios. This may be due to not having enough samples, which limits the predictive power at finer taxonomic levels. For a comprehensive assessment, we utilize the FMS method to identify optimal combinations of normalization, zero replacement, feature selection, and model choices for maximizing prediction accuracy in microbiome data analysis. Contrary to common expectations, our findings do not support an overall superiority of certain taxonomic levels over others. Instead, the FMS model provides nuanced recommendations: for RF, optimal results are achieved by employing data augmentation and focusing on more specific taxonomic levels such as Family and Genus, while for Bayesian NN, utilizing more general taxonomic levels like Phylum without augmentation proves advantageous. Limited predictive power in soil microbiome compared to environment. When including environmental features such as soil physicochemical properties and microbial population density of soil in the model, we achieved higher weighted F1 score values. For pitted scab disease ( Scabpit ), utilizing alpha diversity with the RF model yields a median weighted F1 score of approximately 0.75. When combined with soil population density information, this score increases to 0.85. Similarly, for the Bayesian NN model, the score improves from 0.6 to 0.8 with the addition of soil population density information. For RF, employing OTU3 results in a median weighted F1 score of about 0.8, which further increases to 0.9 when supplemented with soil population density data. Similarly, for Bayesian NN, the median score reaches approximately 0.9. However, incorporating the population density of soil information leads to a narrower range for the box plot. In general, we investigated 14 different models for yield and disease prediction with different combinations of microbiome and environmental data. Results show poor performance in predicting yield across different models. The best results for pitted scab are achieved by combining alpha diversity and soil chemistry (Alpha+Soil) for RF and important OTUs and soil data (OTU-S3+Soil) for Bayesian NN. The median weighted F1 scores for predicting diseases range from 0.8 to 0.9 for RF and from 0.6 to 0.9 for Bayesian NN models (refer to Fig. ). Although the best-performing models include microbiome predictors (Alpha and OTU-S3), it is important to note that the models without microbiome data are comparably powerful as those including microbiome data. Specifically, for the RF method, the F1 score exceeds 0.75 for pitted scabe disease ( Scabpit ) and for black scurf, and surpasses 0.6 for scab and superficial scab ( Scabsuper ) diseases. While the models trained with microbiome data alone show that microbiome can effectively predict pitted scab disease, the fact that models without this type of data continue to perform well provides evidence that microbiome data may not be necessary to achieve reasonable prediction. In fact, when the collection of microbiome data requires much higher cost investment, the increased prediction accuracy by including microbiome predictors may not be enough to justify the extra cost. Cost Analysis and Practical Applications: The public price for soil chemistry analysis from a commercial lab can be $9.75 per sample. Adding the costs for microbial population density analysis, which is $20 per sample, brings the total cost of environmental data used in our analysis to approximately $29.75 per sample. Microbiome data collection, performed at the University of Wisconsin-Madison Biotechnology Center can cost $32 per sample. The goal of this research is to provide yield and disease predictions before planting and guide management decisions that would occur during the growing season. The value of improved prediction accuracy depends on how this tool is used. For instance, fields predicted to have high disease pressure would suggest the need for disease-control methods, such as fumigation or fungicide application. In this case, predictions with a certain threshold of accuracy (e.g.,>5%) would provide sufficient information to make disease-control decisions. Corresponding management actions would reduce disease risks, and the benefits would depend on factors such as the predicted disease level, potato cultivar, and market price. In another scenario, yield prediction can assist with selecting potato cultivars that align with the producer’s goals. The ultimate goal of our research is to help create a decision-making platform where growers can choose between various management options and perform benefit and risk analyses. Growers often aim to produce multiple potato cultivars for various markets, and soil microbiome composition and disease pressure often vary significantly among potato fields on the same farm. By selecting the most suitable cultivars based on the field-specific conditions, growers can optimize production and environmental outcomes, and meet the goal of precision agriculture. Inconsistent Predictive Performance Across Outcomes: One of the key challenges identified in this study is the inconsistency in predictive performance across different outcomes, particularly between disease and yield. While our models demonstrate robust predictive power for diseases such as pitted scab and superficial scab, the results for yield outcomes were notably weaker. This discrepancy likely stems from the complex and multifactorial nature of yield responses, which are influenced by environmental factors, soil properties, farming practices, and other variables not fully captured in our current dataset. Moreover, disease outcomes such as pitted scab tend to have clearer biological indicators, making them easier to model, whereas yield is a more complex and continuous trait, further complicating prediction. Our results also underscore that traits difficult for humans to detect, such as yield potential based on subtle microbial interactions, are also inherently challenging for machine learning models to predict. Furthermore, the limited sample size in this study may have compounded these issues, particularly for yield outcomes, where larger datasets are required to capture the full range of influential factors. Moving forward, we are collecting more samples from different years and locations, which will enhance the diversity and robustness of the dataset. This will enable the development of more generalized and accurate models capable of addressing the variability in agricultural outcomes. Our findings here lay the groundwork for further research, with the ultimate goal of creating models that are both reliable and broadly applicable to diverse agricultural datasets. Implications for disease management. Pitted scab is a severe form of potato common scab, a soil-borne disease that significantly reduces potato yield. The pitted scab is caused by the pathogenic Streptomyces spp, with symptoms of deep, dark lesions on the tuber surface. The disease is known to be sensitive to soil physicochemical properties including soil pH and moisture content . Soil microbial communities can also be related to the severity of the disease in the plants, as suppressive soils with unique microbiomes often cause the pathogen to fail to establish in the plants . The role of soil microbiome in influencing soil-borne disease has captured great research interests in both soil health and disease management. This study shows the close association of pitted scab with soil physicochemical properties when sampling across two different states. Although soil microbiome information contributed a small amount of prediction power to the prediction of this disease, the increased precision of prediction suggests the importance of soil microbiome in disease development. Particularly at finer spatial scales, soil microbiome may explain more of the disease variation among fields with similar physicochemical properties. Soil microbiome represents the most complex and least understood aspect of soil health. In this study, we use ML techniques such as random forest (RF) and Bayesian neural network (Bayesian NN) to determine whether soil microbial information has any predictive power for plant outcomes such as yield and disease. The RF method consistently demonstrates superior performance across all models, underscoring its effectiveness compared to the Bayesian NN method. In this study, Bayesian NN takes far longer to train due to weight sampling and approximation, especially on more refined taxonomic levels such as Order, Family, and Genus. Thus, we believe that based on the current dataset, RF is the preferred decision-making model with high prediction potential for disease outcomes when including a combination of microbiome and environmental predictors. Our results indicate that microbiome data helps predict potato disease, but the best accuracy comes from combining it with environmental data. Given that prediction with environmental factors alone was sufficiently powerful, it is uncertain whether the extra expense to sequence microbiome data is worth the cost. Future work. In our study, we initially attempted to predict the continuous response variable for yield using various machine learning techniques, including RF regression. However, the inherent ambiguity and subjectivity introduced by artificial labeling of the continuous data posed significant challenges, leading to unsatisfactory prediction accuracy . To overcome this limitation, we employed a binary classification approach, which proved to be more effective in capturing the underlying patterns and achieving better predictive performance. Although this approach involves some loss of granularity, it allowed us to focus on the broader trends and mitigate the impact of labeling subjectivity . In addition, we applied the principal component analysis (PCA) method. However, in our specific case, PCA did not yield significant improvements in terms of dimensionality reduction or feature extraction. We hypothesize that this could be due to the inherent complexity and nonlinear relationships present in our data, which may not be effectively captured by linear transformations like PCA . As part of future work, we aim to investigate advanced machine learning techniques that can handle the inherent ambiguity and subjectivity present in continuously labeled data. This could involve ensemble methods, deep learning approaches, or techniques specifically designed to handle noisy or subjective labels . Additionally, we plan to utilize the MiNAA package to detect alignment pairs of taxa in two constructed networks (healthy and diseased). By doing so, we can identify important taxa by comparing the two networks and confirming their significance using three different methods (ML-network-based models and applying Minaa). This work is based on data from one year, 2019. The goal is to find model traits that remain consistent year after year, making them applicable to other datasets. we plan to expand the dataset to include multiple growing seasons and diverse geographic locations, which will provide a more comprehensive representation of environmental and agronomic conditions. This expanded dataset will enable us to revisit continuous prediction methods with a larger and more diverse sample, aiming to improve accuracy by incorporating more sophisticated regression techniques and addressing variability across different regions. By leveraging a richer dataset, we hope to refine our models to capture the nuanced interactions influencing yield outcomes, ultimately developing more generalizable and robust predictive tools for precision agriculture. Supplementary file 1. |
Progress of mesenchymal stem cells affecting extracellular matrix metabolism in the treatment of female stress urinary incontinence | 1eea80a6-daa3-469a-aaf7-e0bc3f8dd692 | 11863768 | Surgical Procedures, Operative[mh] | Stress urinary incontinence (SUI) is a global health challenge predominantly affecting women, usually triggered by increased abdominal pressure from various activities (e.g., sneezing, coughing, and physical movements) . As the most prevalent form of urinary incontinence, recent data indicates that the overall prevalence of SUI among women in mainland China is 24.5% , with the risk of developing SUI increasing progressively with age, reaching a 50% prevalence among women over 40 . The quality of life for women is inversely correlated with the severity of their condition; the frequency and volume of incontinence, as well as the compulsory use of urinary pads in daily life, significantly impact women’s self-esteem and confidence . The median annual management cost for women with urinary incontinence at baseline approaches $500 , imposing a substantial financial burden on patients. Histologically, SUI is characterized by pelvic tissue dysfunction and alterations in connective tissue composition , with pregnancy and childbirth being primary causative factors due to damage to pelvic floor muscles, pubic nerves, and periurethral tissues . Preferred conservative treatments include pelvic muscle exercises, medications, and urethral fillers, though they often fail to achieve success . Surgical treatment is a therapeutic approach taken when conservative treatment is ineffective, including various types of slings and artificial urethral sphincter implantation. Among them, the retropubic mid-urethral sling (RMUS) and transforaminal mid-urethral sling (TMUS) have been measured to have the highest level of evidence for the feasibility and safety of treating SUI . Despite the substantial evidence level supporting their feasibility and safety, surgeries are frequently complicated by issues like poor healing, nerve trauma, bladder damage, and postoperative voiding disorders . In recent years, regenerative medicine, particularly cellular therapies, has advanced as a potential alternative to surgical interventions for treating urinary incontinence, potentially expanding the range of treatment options. Mesenchymal stem cells (MSCs) are a prevalent type of stem cell utilized in the treatment of SUI. These multipotent stem cells are derived from various tissues, including bone marrow , peripheral fat , muscular tissue , umbilical cord , dental pulp , and urine . In clinical applications, MSCs are predominantly harvested from sources such as bone marrow, umbilical cord blood, and adipose tissue . These MSCs have shown significant efficacy in the treatment of SUI , by modulating immune functions and facilitating the repair of damaged tissues through both differentiation and paracrine effects . Paracrine mechanisms, primarily involving the secretion of growth factors (GFs), cytokines, chemokines, and extracellular vesicles (EVs), are believed to be central to the reparative actions of MSCs . These vesicles aid in regulating immunity, promoting cell proliferation, angiogenesis, and differentiation by transferring proteins, mRNAs, and microRNAs (miRs) to target cells . EVs exhibit a wide range of diversity, with the three most extensively studied categories being exosomes(Exos), microvesicles, and apoptotic bodies(ABs) . Among these, Exos are the smallest, typically measuring 30–150 nm in diameter . MSC treatments have demonstrated safe and effective outcomes in phase I and II clinical trials , with minimal ethical concerns, promising a substantial therapeutic future in SUI management. Inadequate urethral closure due to impaired pelvic floor support structures is the main etiology of SUI , with alterations in the ECM serving as a significant pathogenic mechanism . The ECM constitutes a complex, three-dimensional network surrounding cells that facilitates intercellular biosignal transmission and regulates cell proliferation, differentiation, and migration, among other functions . In pathological conditions, extensive remodeling of the ECM is a crucial driver of disease progression . Stem cell therapy has been shown to stimulate ECM remodeling in urethral injuries in SUI, thereby improving urinary incontinence . However, the molecular mechanisms by which MSCs regulate the ECM of pelvic floor tissues to enhance SUI treatment remain unexplored. In this context, we review and propose a prospective study on how MSCs influence ECM metabolism in the treatment of female SUI.
MSCs are a class of pluripotent, differentiable stromal cells with therapeutic roles that include proliferation, differentiation, pluripotency, homing/migration, nutrition, and immunomodulation . These functions vary according to the stem cell type and the extracellular environment. In 2006, MSCs were first proposed as mediators of therapy through their secretory effects , involving GFs, cytokines, chemokines, and EVs as paracrine mediators . Recent preclinical studies have confirmed that MSC treatments primarily operate through these mediators , addressing diseases such as cirrhosis, psoriasis, and pulmonary fibrosis. One study on the role of adipose-derived stem cells (ADSCs) in heart valve tissue engineering found that ADSCs secrete ECM components like collagen and elastin , marking a significant advancement in treating ECM-related diseases. MSCs modulate ECM metabolism in various diseases, including fibrosis , cancer , wound healing , neuroinflammation , and SUI , suggesting that MSCs not only differentiate but also regulate ECM through secreted factors, providing a new direction for female SUI treatment. It has been shown that stem cells facilitate SUI recovery by modulating ECM metabolism, primarily enhancing urethral function, with the ECM playing a pivotal role in maintaining the integrity and functionality of pelvic floor support structures. Collagen and elastin are crucial components of the ECM. Collagen, abundant and diverse, is predominantly found in pelvic floor support tissues like the vaginal wall and fascia, with types I, III, and to a lesser extent, V being most prevalent. Type I collagen plays a role in regulating cellular activity and promoting growth, contributing to tissue support and strength, while type III collagen is linked to tissue elasticity, and type V’s role remains poorly understood . During tissue healing, collagen III is prevalent in the initial stages of ECM formation, transitioning to collagen I during the maturation and remodeling phases . Preclinical studies have shown a significant reduction in collagen types I and III in the anterior vaginal wall of SUI rats, whereas increased pelvic floor collagen has alleviated urinary incontinence . Elastin, the primary component of tissue elastic fibers, facilitates contraction and stretching, playing a vital role in maintaining normal pelvic floor function . Moreover, elastin is an essential element of pelvic floor connective tissue, with abnormalities associated with dysfunction, particularly in SUI and pelvic prolapse . Elastin fiber dysfunction leads to a loss of elasticity, impeding urethral sphincter contraction and disrupting the urethral closure mechanism, leading to SUI . In treating SUI, MSCs can act directly or influence other cells, such as fibroblasts, to remodel the ECM of pelvic floor support structures (Fig. ). and regulate ECM metabolism signaling pathways.
The normal urethral sphincter, intact anterior vaginal wall, and periurethral connective tissue are crucial pelvic floor support structures that regulate urethral opening and closing in response to abdominal pressure changes , along with the anorectal muscle, which also plays a supportive role . The urethra is mainly anchored by the anterior vaginal wall, enriched with a dense ECM produced by fibroblast regulation . In addition, connective tissue contains essential pelvic floor support components, with collagen and elastin fibers being crucial elements of the matrix . MSCs’ remodeling of the ECM in pelvic floor support structures in women with SUI primarily focuses on the connective tissue surrounding the urethral sphincter and the anterior vaginal wall, predominantly enhancing incontinence management by regulating the collagen and elastin levels that maintain tissue stability and relaxation. (Table ). ECM around the urethral sphincter The urethral sphincter, a vital part of the urethral support structure, comprises the internal urethral sphincter (IUS) and external urethral sphincter (EUS). These structures are intrinsic to urethral closure and typically function in conjunction with secondary exogenous closure structures (such as the anal retractor and puborectalis muscles) . Damage to the urethral sphincter or weakening of support structures can cause the urethra to fail to close under increased intra-abdominal pressure, leading to urine leakage . The balance of forces between elastic and collagen fibers is critical for maintaining normal urethral sphincter function, with abnormal collagen remodeling of periurethral tissues and a loss of functional elastic fiber network observed in a rat model of SUI . Therefore, the degradation of collagen and elastin in periurethral tissues is closely linked to the development of urinary incontinence. MSCs are capable of repairing the damaged urethral sphincter and performing injectable treatments that differentiate into smooth muscle cells for direct repair , but also indirectly affect the urethral sphincter and peripheral ECM through paracrine effects . As early as 2010, Lin et al. administered urethral injections of ADSCs using a vaginal balloon dilatation and bilateral ovariectomy-induced SUI model in postpartum rats. Histological analysis revealed significantly higher elastin levels in the treated group compared to the model group, suggesting that ADSCs mediate tissue recovery, including elastin, through cytokines. Subsequently, Dissaranan et al. performed vaginal dilatation (VD) in sprague-dawley (SD) female rats to simulate birth injury, injecting the model rats intravenously with autologous bone marrow mesenchymal stem cells (BMSCs). Measurements indicated that, compared to the control group, VD rats exhibited an increase in elastin and fibers in the peripheral matrix of the EUS, along with a significant improvement in leak point pressure (LPP). Although there was no significant change in external urethral sphincter electromyography (EUS EMG), MSCs demonstrated a remodeling effect in the ECM through paracrine secretion, potentially enhancing EUS function. It is important to note that no specific paracrine factors were identified in this experiment. The following year, Deng et al. developed a similar rat model of birth injury induced by dual vaginal dilatation plus pudendal nerve crush (VD + PNC) injury, treating it with BMSCs via tail vein injection, and reported similar results. Additionally, fewer animal experiments have utilized myogenic stem cells (MDSCs); however, Bilhar et al. explored MDSCs effects on urethral recovery in female SUI model rats. After MDSC treatment, observed were neogenesis, urethral muscle regeneration, and a significant increase in collagen type I alpha 1 (Co1α1) and collagen type III alpha 1 (Col3α1) gene expression, indicating sustained recovery of urethral tissue. This underscores the significant potential of MSCs and their secretions for ECM repair in the treatment and prevention of SUI. All aforementioned studies focused on the effect of a single MSC dose in a rat SUI model. Janssen et al. explored the impact of multiple MSC doses on maintaining urethral function, employing a VD + PNC double injury SD female rat model. This injury model revealed thinner and more susceptible elastic fibers in the urethral sphincter, particularly in the internal urethral sphincter (IUS). Treatment with single and multiple doses of MSCs significantly improved elastin production and fiber thickening, and somewhat prevented disruption of the EUS. Although the three-dose treatment groups showed significant improvement in LPP relative to the control group, differences among these groups were minimal, likely due to multiple comparisons and the small sample size. Notably, peak bladder pressure was significantly higher in the maximum dose MSC-treated group compared to others, offering another metric for assessing urethral function. This study suggests that MSC treatment enhances urethral integrity by increasing periurethral elastin and restoring neuromuscular function, with higher doses amplifying this effect. In recent years, BMSC-derived small extracellular vesicles (sEV) have been reported to increase the synthesis of ECM proteins such as elastin, collagen I, and collagen III around the urethral sphincter, enhancing the length and thickness of elastic and collagen fibers . sEV, also known as Exos. In a novel study, the combinatorial therapy of MSCs overexpressing chemokine receptor 1 (CCR1) and the chemokine (C-C motif) Ligand 7 (CCL7) was found to more effectively promote collagen synthesis and muscle fiber thickening around the EUS . Utilizing a VD + PNC rat model, these MSCs were genetically modified to overexpress CCR1 and administered via tail vein injection prior to treatment with BMSCs, followed by the periurethral injection of CCL7. Recently, Paz et al. extracted decidua mesenchymal stem cells (DMSCs) from the maternal layer of the human placenta, known as the decidua. These DMSCs were injected periurethrally into VD-treated rat models, and histological analysis revealed an increase in elastic fibers around the EUS, coincident with improved urinary incontinence. Furthermore, they isolated myofibroblasts from the suburethral tissue of SUI patients and observed a significant reduction in senescence-associated secretory phenotype (SASP) components, such as monocyte chemoattractant protein (MCP-1) and MCP-3, post-treatment, suggesting that the pathogenesis of SUI may also involve the senescence of pelvic tissue cells. These studies have investigated cell-free therapies, combination therapies, and stem cells from various tissue sources for SUI treatment, enhancing EUS function through regenerative processes and likely by modulating collagen and elastin synthesis. However, differences between animal models and human diseases necessitate further investigation into MSC treatments in humans. ECM in connective tissue The connective tissue surrounding the urethra and vaginal wall is a critical component of the urethral support structure, assuming both supportive and connective roles . Rich in ECM, connective tissue maintains functional integrity as a physical scaffolding for cells, tissues, and organs, significantly contributing to the biomechanical properties of tissues, crucial for elasticity and strength . Impairment of functional ECM in connective tissue is a key change in the pathophysiology of SUI. Altered ECM metabolism in periurethral connective tissue has been reported in patients with SUI . Human umbilical cord mesenchymal stem cells (huc-MSCs) have been found to enhance SUI by augmenting urethral connective tissue , possibly through paracrine and anti-inflammatory effects. Some researchers have observed that in vivo injection of MSCs into rats with periurethral connective tissue hemorrhage following VD restored both urethral and systemic connective tissue and vascularity . This effect may extend to the vaginal wall organization, as Janssen et al. noted that BMSCs improved the biomechanical properties of the vaginal wall, possibly through ECM remodeling. After VD + PNC, while elastin in the vaginal wall was reduced and collagen deposition increased, intravenous MSC administration resulted in an increase in elastin fiber density and a return to normal collagen levels, accompanied by a recovery of vaginal fibrosis and improved urethral function. Increased expression of matrix metalloproteinase-9 (MMP-9) following vaginal injury degrades ECM components, but MSCs can reverse MMP-9 expression and reduce ECM loss, significantly benefiting the pelvic floor support structures and presenting excellent potential for restoring SUI caused by pelvic floor disorders.
The urethral sphincter, a vital part of the urethral support structure, comprises the internal urethral sphincter (IUS) and external urethral sphincter (EUS). These structures are intrinsic to urethral closure and typically function in conjunction with secondary exogenous closure structures (such as the anal retractor and puborectalis muscles) . Damage to the urethral sphincter or weakening of support structures can cause the urethra to fail to close under increased intra-abdominal pressure, leading to urine leakage . The balance of forces between elastic and collagen fibers is critical for maintaining normal urethral sphincter function, with abnormal collagen remodeling of periurethral tissues and a loss of functional elastic fiber network observed in a rat model of SUI . Therefore, the degradation of collagen and elastin in periurethral tissues is closely linked to the development of urinary incontinence. MSCs are capable of repairing the damaged urethral sphincter and performing injectable treatments that differentiate into smooth muscle cells for direct repair , but also indirectly affect the urethral sphincter and peripheral ECM through paracrine effects . As early as 2010, Lin et al. administered urethral injections of ADSCs using a vaginal balloon dilatation and bilateral ovariectomy-induced SUI model in postpartum rats. Histological analysis revealed significantly higher elastin levels in the treated group compared to the model group, suggesting that ADSCs mediate tissue recovery, including elastin, through cytokines. Subsequently, Dissaranan et al. performed vaginal dilatation (VD) in sprague-dawley (SD) female rats to simulate birth injury, injecting the model rats intravenously with autologous bone marrow mesenchymal stem cells (BMSCs). Measurements indicated that, compared to the control group, VD rats exhibited an increase in elastin and fibers in the peripheral matrix of the EUS, along with a significant improvement in leak point pressure (LPP). Although there was no significant change in external urethral sphincter electromyography (EUS EMG), MSCs demonstrated a remodeling effect in the ECM through paracrine secretion, potentially enhancing EUS function. It is important to note that no specific paracrine factors were identified in this experiment. The following year, Deng et al. developed a similar rat model of birth injury induced by dual vaginal dilatation plus pudendal nerve crush (VD + PNC) injury, treating it with BMSCs via tail vein injection, and reported similar results. Additionally, fewer animal experiments have utilized myogenic stem cells (MDSCs); however, Bilhar et al. explored MDSCs effects on urethral recovery in female SUI model rats. After MDSC treatment, observed were neogenesis, urethral muscle regeneration, and a significant increase in collagen type I alpha 1 (Co1α1) and collagen type III alpha 1 (Col3α1) gene expression, indicating sustained recovery of urethral tissue. This underscores the significant potential of MSCs and their secretions for ECM repair in the treatment and prevention of SUI. All aforementioned studies focused on the effect of a single MSC dose in a rat SUI model. Janssen et al. explored the impact of multiple MSC doses on maintaining urethral function, employing a VD + PNC double injury SD female rat model. This injury model revealed thinner and more susceptible elastic fibers in the urethral sphincter, particularly in the internal urethral sphincter (IUS). Treatment with single and multiple doses of MSCs significantly improved elastin production and fiber thickening, and somewhat prevented disruption of the EUS. Although the three-dose treatment groups showed significant improvement in LPP relative to the control group, differences among these groups were minimal, likely due to multiple comparisons and the small sample size. Notably, peak bladder pressure was significantly higher in the maximum dose MSC-treated group compared to others, offering another metric for assessing urethral function. This study suggests that MSC treatment enhances urethral integrity by increasing periurethral elastin and restoring neuromuscular function, with higher doses amplifying this effect. In recent years, BMSC-derived small extracellular vesicles (sEV) have been reported to increase the synthesis of ECM proteins such as elastin, collagen I, and collagen III around the urethral sphincter, enhancing the length and thickness of elastic and collagen fibers . sEV, also known as Exos. In a novel study, the combinatorial therapy of MSCs overexpressing chemokine receptor 1 (CCR1) and the chemokine (C-C motif) Ligand 7 (CCL7) was found to more effectively promote collagen synthesis and muscle fiber thickening around the EUS . Utilizing a VD + PNC rat model, these MSCs were genetically modified to overexpress CCR1 and administered via tail vein injection prior to treatment with BMSCs, followed by the periurethral injection of CCL7. Recently, Paz et al. extracted decidua mesenchymal stem cells (DMSCs) from the maternal layer of the human placenta, known as the decidua. These DMSCs were injected periurethrally into VD-treated rat models, and histological analysis revealed an increase in elastic fibers around the EUS, coincident with improved urinary incontinence. Furthermore, they isolated myofibroblasts from the suburethral tissue of SUI patients and observed a significant reduction in senescence-associated secretory phenotype (SASP) components, such as monocyte chemoattractant protein (MCP-1) and MCP-3, post-treatment, suggesting that the pathogenesis of SUI may also involve the senescence of pelvic tissue cells. These studies have investigated cell-free therapies, combination therapies, and stem cells from various tissue sources for SUI treatment, enhancing EUS function through regenerative processes and likely by modulating collagen and elastin synthesis. However, differences between animal models and human diseases necessitate further investigation into MSC treatments in humans.
The connective tissue surrounding the urethra and vaginal wall is a critical component of the urethral support structure, assuming both supportive and connective roles . Rich in ECM, connective tissue maintains functional integrity as a physical scaffolding for cells, tissues, and organs, significantly contributing to the biomechanical properties of tissues, crucial for elasticity and strength . Impairment of functional ECM in connective tissue is a key change in the pathophysiology of SUI. Altered ECM metabolism in periurethral connective tissue has been reported in patients with SUI . Human umbilical cord mesenchymal stem cells (huc-MSCs) have been found to enhance SUI by augmenting urethral connective tissue , possibly through paracrine and anti-inflammatory effects. Some researchers have observed that in vivo injection of MSCs into rats with periurethral connective tissue hemorrhage following VD restored both urethral and systemic connective tissue and vascularity . This effect may extend to the vaginal wall organization, as Janssen et al. noted that BMSCs improved the biomechanical properties of the vaginal wall, possibly through ECM remodeling. After VD + PNC, while elastin in the vaginal wall was reduced and collagen deposition increased, intravenous MSC administration resulted in an increase in elastin fiber density and a return to normal collagen levels, accompanied by a recovery of vaginal fibrosis and improved urethral function. Increased expression of matrix metalloproteinase-9 (MMP-9) following vaginal injury degrades ECM components, but MSCs can reverse MMP-9 expression and reduce ECM loss, significantly benefiting the pelvic floor support structures and presenting excellent potential for restoring SUI caused by pelvic floor disorders.
The relationship between ECM formation and fibroblasts is well-established, highlighting mechanisms associated with MSC treatment of SUI involving fibroblast-driven ECM metabolism.(Fig. ). Fibroblasts, the primary cellular component of connective tissue, secrete collagen, elastin, and glycoproteins, which are integral to ECM composition and remodeling . Their dysfunction is linked to the development of SUI. Excessive mechanical stress, such as childbirth, damages fibroblasts, increasing intracellular levels of reactive oxygen species (ROS) and apoptosis . It has been reported that female vaginal fibroblasts can secrete sEV, and increased secretion of fibroblast-sEV during SUI impairs the normal function of fibroblasts to express collagen. Differential analyses suggest this may be related to tissue inhibitor of metalloproteinase-2 (TIMP-2), transforming growth factor-beta (TGF-β), and ATP-binding cassette subfamily C member 4 (ABCC4), among other mechanisms. Regulation of ECM by fibroblasts has also been associated with non-coding RNAs such as miR-34a, miR-93, and miR-328a-3p . miR-34a down-regulates nicotinamide phosphoribosyltransferase (Nampt) expression, miR-93 suppresses coagulation factor III (F3) and calpain-2 expression, and miR-328a-3p down-regulates sirtuin7(SIRT7) expression, reductions in these substances could promote collagen synthesis to regulate ECM remodeling. Non-coding RNAs, including miRs, negatively regulate gene expression at the transcriptional level by binding to their target RNAs . MSCs are closely associated with fibroblasts, able both to differentiate into fibroblasts for repairing damaged connective tissues after injury and to regulate fibroblasts and thus ECM remodeling through the secretion of EVs . MSCs-derived Exos were shown in skin healing studies to be taken up by fibroblasts, stimulating cell migration, proliferation, and collagen synthesis . In SUI treatment, this mechanism was demonstrated by Jiang et al. , who used BMSCs-conditioned medium (CM) to culture vaginally isolated adventitial fibroblasts, treated with increased proliferation, migration rate, and production of collagens I and III, facilitating recovery in rats with simulated SUI after VD. Liu et al. assessed the effect of ADSCs-Exos on collagen metabolism in cultured fibroblasts from women with SUI. They isolated fibroblasts from periurethral vaginal wall tissues of women with SUI who had no severe pelvic disease or prior pelvic surgery, noting significantly reduced collagen levels compared to controls. After treating these fibroblasts with ADSC-Exos medium for 6 h, they found down-regulation of MMP-1 and MMP-2, up-regulation of tissue inhibitor of TIMP-1 and TIMP-3, and significantly higher collagen I content than in the phosphate buffer-treated group. Wang et al. demonstrated that ADSCs-EVs could regulate fibroblasts to contribute to pelvic floor tissue ECM remodeling in SUI. They explored the specific mechanism of the miR-93/F3 axis involved in SUI in ADSCs-EVs: after ADSCs-EVs carrying miR-93 were engulfed by fibroblasts, the expression of the target gene F3 was suppressed. In another similar study , miR-328a-3p was found to downregulate SIRT7 in human primary fibroblasts and SUI rats, promoting elastin and collagen I production. The downregulation of F3 and SIRT7 in fibroblasts in both studies promoted ECM metabolism.
The potential mechanisms through which MSCs influence ECM metabolism are extensive and form a complex regulatory network, with some key signaling pathways now well-defined. This study details the cytokine networks involved in ECM production by MSC-activated cells and several major signaling pathways that may contribute to ECM formation, either independently or interactively.(Fig. ). TGF-β/SMAD signaling pathway The TGF-β superfamily consists of structurally related cytokines that are crucial within cells. The TGF-β/SMAD pathway is the principal route regulating collagen synthesis in fibroblasts and plays a vital role in activating fibroblasts that promote ECM synthesis in skin and other organ tissues . This pathway is intimately linked with ECM gene expression and fibrosis, encompassing heart, liver, kidney, lung, and skin fibrosis , and extends to ECM-associated disorders such as SUI, urge incontinence, and pelvic prolapse . TGF-β, a pleiotropic cytokine, regulates cellular behavior and plasticity across various tissues. Its myriad cellular responses are primarily mediated through the classical SMAD signaling pathway, although it also employs non-classical pathways like PI3K/AKT to regulate collagen and ECM homeostasis efficiently . TGF-β1/SMAD is pivotal in ECM metabolism, implicated in the pathogenesis of mechanical injury-induced SUI, though the specific effects and underlying mechanisms remain poorly defined . Classically, TGF-β activates the type I receptor (TGFβRI), which phosphorylates SMAD2/3; these then bind to SMAD4 and translocate into the nucleus to influence gene expression, a process inhibited by SMAD7 . Zhang et al. explored the specific impact of BMSCs-derived sEV on the TGF-β1 signaling pathway in urothelial function and ECM remodeling both in vivo and ex vivo. sEV are readily absorbed and internalized by fibroblasts and tissues surrounding the urethra. They discovered that miR-328a-3p carried by BMSCs-sEV is a crucial upstream regulator of ECM, antagonizing the expression of SIRT7. SIRT7 is a nucleolar-dependent deacetylase that opposes TGF-β1 signaling and total SMAD2/3 protein phosphorylation, thus regulating ECM . Therefore, the remodeling of the urethral sphincter ECM by BMSCs is facilitated by the secretion of sEV containing miR-328a-3p, which is internalized by periurethral tissue cells and down-regulates SIRT7 to enhance ECM secretion. This mechanism improves the ECM of the damaged urethral sphincter and represents a novel approach to treating SUI. In recent years, efforts have increased to enhance SUI management by modulating the ECM through TGF-β/SMAD signaling. Liu et al. discovered that dimethyl fumarate (DMF) upregulated nuclear factor erythroid 2-related factor 2 (Nrf2) levels, activating the TGF-β1/SMAD3 pathway to regulate collagen and elastin in the pelvic floor, thereby improving incontinence. Li et al. found that puerarin might serve as a therapeutic agent for SUI, with its mechanism involving the regulation of collagen metabolism in mouse L929 fibroblasts by promoting the Nrf 2/TGF-β1 signaling pathway while also protecting fibroblasts from mechanical traction injury. Nrf2, a critical antioxidant gene inducer, can be modulated by huc-MSCs in some cells in vivo via EVs . Therefore, Nrf2 emerges as a potential therapeutic target for MSCs in SUI treatment to mitigate fibrosis and inflammation. Future studies may reveal MSCs’ ability to reduce ECM damage in pelvic floor tissues through activation of Nrf2/TGF-β/SMAD, aiming to ameliorate or prevent SUI. JAK/STAT signaling pathway The janus kinase/signal transducer and activator of transcription (JAK/STAT) signaling pathway, while relatively straightforward, is crucial for cell proliferation, differentiation, apoptosis, and immune response. The classical STAT pathway involves cytokine binding to its receptor, leading to tyrosine phosphorylation and activation of receptor-associated JAK, followed by phosphorylation and activation of STAT. Once activated, tyrosine-phosphorylated STAT (p-STAT) forms a dimer that translocates to the nucleus, binds to target DNA sequences, and regulates gene expression. In the non-classical pathway, unphosphorylated STAT also forms a dimer that enters the nucleus to regulate transcription . The JAK family includes JAK1, JAK2, JAK3, and Tyrosine kinase 2 (Tyk2), with seven identified STAT family members (STAT1, STAT2, STAT3, STAT4, STAT5a, STAT5b, STAT6) . Studies on JAK/STAT in SUI are limited. Jiang et al. found that BMSCs enhanced SUI treatment by promoting fibroblast proliferation, migration, and collagen production in the anterior vaginal wall of rats via the JAK2/STAT4 pathway. Utilizing the properties of BMSCs-secreted proteome, they cultured fibroblasts from the vaginal tissue of SUI model rats with BMSCs-CM, observing significantly higher levels of phosphorylated janus kinase 2 (p-JAK2) and p-STAT4 compared to controls. GFs and cytokines binding to membrane receptors in BMSCs-CM elevated p-JAK2 and p-STAT4 levels in fibroblasts, with JAK2 inhibitor treatment reversing these effects. This led to a notable increase in LPP, enhancing vaginal antrum fibroblast survival and collagen fiber regeneration in the treated group. In addition, a study on self-healing biomaterials discovered that a novel basic fibroblast growth factor (bFGF)/stromal cell-derived factor (SDF-1)/hydrogel cross-linking material effectively induced BMSCs homing and enhanced collagen production in vaginal tissues, with kyoto encyclopedia of genes and genomes (KEGG) enrichment analysis confirming that the JAK-STAT signaling pathway plays a key role associated with collagen production, though more specific mechanisms need further exploration. Therefore, MSCs play a broad role in treating diseases like SUI caused by ECM injury in pelvic floor tissues through regulation of the JAK/STAT signaling pathway. Wnt/β-catenin signaling pathway The Wnt/β-catenin signaling pathway is central to various biological processes, including cell proliferation, differentiation, apoptosis, stem cell self-renewal, tissue homeostasis, and wound healing. It is an evolutionarily conserved pathway, typically divided into β-catenin-dependent and independent types . Classical pathway transduction primarily involves extracellular Wnt protein binding to frizzled(Frz) protein and co-receptor low-density lipoprotein receptor-related protein 5/6 (LRP5/6). This triggers phosphorylation of disheveled protein (Dsh), transmitting signals into the cytoplasm to inhibit glycogen synthase kinase 3 β (GSK-3β) activation, leading to accumulation of intracellular cytoplasmic free β-catenin. Subsequently, β-catenin enters the nucleus and binds to lymphocyte enhancer factor/T cell factor (LEF/TCF) to promote downstream gene expression . Indeed, the Wnt pathway plays various roles in ECM-related diseases such as fibrosis but is considered a challenge for therapeutic targeting . Identifying potent and specific Wnt pathway inhibitors for the treatment of cancer and other diseases remains one of the most significant challenges for future research in this field . In recent years, the Wnt protein signaling pathway has been recognized for its key roles in various diseases, including tumors , fibrosis , pelvic prolapse , and SUI . Studies involving the ECM have shown that activation of β-catenin can be involved in both the synthesis and degradation of the ECM by secreted substances such as MMPs. In pelvic floor disorders, reduced β-catenin expression in vaginal fibroblasts may be associated with pelvic organ prolapse. One reason for this association is that a significant reduction in collagen I expression by vaginal fibroblasts was found in cases of pelvic organ prolapse, which was reversed with the use of Wnt/β-catenin activators . This partially reflects the link between Wnt/β-catenin and pelvic floor dysfunction. Fewer studies have shown that MSCs improve SUI by modulating Wnt/β-catenin to regulate ECM. In other diseases, some researchers have found that EVs secreted by BMSCs, like ABs, can regulate the Wnt/β-catenin pathway to ameliorate endometrial fibrosis, though the specific molecular mechanism has not been explored . Furthermore, in a study exploring chondrocyte damage in arthritis, it was found that MSC-derived EVs could load miR-3960 into cartilage tissue cells to downregulate pleckstrin homology-like domain family a member 2 (PHLDA2) proteins, thereby inhibiting the syndecan-1 (SDC-1)/ Wnt/β-catenin axis and reducing ECM degradation . In conclusion, based on the effect of β-catenin activation on collagen production in vaginal fibroblasts and its regulation by paracrine secretions from MSCs, we hypothesized that this signaling pathway may play a key role in the treatment of SUI in MSCs and may represent a promising therapeutic target. PI3K/AKT signaling pathway Akt, a serine/threonine kinase also known as protein kinase B (PKB), is activated by a variety of growth factors, including vascular endothelial growth factor (VEGF), fibroblast growth factor (FGF), insulin-like growth factor (IGF), and TGF-β. Upon activation, Akt phosphorylates numerous downstream effectors crucial for apoptosis, transcription, and carcinogenesis regulation . The most prevalent activation pathway for Akt is the phosphatidylinositol 3-kinase (PI3K)-dependent route, involving two subunits, the regulatory p85 and the catalytic p110, which together activate Akt. The mammalian target of rapamycin (mTOR) is a major downstream target of Akt, completing the primary nodes of this pathway . The PI3K/AKT pathway plays a significant role in ECM synthesis regulation. Recent findings suggest that silk fibroin (SF) can activate the integrin/PI3K/AKT signaling pathway in MSCs, either directly or indirectly, enhancing their paracrine function and promoting collagen production . Studies focusing on the PI3K/AKT pathway in MSCs for improving SUI are limited. It has been reported that MSCs can secrete insulin-like growth factor-1 (IGF-1) , which activates the Akt signaling pathway, accelerating pelvic floor nerve and tissue recovery in rat-induced SUI models . While IGF-1 is known to influence a variety of cellular processes, such as growth, motility, and differentiation, the impact of Akt activation by IGF-1 on pelvic floor ECM has not been thoroughly investigated. However, other studies have shown that IGF-1 can promote collagen synthesis in tendon tissues , though this effect is not direct but mediated through the co-regulation of the PI3K/Akt and ERK pathways . It remains to be determined whether MSCs can regulate pelvic floor collagen synthesis via the IGF-1/PI3K/AKT axis. Consequently, the PI3K/AKT signaling pathway could be a promising therapeutic target for treating SUI. ERK/MAPK signaling pathway The mitogen-activated protein kinase (MAPK) signaling pathway, commonly mediated by protein kinase-coupled receptors, is evolutionarily conserved and involves receptors like the epidermal growth factor receptor (EGFR), fibroblast growth factor receptor (FGFR), platelet-derived growth factor receptor (PDGFR), and vascular endothelial growth factor receptor (VEGFR) in a receptor tyrosine kinase (RTK) signaling cascade . The well-characterized MAPK family includes extracellular signal-regulated kinase 1/2 (ERK1/2), c-Jun amino-terminal kinase (JNK), p38, and ERK5. The ERK/MAPK pathway involves a multilayered kinase cascade activated by receptor-ligand binding, initiating the MAPK signaling pathway through Ras protein activation. This leads to the phosphorylation of Raf, MEK1/2, and ERK1/2 target proteins. Phosphorylated ERK1/2 activates various transcription factors to regulate gene expression and can also affect subcellular responses within the cytoplasm . The MAPK pathway significantly influences ECM synthesis, degradation, and remodeling, controlling ECM dynamic stability by regulating ECM-related gene expression, such as MMPs . In a study examining the relationship between ERK1/2 and SUI in periurethral support tissue fibroblasts, the use of an ERK kinase inhibitor was shown to influence the expression of mRNAs and proteins for collagen I and III in vaginal fibroblast cultures, suggesting that the ERK/MAPK pathway affects pelvic floor collagen tissue and may be implicated in SUI pathogenesis . Additionally, ADSCs were found to potentially improve SUI by activating ERK1/2 through vascular endothelial growth factor (VEGF) , which binds to receptor proteins and triggers a series of responses in target cells. Tissue analysis reveals an increase in collagen I/III ratios (enhancing urothelial tensile strength )and more dense and organized elastin. Therefore, MSCs may regulate the ERK/MAPK signaling pathway to affect ECM remodeling, offering a potential treatment for SUI and a key target for future therapeutic strategies.
The TGF-β superfamily consists of structurally related cytokines that are crucial within cells. The TGF-β/SMAD pathway is the principal route regulating collagen synthesis in fibroblasts and plays a vital role in activating fibroblasts that promote ECM synthesis in skin and other organ tissues . This pathway is intimately linked with ECM gene expression and fibrosis, encompassing heart, liver, kidney, lung, and skin fibrosis , and extends to ECM-associated disorders such as SUI, urge incontinence, and pelvic prolapse . TGF-β, a pleiotropic cytokine, regulates cellular behavior and plasticity across various tissues. Its myriad cellular responses are primarily mediated through the classical SMAD signaling pathway, although it also employs non-classical pathways like PI3K/AKT to regulate collagen and ECM homeostasis efficiently . TGF-β1/SMAD is pivotal in ECM metabolism, implicated in the pathogenesis of mechanical injury-induced SUI, though the specific effects and underlying mechanisms remain poorly defined . Classically, TGF-β activates the type I receptor (TGFβRI), which phosphorylates SMAD2/3; these then bind to SMAD4 and translocate into the nucleus to influence gene expression, a process inhibited by SMAD7 . Zhang et al. explored the specific impact of BMSCs-derived sEV on the TGF-β1 signaling pathway in urothelial function and ECM remodeling both in vivo and ex vivo. sEV are readily absorbed and internalized by fibroblasts and tissues surrounding the urethra. They discovered that miR-328a-3p carried by BMSCs-sEV is a crucial upstream regulator of ECM, antagonizing the expression of SIRT7. SIRT7 is a nucleolar-dependent deacetylase that opposes TGF-β1 signaling and total SMAD2/3 protein phosphorylation, thus regulating ECM . Therefore, the remodeling of the urethral sphincter ECM by BMSCs is facilitated by the secretion of sEV containing miR-328a-3p, which is internalized by periurethral tissue cells and down-regulates SIRT7 to enhance ECM secretion. This mechanism improves the ECM of the damaged urethral sphincter and represents a novel approach to treating SUI. In recent years, efforts have increased to enhance SUI management by modulating the ECM through TGF-β/SMAD signaling. Liu et al. discovered that dimethyl fumarate (DMF) upregulated nuclear factor erythroid 2-related factor 2 (Nrf2) levels, activating the TGF-β1/SMAD3 pathway to regulate collagen and elastin in the pelvic floor, thereby improving incontinence. Li et al. found that puerarin might serve as a therapeutic agent for SUI, with its mechanism involving the regulation of collagen metabolism in mouse L929 fibroblasts by promoting the Nrf 2/TGF-β1 signaling pathway while also protecting fibroblasts from mechanical traction injury. Nrf2, a critical antioxidant gene inducer, can be modulated by huc-MSCs in some cells in vivo via EVs . Therefore, Nrf2 emerges as a potential therapeutic target for MSCs in SUI treatment to mitigate fibrosis and inflammation. Future studies may reveal MSCs’ ability to reduce ECM damage in pelvic floor tissues through activation of Nrf2/TGF-β/SMAD, aiming to ameliorate or prevent SUI.
The janus kinase/signal transducer and activator of transcription (JAK/STAT) signaling pathway, while relatively straightforward, is crucial for cell proliferation, differentiation, apoptosis, and immune response. The classical STAT pathway involves cytokine binding to its receptor, leading to tyrosine phosphorylation and activation of receptor-associated JAK, followed by phosphorylation and activation of STAT. Once activated, tyrosine-phosphorylated STAT (p-STAT) forms a dimer that translocates to the nucleus, binds to target DNA sequences, and regulates gene expression. In the non-classical pathway, unphosphorylated STAT also forms a dimer that enters the nucleus to regulate transcription . The JAK family includes JAK1, JAK2, JAK3, and Tyrosine kinase 2 (Tyk2), with seven identified STAT family members (STAT1, STAT2, STAT3, STAT4, STAT5a, STAT5b, STAT6) . Studies on JAK/STAT in SUI are limited. Jiang et al. found that BMSCs enhanced SUI treatment by promoting fibroblast proliferation, migration, and collagen production in the anterior vaginal wall of rats via the JAK2/STAT4 pathway. Utilizing the properties of BMSCs-secreted proteome, they cultured fibroblasts from the vaginal tissue of SUI model rats with BMSCs-CM, observing significantly higher levels of phosphorylated janus kinase 2 (p-JAK2) and p-STAT4 compared to controls. GFs and cytokines binding to membrane receptors in BMSCs-CM elevated p-JAK2 and p-STAT4 levels in fibroblasts, with JAK2 inhibitor treatment reversing these effects. This led to a notable increase in LPP, enhancing vaginal antrum fibroblast survival and collagen fiber regeneration in the treated group. In addition, a study on self-healing biomaterials discovered that a novel basic fibroblast growth factor (bFGF)/stromal cell-derived factor (SDF-1)/hydrogel cross-linking material effectively induced BMSCs homing and enhanced collagen production in vaginal tissues, with kyoto encyclopedia of genes and genomes (KEGG) enrichment analysis confirming that the JAK-STAT signaling pathway plays a key role associated with collagen production, though more specific mechanisms need further exploration. Therefore, MSCs play a broad role in treating diseases like SUI caused by ECM injury in pelvic floor tissues through regulation of the JAK/STAT signaling pathway.
The Wnt/β-catenin signaling pathway is central to various biological processes, including cell proliferation, differentiation, apoptosis, stem cell self-renewal, tissue homeostasis, and wound healing. It is an evolutionarily conserved pathway, typically divided into β-catenin-dependent and independent types . Classical pathway transduction primarily involves extracellular Wnt protein binding to frizzled(Frz) protein and co-receptor low-density lipoprotein receptor-related protein 5/6 (LRP5/6). This triggers phosphorylation of disheveled protein (Dsh), transmitting signals into the cytoplasm to inhibit glycogen synthase kinase 3 β (GSK-3β) activation, leading to accumulation of intracellular cytoplasmic free β-catenin. Subsequently, β-catenin enters the nucleus and binds to lymphocyte enhancer factor/T cell factor (LEF/TCF) to promote downstream gene expression . Indeed, the Wnt pathway plays various roles in ECM-related diseases such as fibrosis but is considered a challenge for therapeutic targeting . Identifying potent and specific Wnt pathway inhibitors for the treatment of cancer and other diseases remains one of the most significant challenges for future research in this field . In recent years, the Wnt protein signaling pathway has been recognized for its key roles in various diseases, including tumors , fibrosis , pelvic prolapse , and SUI . Studies involving the ECM have shown that activation of β-catenin can be involved in both the synthesis and degradation of the ECM by secreted substances such as MMPs. In pelvic floor disorders, reduced β-catenin expression in vaginal fibroblasts may be associated with pelvic organ prolapse. One reason for this association is that a significant reduction in collagen I expression by vaginal fibroblasts was found in cases of pelvic organ prolapse, which was reversed with the use of Wnt/β-catenin activators . This partially reflects the link between Wnt/β-catenin and pelvic floor dysfunction. Fewer studies have shown that MSCs improve SUI by modulating Wnt/β-catenin to regulate ECM. In other diseases, some researchers have found that EVs secreted by BMSCs, like ABs, can regulate the Wnt/β-catenin pathway to ameliorate endometrial fibrosis, though the specific molecular mechanism has not been explored . Furthermore, in a study exploring chondrocyte damage in arthritis, it was found that MSC-derived EVs could load miR-3960 into cartilage tissue cells to downregulate pleckstrin homology-like domain family a member 2 (PHLDA2) proteins, thereby inhibiting the syndecan-1 (SDC-1)/ Wnt/β-catenin axis and reducing ECM degradation . In conclusion, based on the effect of β-catenin activation on collagen production in vaginal fibroblasts and its regulation by paracrine secretions from MSCs, we hypothesized that this signaling pathway may play a key role in the treatment of SUI in MSCs and may represent a promising therapeutic target.
Akt, a serine/threonine kinase also known as protein kinase B (PKB), is activated by a variety of growth factors, including vascular endothelial growth factor (VEGF), fibroblast growth factor (FGF), insulin-like growth factor (IGF), and TGF-β. Upon activation, Akt phosphorylates numerous downstream effectors crucial for apoptosis, transcription, and carcinogenesis regulation . The most prevalent activation pathway for Akt is the phosphatidylinositol 3-kinase (PI3K)-dependent route, involving two subunits, the regulatory p85 and the catalytic p110, which together activate Akt. The mammalian target of rapamycin (mTOR) is a major downstream target of Akt, completing the primary nodes of this pathway . The PI3K/AKT pathway plays a significant role in ECM synthesis regulation. Recent findings suggest that silk fibroin (SF) can activate the integrin/PI3K/AKT signaling pathway in MSCs, either directly or indirectly, enhancing their paracrine function and promoting collagen production . Studies focusing on the PI3K/AKT pathway in MSCs for improving SUI are limited. It has been reported that MSCs can secrete insulin-like growth factor-1 (IGF-1) , which activates the Akt signaling pathway, accelerating pelvic floor nerve and tissue recovery in rat-induced SUI models . While IGF-1 is known to influence a variety of cellular processes, such as growth, motility, and differentiation, the impact of Akt activation by IGF-1 on pelvic floor ECM has not been thoroughly investigated. However, other studies have shown that IGF-1 can promote collagen synthesis in tendon tissues , though this effect is not direct but mediated through the co-regulation of the PI3K/Akt and ERK pathways . It remains to be determined whether MSCs can regulate pelvic floor collagen synthesis via the IGF-1/PI3K/AKT axis. Consequently, the PI3K/AKT signaling pathway could be a promising therapeutic target for treating SUI.
The mitogen-activated protein kinase (MAPK) signaling pathway, commonly mediated by protein kinase-coupled receptors, is evolutionarily conserved and involves receptors like the epidermal growth factor receptor (EGFR), fibroblast growth factor receptor (FGFR), platelet-derived growth factor receptor (PDGFR), and vascular endothelial growth factor receptor (VEGFR) in a receptor tyrosine kinase (RTK) signaling cascade . The well-characterized MAPK family includes extracellular signal-regulated kinase 1/2 (ERK1/2), c-Jun amino-terminal kinase (JNK), p38, and ERK5. The ERK/MAPK pathway involves a multilayered kinase cascade activated by receptor-ligand binding, initiating the MAPK signaling pathway through Ras protein activation. This leads to the phosphorylation of Raf, MEK1/2, and ERK1/2 target proteins. Phosphorylated ERK1/2 activates various transcription factors to regulate gene expression and can also affect subcellular responses within the cytoplasm . The MAPK pathway significantly influences ECM synthesis, degradation, and remodeling, controlling ECM dynamic stability by regulating ECM-related gene expression, such as MMPs . In a study examining the relationship between ERK1/2 and SUI in periurethral support tissue fibroblasts, the use of an ERK kinase inhibitor was shown to influence the expression of mRNAs and proteins for collagen I and III in vaginal fibroblast cultures, suggesting that the ERK/MAPK pathway affects pelvic floor collagen tissue and may be implicated in SUI pathogenesis . Additionally, ADSCs were found to potentially improve SUI by activating ERK1/2 through vascular endothelial growth factor (VEGF) , which binds to receptor proteins and triggers a series of responses in target cells. Tissue analysis reveals an increase in collagen I/III ratios (enhancing urothelial tensile strength )and more dense and organized elastin. Therefore, MSCs may regulate the ERK/MAPK signaling pathway to affect ECM remodeling, offering a potential treatment for SUI and a key target for future therapeutic strategies.
The preceding discourse has methodically delineated the impact of MSCs on female SUI from the perspectives of tissue structure, cellular function, and molecular mechanisms. Current research is predominantly confined to preclinical experiments, employing MSCs in animal models via intravenous or periurethral injections to investigate histological and molecular changes in the periurethral tissues, while monitoring parameters such as LPP, EMG, PNSBP, PNMBP, and PNENG to assess the recovery from SUI, yielding promising outcomes. However, clinical trials that delve into histological and molecular studies are scarce, focusing primarily on symptomatic improvement due to the inability to obtain post-treatment tissue samples from patients. Additionally, the administration route, therapeutic dosage, in vivo survival time, and therapeutic mechanisms of these treatments necessitate confirmation through large-sample studies. Moreover, within the realm of research concerning the molecular underpinnings of the extracellular matrix associated with MSCs in the treatment of SUI, the TGF-β/SMAD signaling cascade emerges as the predominant pathway orchestrating collagen homeostasis, with Nrf2 potentially representing a therapeutic target. Although the JAK/STAT, Wnt/β-catenin, and PI3K/AKT pathways have been identified, their precise roles in therapeutic regulation remain enigmatic. In the ERK/MAPK signaling axis, VEGF is capable of activating ERK1/2, thereby ameliorating SUI symptoms. Collectively, MSCs are posited to enhance ECM metabolism across a spectrum of signaling pathways, primarily modulating the expression of collagen I and III, as well as elastin. However, the elucidation of specific target genes or molecular targets necessitates further investigation, underscoring a critical challenge and focal point for forthcoming research endeavors. Embryonic stem cell (ESC) therapy is widely debated due to ethical concerns. To mitigate these disputes, researchers are continuously seeking stem cell types that align with ethical standards and offer optimal efficacy, with induced pluripotent stem cells (iPSCs) emerging as a viable alternative that significantly reduces the risk of teratoma formation . Additionally, since their initial extraction from bone marrow in 1974 , MSCs have been extensively utilized in preclinical studies for the treatment of various diseases, particularly in rodents, with their safety and efficacy being evident. Concomitantly, in the systematic review of clinical trials by Lalu et al. , it was observed that, aside from transient pyrexia, no other adverse effects were identified following MSC therapy. MSCs, derived from a variety of adult tissues, possess the unique capability to differentiate into endodermal, mesodermal, and ectodermal lineages, akin to the properties of embryonic stem cells, yet they circumvent the ethical dilemmas associated with human embryonic involvement . Consequently, ethical controversies surrounding MSCs are virtually nonexistent. However, the primary limitations of MSCs may lie in their heterogeneity and the efficiency of their differentiation processes .
The molecular pathogenesis of SUI is not yet fully understood, with a growing body of evidence suggesting that pathological changes in the ECM of pelvic floor support tissues play a role in its pathogenesis. To date, the modulation of ECM metabolism by MSCs has been increasingly investigated in the treatment of various conditions, including SUI. Our review indicates that MSCs exhibit significant potential in enhancing urethral sphincter functionality, modulating connective tissue structure, and stimulating fibroblast activity, thereby reconstructing the ECM and restoring aberrant pelvic floor support structures through the influence on ECM-related signaling pathways such as TGF-β/SMAD, JAK/STAT, Wnt/β-catenin, PI3K/AKT, and ERK/MAPK. However, these studies are often limited by small sample sizes and are conducted exclusively in animal models, which to some extent restricts the extrapolation of findings to human conditions and heightens the uncertainty of their clinical application in humans. Within the bounds of ethics, future research on the modulation of ECM by MSCs for the treatment of SUI may progressively employ primate models to investigate the histological changes in pelvic floor tissues induced by MSC therapy in humans, while also exploring optimal routes, dosages, and long-term effects, with the aim of providing more effective treatment regimens for clinical SUI patients. Additionally, in recent years, the technology of MSCs improving damaged tissues by secreting Exos to regulate relevant signaling pathways in target cells has gradually matured. Nevertheless, studies involving signaling pathways within cells related to pelvic floor support tissues remain scarce, with most research being confined to macroscopic histological changes, which significantly limits the understanding of underlying mechanisms. Therefore, research on MSCs regulating target cell signaling pathways to reshape ECM in the treatment of SUI may emerge as a mainstream direction with broad prospects. In summary, MSCs hold the potential to cure SUI, and the underlying mechanisms are likely to be progressively elucidated.
|
Resistance towards and biotransformation of a | a5c2f8c5-9e04-4a18-a301-afbf866d9c24 | 11203913 | Microbiology[mh] | The use of bacteria and their antimicrobial metabolites has emerged as a potential alternative to maurhofer 1994synthetic agrochemicals in control of phytopathogens. In Pseudomonas protegens , a model organism for biocontrol, secondary metabolites such as 2,4-diacetylphloroglucinol (DAPG), pyoluteorin, and orfamide A have been shown to play essential roles in suppression of plant pathogens . For example, DAPG suppress the fungal take-all disease in cereal plants , pyoluteorin is involved in the suppression of Pythium damping-off of cress , and orfamide A can reduce bacterial wilt disease in tomato . However, as bacterial biocontrol is often associated with varying efficiencies across fields and plants , it remains an important challenge to identify and mitigate the processes responsible for these variations. Although biocontrol is a complex phenomenon that relies on multiple processes for its effectiveness , central processes involve interactions between an invading biocontrol strain and its secreted antibiotics, and the resident microbial community. For example, invasion and competition with existing microbial populations in the soil and rhizosphere is a prerequisite for effective biocontrol. Although recent results have shown that specific Pseudomonas -produced secondary metabolites are indeed required for efficient invasion of the rhizosphere of Arabidopsis roots and alters the structure of the resident, synthetic microbial community (SynCom) , other studies have found microbial communities that are more resilient towards invasion. For example, earlier studies have shown that P. protegens inoculants, including engineered variants with enhanced production of DAPG and pyoluteorin, had no impact on the structure of the bacterial community on cucumber roots or on the proportion of bacteria that were tolerant or sensitive to DAPG and pyoluteorin . Similarly, inoculation with various species of secondary metabolite producing fluorescent Pseudomonas may result in vastly different outcomes, ranging from temporary, spatially limited, and transient effects on the natural rhizosphere microbiomes to significant perturbations within indigenous microbiomes . Although these results suggest that resilience towards invading Pseudomonas biocontrol strains and their antibiotics may depend on the composition and activity of the invaded microbial community, little is known about the underlying community traits or mechanisms that may operate to tolerate otherwise toxic levels of antibiotics and to constrain Pseudomonas invasion. In this study we use a hydrogel-based bead system as in vitro model system to systematically explore the contribution of the secondary metabolites, DAPG, pyoluteorin, and orfamide A, to the ability of P. protegens DTU9.1 to invade and establish in a four-membered SynCom of commonly isolated soil bacteria . The hydrogel environment has been shown to mimic soil characteristics allowing for spatial distribution of microbes, as well as surface colonization , and thus enabling systematic analyses of community-level interactions affecting secondary metabolism in P. protegens DTU9.1 in an artificial yet soil-like environment. We showed that community traits affect SynCom susceptibility towards the toxic antibiotics, and that one of the underlying mechanisms is hydrolysis and subsequent degradation of the cyclic lipopeptide orfamide A involving several members of the four-species SynCom. These results provide insight into how levels and activity of antibiotic metabolites from invading Pseudomonas strains may be shaped by interspecies interactions in microbial communities and provide a framework for studying community mechanisms that affect invasion or efficacy of biocontrol strains.
Microorganisms and cultivation Plasmid cloning was performed in Escherichia coli CC118- λpir . Cells were cultured in lysogeny broth (LB; Lennox, Merck, St. Louis, MO, USA) with appropriate antibiotics. The antibiotic concentration used was 10 μg/mL for chloramphenicol, and 8 μg/mL and 50 μg/mL for tetracycline for E. coli and P. protegens , respectively. E. coli CC118 λ pir was cultured by inoculating a single colony in 5 mL LB broth and incubating overnight at 37°C with shaking (200 rpm). P. protegens DTU9.1 and members of the synthetic community were cultured by inoculating a single colony in 5 mL LB broth and incubating overnight at 30°C with shaking (200 rpm). The community members include Pedobacter sp . D749 (Accession: CP079218), Rhodococcus globerulus D757 (Accession: CP079698), S. indicatrix D763 (Accession: CP079106), and Chryseobacterium sp . D764 (Accession: CP079219) . Minimal inhibitory concentration (MIC) assay A MIC assay was conducted to determine the susceptibility of each SynCom member towards orfamide A. SynCom members were cultured in four biological replicates in LB O/N. Cells were washed twice in 0.9% NaCl. A clear 96-well flat-bottom microplate (Greiner Bio-One) was prepared with 200 μL 0.1x TSB per well inoculated with bacteria to an initial OD 600 of 0.01 and appropriate serial dilutions of metabolites. The microplate was covered with semi-permeable membrane (Breathe-Easy, Merck) and incubated at room temperature with 600 rpm shaking for 24 hours, followed by MIC-value determination. Generation of secondary metabolite deficient mutants To generate mutants in P. protegens DTU9.1 incapable of synthesizing DAPG, pyoluteorin, and orfamide A, genes required for biosynthesis were deleted by allelic replacement as reported previously . Primers used for cloning and verification are summarized in . In short, DNA fragments directly upstream and directly downstream of the gene of interest were PCR amplified and subsequently joined by splicing-by-overlap extension PCR with XbaI and SacI overhangs. The purified PCR product was restriction-digested and inserted in pNJ1 . The resulting plasmid was mobilized into P. protegens DTU9.1 via triparental mating with E. coli HB101 harboring the helper plasmid pRK600. Merodiploid transconjugants were initially selected on Pseudomonas Isolation Agar (PIA, Merck) supplemented with 50 μg/mL tetracycline. A second selection was performed on NSLB agar (10 g/L tryptone, 5 g/L yeast extract, 15 g/L Bacto agar) with 15% v/v sucrose. Candidates for successful deletion were confirmed by PCR and verified by Sanger sequencing at Eurofins Genomics. Integration of P. protegens DTU9.1 in a synthetic microbial community The effect of introducing P. protegens DTU9.1 into a synthetic bacterial community was investigated in an artificial soil medium composed of spherical hydrogel beads. The beads were prepared according to a previously published method . In short, a polymer solution was prepared as a 4:1 mixture of 9.6 g/L gellan gum (Phytagel, Sigma) and 2.4 g/L sodium alginate (Sigma) dissolved in distilled water. Spherical beads with a diameter of approximately 3–4 mm were formed by dropping polymer solution into a cross-linker solution containing 20 g/L CaCl 2 with a 10 mL syringe. Then, the beads were soaked in 0.1x TSB (Sigma) for 1 hour followed by sieving the beads to remove residual TSB medium. Finally, 20 mL beads were transferred to 50 mL Falcon tubes. Cultures of the four community members and P. protegens WT and Δ ofaA were grown overnight (O/N). The optical density at 600 nm (OD 600 ) of Pedobacter and Rhodococcus was set to 2.0, for Stenotrophomonas and Chryseobacterium it was set to 0.1, and for P. protegens DTU9.1 and mutants it was set to 0.001 (see for CFU/mL). Bacterial inoculation suspensions were prepared by mixing equal volumes in a total volume of 2 mL 0.1x TSB. Lastly, the prepared beads were inoculated with the 2 mL bacterial suspension. Inoculated bead systems were incubated static at RT and samples collected after 1, 4, and 7 days. Sampling was performed by briefly shaking the bead systems followed by extracting approximately 1 mL beads into new 15 mL Falcon tubes. Extracted beads were subsequently diluted in 0.9% (w/v) NaCl according to their weight to normalize the amount of bacterial cells. The tubes were shaken on a vortex for 10 minutes at maximum speed to disrupt the hydrogel beads. After vortexing, dilutions were spread on 0.1x TSA plates and incubated at RT for 48 hours before counting CFU/mL. The remaining liquid (approx. 5 mL) of the processed samples were saved for chemical detection of secondary metabolites. Detection of secondary metabolites with LC-HRMS To extract secondary metabolites in the hydrogel bead samples and supernatants of O/N cultures, an equal volume of ethyl acetate was added to the samples followed by shaking the tubes briefly. For extraction of metabolites from 0.1x TSA plates, an agar plug covering entire bacterial colonies (approx. 6 mm diameter for normal plates and 30 mm diameter for swarming plates) was suspended in 1 mL isopropanol:ethyl acetate (1:3 v/v) with 1% formic acid and shaken briefly. For both types of extractions, tubes were subsequently centrifuged for 3 minutes at 5000 x g and the top layer was transferred to new tubes. Extracts were then evaporated under N 2 . The dried extracts were re-suspended in 200 μL methanol (MeOH) and centrifuged for 3 minutes at 13000 x g. The supernatant was transferred to HPLC vials and subjected to ultra high-performance liquid chromatography electrospray ionization time-of-flight mass spectrometry (UHPLC-HRMS) analysis. LC-HRMS was performed on an Agilent Infinity 1290 UHPLC system. Liquid chromatography of 1 μL or 5 μL extract was performed using an Agilent Poroshell 120 phenyl-C 6 column (2.1 × 150 mm, 1.9 μm) at 60°C using CH 3 CN and H 2 O, both containing 20 mM formic acid. Initially, a linear gradient of 10% CH 3 CN/H 2 O to 100% CH 3 CN over 10 min was employed, followed by isocratic elution of 100% CH 3 CN for 2 min. Then, the gradient was returned to 10% CH 3 CN/H 2 O in 0.1 min and finally isocratic condition of 10% CH 3 CN/H 2 O for 1.9 min, all at a flow rate of 0.35 min/mL. HRMS data was recorded in positive ionization on an Agilent 6545 QTOF MS equipped with an Agilent Dual Jet Stream electrospray ion (ESI) source with a drying gas temperature of 250°C, drying gas flow of 8 min/L, sheath gas temperature of 300°C and sheath gas flow of 12 min/L. Capillary voltage was 4000 V and nozzle voltage was set to 500 V. Fragmentation data was collected using auto MS/MS at three collision energies (10, 20, 40 eV). The HRMS data was processed and analyzed using Agilent MassHunter Qualitative Analysis B.07.00. HPLC grade solvents (VWR Chemicals) were used for extractions whereas LCMS grade solvents (VWR Chemicals) were used for LCMS. GNPS molecular networking A molecular network was created using the Feature-Based Molecular Networking workflow on GNPS ( https://gnps.ucsd.edu , ). The workflow run can be found at this link: https://gnps.ucsd.edu/ProteoSAFe/status.jsp?task=1ad207802221433ca5431d22f2638d0e . Raw data was processed using MZmine2.53 . Data was filtered by removing all MS/MS fragment ions within +/− 17 Da of the precursor m/z. MS/MS spectra were window filtered by choosing only the top 6 fragment ions in the +/− 50 Da window throughout the spectrum. Additional settings include: precursor ion mass tolerance was set to 0.05 Da, MS/MS fragment ion tolerance to 0.05 Da, and edges were filtered to have a cosine score above 0.7 and more than 10 matched peaks. The spectra in the network were then searched against GNPS spectral libraries . The library spectra were filtered in the same manner as the input data. All matches kept between network spectra and library spectra were required to have a score above 0.7 and at least 6 matched peaks. The DEREPLICATOR was used to annotate MS/MS spectra . The molecular networks were visualized using Cytoscape 3.8.2 . MALDI sample preparation Bacterial strains were cultured on 10 mL 0.1x TSA plates at 22°C. After 4 days of incubation, the microbial colony and surrounding agar was sectioned and mounted on an IntelliSlides conductive tin oxide glass slide (Bruker). The sample was covered with matrix by spraying 1.75 mL of a matrix solution in a nitrogen atmosphere. The matrix solution was 2,5-dihydrobenzoic acid (DHB) of 20 mg/mL concentration in ACN/MeOH/H 2 O (70:25:5, v/v/v) according to . MALDI mass spectrometry imaging (MSI) Samples were dried in a desiccator overnight prior to MSI measurement. The samples were then subjected to timsTOF flex mass spectrometer (Bruker) for MALDI-IMS acquisition. Calibration was done using red phosphorus. The samples were run in positive MS scan mode with 100 μm raster width and a m/z range of 100–2000. Briefly, a photograph of the colonies was loaded onto Fleximaging software, three teach points were selected to align the background image with the sample slide, measurement regions were defined, and the automatic run mode was then employed. The settings in the timsControl were as follow: Laser: imaging 100 μm, Power Boost 3.0%, scan range 26 μm in the XY interval, and laser power 90%; Tune: Funnel 1 RF 300 Vpp, Funnel 2 RF 300 Vpp, Multipole RF 300 Vpp, isCID 0 eV, Deflection Delta 70 V, MALDI plate offset 100 V, quadrupole ion energy 5 eV, quadrupole loss mass 100 m/z , collision energy 10 eV, focus pre TOF transfer time 75 μs, pre-pulse storage 8 μs. After data acquisition, the data was analyzed using SCiLS software. Swarming assay For the swarming assay R. globerulus D757, as well as the two variants of P. protegens DTU9.1 (WT and Δ ofaA ) were cultured in three biological replicates in LB broth O/N. Cells were washed twice in 0.9% NaCl prior to adjusting OD 600 to 1 for Rhodococcus and 0.001 for Pseudomonas . For cocultures equal volumes of culture suspensions were mixed, whereas for axenic plates cell suspensions were mixed with an equal volume of 0.9% NaCl. Aliquots of 5 μL were spotted in the center of 0.1x TSA plates with 0.6% agar. Plates were incubated for 48 hours at 30°C prior to taking pictures. Swarming areas were analyzed with ImageJ. Chemical hydrolysis of orfamide A Pure orfamide A (Cayman, United States) was chemically linearized by hydrolysis by mixing 100 μL (0.386 μmol, suspended in methanol) and 193 μL 0.1 M aqueous LiOH (1.93 μmol, 5 equimolar). The solution was stirred at room temperature for 21 h. The reaction mixture was quenched by addition of 29.3 μL 1 M HCl. This led to the formation of a white precipitate, which was re-dissolved by addition of 677.7 μL methanol. Complete hydrolysis was verified by LC-HRMS. Statistics Multivariate analysis of community composition was performed using PERMANOVA on Bray-Curtis dissimilarities and the model formulation Y ~ Time + Variant + Time:Variant. Follow-up PERMANOVAs were performed on each time point with only Variant as the dependent variable. A univariate comparison of CFU counts and metabolite concentrations in SynCom versus axenic culture of P. protegens DTU9.1 was carried out using Student’s t -tests assuming equal variance. Data availability LC-HRMS data has been deposited at MassIVE with the identifier, MSV000092145. MALDI-MSI has been uploaded to Metaspace ( https://metaspace2020.eu/project/Hansen-2023 ). Demultiplexed 16S rRNA sequencing reads were uploaded to NCBI SRA database under BioProject number PRJNA983551.
Plasmid cloning was performed in Escherichia coli CC118- λpir . Cells were cultured in lysogeny broth (LB; Lennox, Merck, St. Louis, MO, USA) with appropriate antibiotics. The antibiotic concentration used was 10 μg/mL for chloramphenicol, and 8 μg/mL and 50 μg/mL for tetracycline for E. coli and P. protegens , respectively. E. coli CC118 λ pir was cultured by inoculating a single colony in 5 mL LB broth and incubating overnight at 37°C with shaking (200 rpm). P. protegens DTU9.1 and members of the synthetic community were cultured by inoculating a single colony in 5 mL LB broth and incubating overnight at 30°C with shaking (200 rpm). The community members include Pedobacter sp . D749 (Accession: CP079218), Rhodococcus globerulus D757 (Accession: CP079698), S. indicatrix D763 (Accession: CP079106), and Chryseobacterium sp . D764 (Accession: CP079219) .
A MIC assay was conducted to determine the susceptibility of each SynCom member towards orfamide A. SynCom members were cultured in four biological replicates in LB O/N. Cells were washed twice in 0.9% NaCl. A clear 96-well flat-bottom microplate (Greiner Bio-One) was prepared with 200 μL 0.1x TSB per well inoculated with bacteria to an initial OD 600 of 0.01 and appropriate serial dilutions of metabolites. The microplate was covered with semi-permeable membrane (Breathe-Easy, Merck) and incubated at room temperature with 600 rpm shaking for 24 hours, followed by MIC-value determination.
To generate mutants in P. protegens DTU9.1 incapable of synthesizing DAPG, pyoluteorin, and orfamide A, genes required for biosynthesis were deleted by allelic replacement as reported previously . Primers used for cloning and verification are summarized in . In short, DNA fragments directly upstream and directly downstream of the gene of interest were PCR amplified and subsequently joined by splicing-by-overlap extension PCR with XbaI and SacI overhangs. The purified PCR product was restriction-digested and inserted in pNJ1 . The resulting plasmid was mobilized into P. protegens DTU9.1 via triparental mating with E. coli HB101 harboring the helper plasmid pRK600. Merodiploid transconjugants were initially selected on Pseudomonas Isolation Agar (PIA, Merck) supplemented with 50 μg/mL tetracycline. A second selection was performed on NSLB agar (10 g/L tryptone, 5 g/L yeast extract, 15 g/L Bacto agar) with 15% v/v sucrose. Candidates for successful deletion were confirmed by PCR and verified by Sanger sequencing at Eurofins Genomics.
P. protegens DTU9.1 in a synthetic microbial community The effect of introducing P. protegens DTU9.1 into a synthetic bacterial community was investigated in an artificial soil medium composed of spherical hydrogel beads. The beads were prepared according to a previously published method . In short, a polymer solution was prepared as a 4:1 mixture of 9.6 g/L gellan gum (Phytagel, Sigma) and 2.4 g/L sodium alginate (Sigma) dissolved in distilled water. Spherical beads with a diameter of approximately 3–4 mm were formed by dropping polymer solution into a cross-linker solution containing 20 g/L CaCl 2 with a 10 mL syringe. Then, the beads were soaked in 0.1x TSB (Sigma) for 1 hour followed by sieving the beads to remove residual TSB medium. Finally, 20 mL beads were transferred to 50 mL Falcon tubes. Cultures of the four community members and P. protegens WT and Δ ofaA were grown overnight (O/N). The optical density at 600 nm (OD 600 ) of Pedobacter and Rhodococcus was set to 2.0, for Stenotrophomonas and Chryseobacterium it was set to 0.1, and for P. protegens DTU9.1 and mutants it was set to 0.001 (see for CFU/mL). Bacterial inoculation suspensions were prepared by mixing equal volumes in a total volume of 2 mL 0.1x TSB. Lastly, the prepared beads were inoculated with the 2 mL bacterial suspension. Inoculated bead systems were incubated static at RT and samples collected after 1, 4, and 7 days. Sampling was performed by briefly shaking the bead systems followed by extracting approximately 1 mL beads into new 15 mL Falcon tubes. Extracted beads were subsequently diluted in 0.9% (w/v) NaCl according to their weight to normalize the amount of bacterial cells. The tubes were shaken on a vortex for 10 minutes at maximum speed to disrupt the hydrogel beads. After vortexing, dilutions were spread on 0.1x TSA plates and incubated at RT for 48 hours before counting CFU/mL. The remaining liquid (approx. 5 mL) of the processed samples were saved for chemical detection of secondary metabolites.
To extract secondary metabolites in the hydrogel bead samples and supernatants of O/N cultures, an equal volume of ethyl acetate was added to the samples followed by shaking the tubes briefly. For extraction of metabolites from 0.1x TSA plates, an agar plug covering entire bacterial colonies (approx. 6 mm diameter for normal plates and 30 mm diameter for swarming plates) was suspended in 1 mL isopropanol:ethyl acetate (1:3 v/v) with 1% formic acid and shaken briefly. For both types of extractions, tubes were subsequently centrifuged for 3 minutes at 5000 x g and the top layer was transferred to new tubes. Extracts were then evaporated under N 2 . The dried extracts were re-suspended in 200 μL methanol (MeOH) and centrifuged for 3 minutes at 13000 x g. The supernatant was transferred to HPLC vials and subjected to ultra high-performance liquid chromatography electrospray ionization time-of-flight mass spectrometry (UHPLC-HRMS) analysis. LC-HRMS was performed on an Agilent Infinity 1290 UHPLC system. Liquid chromatography of 1 μL or 5 μL extract was performed using an Agilent Poroshell 120 phenyl-C 6 column (2.1 × 150 mm, 1.9 μm) at 60°C using CH 3 CN and H 2 O, both containing 20 mM formic acid. Initially, a linear gradient of 10% CH 3 CN/H 2 O to 100% CH 3 CN over 10 min was employed, followed by isocratic elution of 100% CH 3 CN for 2 min. Then, the gradient was returned to 10% CH 3 CN/H 2 O in 0.1 min and finally isocratic condition of 10% CH 3 CN/H 2 O for 1.9 min, all at a flow rate of 0.35 min/mL. HRMS data was recorded in positive ionization on an Agilent 6545 QTOF MS equipped with an Agilent Dual Jet Stream electrospray ion (ESI) source with a drying gas temperature of 250°C, drying gas flow of 8 min/L, sheath gas temperature of 300°C and sheath gas flow of 12 min/L. Capillary voltage was 4000 V and nozzle voltage was set to 500 V. Fragmentation data was collected using auto MS/MS at three collision energies (10, 20, 40 eV). The HRMS data was processed and analyzed using Agilent MassHunter Qualitative Analysis B.07.00. HPLC grade solvents (VWR Chemicals) were used for extractions whereas LCMS grade solvents (VWR Chemicals) were used for LCMS.
A molecular network was created using the Feature-Based Molecular Networking workflow on GNPS ( https://gnps.ucsd.edu , ). The workflow run can be found at this link: https://gnps.ucsd.edu/ProteoSAFe/status.jsp?task=1ad207802221433ca5431d22f2638d0e . Raw data was processed using MZmine2.53 . Data was filtered by removing all MS/MS fragment ions within +/− 17 Da of the precursor m/z. MS/MS spectra were window filtered by choosing only the top 6 fragment ions in the +/− 50 Da window throughout the spectrum. Additional settings include: precursor ion mass tolerance was set to 0.05 Da, MS/MS fragment ion tolerance to 0.05 Da, and edges were filtered to have a cosine score above 0.7 and more than 10 matched peaks. The spectra in the network were then searched against GNPS spectral libraries . The library spectra were filtered in the same manner as the input data. All matches kept between network spectra and library spectra were required to have a score above 0.7 and at least 6 matched peaks. The DEREPLICATOR was used to annotate MS/MS spectra . The molecular networks were visualized using Cytoscape 3.8.2 .
Bacterial strains were cultured on 10 mL 0.1x TSA plates at 22°C. After 4 days of incubation, the microbial colony and surrounding agar was sectioned and mounted on an IntelliSlides conductive tin oxide glass slide (Bruker). The sample was covered with matrix by spraying 1.75 mL of a matrix solution in a nitrogen atmosphere. The matrix solution was 2,5-dihydrobenzoic acid (DHB) of 20 mg/mL concentration in ACN/MeOH/H 2 O (70:25:5, v/v/v) according to .
Samples were dried in a desiccator overnight prior to MSI measurement. The samples were then subjected to timsTOF flex mass spectrometer (Bruker) for MALDI-IMS acquisition. Calibration was done using red phosphorus. The samples were run in positive MS scan mode with 100 μm raster width and a m/z range of 100–2000. Briefly, a photograph of the colonies was loaded onto Fleximaging software, three teach points were selected to align the background image with the sample slide, measurement regions were defined, and the automatic run mode was then employed. The settings in the timsControl were as follow: Laser: imaging 100 μm, Power Boost 3.0%, scan range 26 μm in the XY interval, and laser power 90%; Tune: Funnel 1 RF 300 Vpp, Funnel 2 RF 300 Vpp, Multipole RF 300 Vpp, isCID 0 eV, Deflection Delta 70 V, MALDI plate offset 100 V, quadrupole ion energy 5 eV, quadrupole loss mass 100 m/z , collision energy 10 eV, focus pre TOF transfer time 75 μs, pre-pulse storage 8 μs. After data acquisition, the data was analyzed using SCiLS software.
For the swarming assay R. globerulus D757, as well as the two variants of P. protegens DTU9.1 (WT and Δ ofaA ) were cultured in three biological replicates in LB broth O/N. Cells were washed twice in 0.9% NaCl prior to adjusting OD 600 to 1 for Rhodococcus and 0.001 for Pseudomonas . For cocultures equal volumes of culture suspensions were mixed, whereas for axenic plates cell suspensions were mixed with an equal volume of 0.9% NaCl. Aliquots of 5 μL were spotted in the center of 0.1x TSA plates with 0.6% agar. Plates were incubated for 48 hours at 30°C prior to taking pictures. Swarming areas were analyzed with ImageJ.
Pure orfamide A (Cayman, United States) was chemically linearized by hydrolysis by mixing 100 μL (0.386 μmol, suspended in methanol) and 193 μL 0.1 M aqueous LiOH (1.93 μmol, 5 equimolar). The solution was stirred at room temperature for 21 h. The reaction mixture was quenched by addition of 29.3 μL 1 M HCl. This led to the formation of a white precipitate, which was re-dissolved by addition of 677.7 μL methanol. Complete hydrolysis was verified by LC-HRMS.
Multivariate analysis of community composition was performed using PERMANOVA on Bray-Curtis dissimilarities and the model formulation Y ~ Time + Variant + Time:Variant. Follow-up PERMANOVAs were performed on each time point with only Variant as the dependent variable. A univariate comparison of CFU counts and metabolite concentrations in SynCom versus axenic culture of P. protegens DTU9.1 was carried out using Student’s t -tests assuming equal variance.
LC-HRMS data has been deposited at MassIVE with the identifier, MSV000092145. MALDI-MSI has been uploaded to Metaspace ( https://metaspace2020.eu/project/Hansen-2023 ). Demultiplexed 16S rRNA sequencing reads were uploaded to NCBI SRA database under BioProject number PRJNA983551.
Utilizing a hydrogel bead system to explore invasion of P. protegens DTU9.1 in a synthetic microbial community To explore the microbial interactions and their effects on community composition over time during invasion of P. protegens DTU9.1, a porous hydrogel bead system was chosen as cultivation system, as it has been shown to mimic soil characteristics and allow for spatial distribution of microbes . The four-membered SynCom ( Pedobacter sp. D749, R. globerulus D757, Stenotrophomonas indicatrix D763, and Chryseobacterium sp. D764) was comprised of co-isolated species from a soil sample site that we previously demonstrated to also contain P. protegens . Initially, it was verified that the four SynCom members could establish and co-exist in the hydrogel beads over a 7-day period . As shown, all members could be detected, although Pedobacter sp . D749 reached an abundance close to the limit of detection (10 5 CFU/mL). Additionally, it was determined that P. protegens DTU9.1 could colonize and maintain itself in the hydrogel bead environment when cultivated axenically . After four days of cultivation, Pseudomonas reached a plateau of approximately 10 8 CFU/mL, which was maintained at the seventh and final day of the experiment. Lastly, it was confirmed that wildtype P. protegens DTU9.1 produced detectable amounts of DAPG, pyoluteorin, and orfamide A after seven days of axenic cultivation in the bead system, as verified using liquid chromatography coupled to high resolution-mass spectrometry (LC-HRMS) (Fig. S1). P. protegens DTU9.1 invades a four-membered synthetic microbial community in a soil-mimicking environment Next, we sought to investigate the sensitivity of each SynCom member towards the three antimicrobial Pseudomonas -produced metabolites (DAPG, pyoluteorin, and orfamide A). R. globerulus D757 displayed sensitivity towards all three metabolites, whereas three members ( Pedobacter sp. D749, R. globerulus D757 and Chryseobacterium sp. D764) were equally susceptible to pyoluteorin with an MIC of 8 μg/mL . Stenotrophomonas indicatrix D763 was generally more resistant towards all three antimicrobial metabolites . Thus, we hypothesized that several community members would be affected upon invasion by wildtype P. protegens DTU9.1 given its ability to produce all three metabolites in the hydrogel bead system. To test this hypothesis, we cultured P. protegens DTU9.1 wildtype along with the four-membered SynCom for seven days in the hydrogel bead system. After one day of cultivation, all SynCom members were still detected , whereas the amount of cultivable P. protegens DTU9.1 cells were 100–1000 fold lower than the SynCom members, likely owing to the low inoculum size of P. protegens DTU9.1 (see Methods). However, on the fourth day, P. protegens DTU9.1 had reached high cell numbers (≈10 8 CFU/mL), which were maintained throughout the experiment and comparable to that observed in axenic cultivation (c.f. ). On the seventh day of incubation P. protegens DTU9.1 constituted the majority of the bacterial biomass, although three of the four SynCom members remained detectable with our method of colony counting . On the fourth day and onwards, colony forming units of Pedobacter sp. D749 were no longer detectable in the systems inoculated with P. protegens DTU9.1. However Pedobacter sp. D749 could be detected at higher cell titers throughout the experiment without the addition of Pseudomonas , albeit with cell counts close to the detection limit on the seventh day . We collected DNA from the hydrogel bead systems for each sampling point and performed amplicon sequencing of the 16S rRNA gene to verify the presence of Pedobacter sp . D749. The presence of the bacterium was confirmed throughout the seven days in both the control system and the system with P. protegens DTU9.1 wildtype (Fig. S2). To investigate further if the changes in community composition upon invasion by P. protegens DTU9.1 was due to production of toxic levels of the three antibiotics, we extracted metabolites from the hydrogel bead systems after seven days of incubation and analyzed the samples with LC-HRMS to verify the production and quantity of the three secondary metabolites in question. During axenic cultivation, P. protegens DTU9.1 reached similar levels of colony forming units as compared to growth alongside the SynCom . The concentrations of the secondary metabolites produced by P. protegens DTU9.1 differed markedly between the two systems. In the case of both DAPG and pyoluteorin, concentrations were elevated during cocultivation compared to axenic growth, although this was not significant for pyoluteorin ( P = 0.056). This could suggest that P. protegens DTU9.1 responded to the presence of competing microorganisms by increasing the production of its antimicrobial secondary metabolites (i.e. DAPG and pyoluteorin). However, in the case of the cyclic lipopeptide, orfamide A, the concentration was significantly lower ( P = 0.0002) when P. protegens DTU9.1 was cultivated with the SynCom compared to axenic growth . Considering that the level of orfamide A in the SynCom system is below the measured MIC values , we hypothesized that introducing an orfamide A-deficient mutant would not cause significant perturbations to the system compared to introducing wildtype P. protegens DTU9.1. Thus, to test this, we constructed an orfamide A-deficient mutant (Δ ofaA ) and introduced it into the SynCom . Expectedly, the changes to community composition over a 7-day period were similar to the system inoculated with wildtype P. protegens DTU9.1. The overall effect on community composition over time compared between the two systems (SynCom + WT and SynCom + Δ ofaA ) was evaluated with a principal coordinate analysis (PCoA) using Bray-Curtis dissimilarities . The analysis suggested that the major factor affecting abundance was the sampling time, as separate clusters representing the three sampling times appeared across the first principal component, which explained the largest portion of the variance in the data (50.3%). An overall PERMANOVA using sampling time, genotypic variant of invading P. protegens DTU9.1, and their interaction as fixed effects further confirmed the significance of sampling time on the community composition ( P = 9.99 ∙ 10 −5 , R 2 = 0.75). A growth assay of each SynCom member, as well as P. protegens DTU9.1 WT and Δ ofaA in liquid broth revealed that the invading Pseudomonas had a significantly faster doubling time than all SynCom members (Fig. S3), whereas Pedobacter sp. D749 had the longest doubling time. This could explain the observed significant effect of sampling time on community composition, as visualized by the PCoA . The levels of both DAPG and pyoluteorin were elevated during coculture of P. protegens DTU9.1 and the SynCom . Furthermore, these levels exceeded the MIC values towards several SynCom members . Thus, to investigate if all three antimicrobial metabolites played a role in the ability of P. protegens DTU9.1 to establish within the bead system co-inoculated with the SynCom, we constructed a Δ phlACB , Δ pltA , Δ ofaA (from here on referred to as ΔTriple) knockout mutant deficient in production of DAPG, pyoluteorin, and orfamide A. The ΔTriple mutant was not able to establish and reach cell titers comparable to that of P. protegens DTU9.1 wildtype . Furthermore, a PCoA and subsequent overall PERMANOVA confirmed that variant rather than sampling time was the main driver of perturbations to community composition over time when comparing the SynCom + WT and SynCom + ΔTriple systems (Fig. S4). However, introduction of mutants deficient in production of either DAPG or pyoluteorin (Δ phlACB and Δ pltA , respectively) into the SynCom caused similar perturbations to community composition over time as wildtype (Fig. S5A). Additionally, we observed that levels of DAPG and pyoluteorin increased in their respective opposite knockout mutant during axenic growth of P. protegens DTU9.1 (Fig. S5B). This could suggest that P. protegens DTU9.1 utilizes a secondary metabolite-mediated approach to establish within the SynCom, which relies on production of at least DAPG or pyoluteorin. Orfamide A is degraded during cocultivation of P. protegens DTU9.1 and SynCom in the hydrogel bead system We observed that the levels of orfamide A were significantly reduced during coculture of P. protegens DTU9.1 and the SynCom compared to axenic cultivation of Pseudomonas . Given that one of the SynCom members is sensitive towards orfamide A and that orfamide A is required for swarming behavior to colonize new niches , our observation could indicate that members of the synthetic community either inhibited production of orfamide A or affected the persistence of the metabolite over time as a potential mechanism of resistance. To further investigate this observation and the fate of orfamide A ( 1 , ) further, the MS data from the LC-HRMS analyses was subjected to Global Natural Product Social (GNPS) molecular network analysis to identify potential chemical relationships between features across the three systems (Fig. S6). This revealed a distinct molecular family displaying the presence of orfamide A ( m/z 1295.8509) during both axenic cultivation of P. protegens DTU9.1 and cocultivation with the SynCom . This molecular family also contained features with related fragmentation patterns to orfamide A that only appeared during cocultivation of P. protegens DTU9.1 and the SynCom. One feature, m/z 1313.8542, corresponded to the mass of hydrolyzed orfamide A, with the addition of 18 Da . Fragmentation analysis suggested hydrolysis of the ester bond connecting the macrocyclic ring , which created a linearized congener of orfamide A ( 2 , ) when P. protegens DTU9.1 was cultivated alongside the SynCom in the bead system. Potential degradation products of orfamide A were also present in the same molecular family . The suspected degradation products were verified in the LC-HRMS data of the extracts. Differences in the retention times of each of these features provided confirmation that they were not arising from in-source fragmentation (Fig. S7). Fragmentation patterns were subsequently analyzed, confirming degradation products emanating from the hydrolyzed ester bond (Fig. S8). A similar phenomenon was observed for orfamide B through the presence of degradation products ( m/z 1113.7369, m/z 1000.6529, and m/z 887.5683 in ). The remaining degradation products of orfamide B were present in concentrations too low to be selected for MS/MS, but were observed in the raw data (available from the raw data. See Methods). Taken together, these results demonstrate that one or multiple SynCom members were able to hydrolyze orfamide A (and orfamide B) from P. protegens DTU9.1 and degrade the linearized lipopeptide. Linearization of orfamide A is caused by R. globerulus D757 and inhibits the motility of P. protegens DTU9.1 To explore the degradation of orfamide A, we turned to dual-species interactions on agar surfaces to investigate if a single community member was responsible for linearization by hydrolysis and degradation. First, P. protegens DTU9.1 was cocultivated with each of the four SynCom members individually in mixed species colonies on 0.1x TSA. MS data of the metabolites extracted from an agar plug covering the entire bacterial colony revealed that the Gram-positive R. globerulus D757 was able to hydrolyze the ester bond of 1 yielding the linearized product, 2 . Additionally, we monitored the interaction between P. protegens DTU9.1 and R. globerulus D757 by matrix-assisted laser desorption-ionization mass spectrometry imaging (MALDI-MSI) to validate that the linearization of orfamide A indeed occurs in the interface between the two bacteria. The two species were cultivated for 4 days on 0.1x TSA plate prior to matrix application and imaging. Orfamide A ( 1 ) was secreted evenly around the P. protegens DTU9.1 colony, whereas the linearized product ( 2 ) was observed only in the interface between the two bacteria . Additionally, we investigated the effect of coculturing R. globerulus D757 and P. protegens DTU9.1 on the swarming motility of the Pseudomonas . Axenic cultivation of P. protegens DTU9.1 led to a uniform faint circular spread from the point of inoculation after 48 hours of incubation . Expectedly, the Δ ofaA knockout mutant was completely impaired in the ability to swarm. Cocultivation with R. globerulus D757 significantly inhibited the swarming motility of wildtype P. protegens DTU9.1, where only a slight faint zone was observed surrounding the primary colony ( , see also Fig. S9). The presence of orfamide A ( 1 ) and its hydrolyzed congener ( 2 ) was verified by extracting metabolites from an agar plug covering the entire swarming area followed by LC-HRMS (Fig. S10A). Additionally, both variants of P. protegens DTU9.1 (WT and Δ ofaA ) were chromosomally tagged with gfp under strong constitutive expression, which allowed us to determine the abundance of the non-fluorescent R. globerulus D757 and green fluorescent P. protegens DTU9.1 in the swarming colony by flow cytometry (Fig. S10B, Supplementary methods). This analysis showed that P. protegens DTU9.1 constituted half of the microbial biomass in the coculture with R. globerulus D757 after 48 hours, although swarming was inhibited. This suggests that hydrolysis of orfamide A by R. globerulus D757 may serve as both a defensive resistance mechanism to reduce toxic levels of the antimicrobial metabolite, while simultaneously blocking the swarming-mediated motility of the metabolite-producer, P. protegens DTU9.1. Fate of orfamide A is affected by a multi-species interaction involving R. globerulus D757 and S. indicatrix D763 In the dual species assays conducted above we only observed linearization of orfamide A by hydrolysis, but no subsequent degradation. Thus, we hypothesized that hydrolysis was a prerequisite for one or more of the remaining three SynCom members ( Pedobacter sp. D749, S. indicatrix D763, and Chryseobacterium sp. D764) to further degrade the hydrolyzed orfamide A ( 2 ). To investigate this hypothesis, we chemically hydrolyzed pure orfamide A by prolonged incubation in an alkaline solution (see Methods) and exposed the remaining three SynCom members individually to hydrolyzed orfamide A over 24 hours in liquid broth. Subsequent extraction of metabolites from the supernatants revealed that S. indicatrix D763 could degrade the hydrolyzed lipopeptide . Here, we show the feature, m/z 1014.6702 ( 3 ), as an example of degradation. This feature corresponds to the mass of 2 excluding three amino acids from the C-terminal end . However, we did also observe several of the larger degradation products as shown in (and Fig. S11). The lack of the smaller degradation products as initially observed in the bead system could be explained by the reduced incubation time of 24 hours compared to seven days in the beads. We also observed small amounts of the two features ( m/z 901.5862 and m/z 814.5541) in the culture with Chryseobacterium sp . D764. This could indicate that Chryseobacterium sp . D764 also secretes enzymes that can break down 2 , although to a much lesser extent than S. indicatrix D763. Thus, the fate of orfamide A ( 1 ) in our experimental setup was affected by a sequential, multi-species interaction involving the initial hydrolysis by R. globerulus D757 and subsequent degradation by primarily S. indicatrix D763 .
P. protegens DTU9.1 in a synthetic microbial community To explore the microbial interactions and their effects on community composition over time during invasion of P. protegens DTU9.1, a porous hydrogel bead system was chosen as cultivation system, as it has been shown to mimic soil characteristics and allow for spatial distribution of microbes . The four-membered SynCom ( Pedobacter sp. D749, R. globerulus D757, Stenotrophomonas indicatrix D763, and Chryseobacterium sp. D764) was comprised of co-isolated species from a soil sample site that we previously demonstrated to also contain P. protegens . Initially, it was verified that the four SynCom members could establish and co-exist in the hydrogel beads over a 7-day period . As shown, all members could be detected, although Pedobacter sp . D749 reached an abundance close to the limit of detection (10 5 CFU/mL). Additionally, it was determined that P. protegens DTU9.1 could colonize and maintain itself in the hydrogel bead environment when cultivated axenically . After four days of cultivation, Pseudomonas reached a plateau of approximately 10 8 CFU/mL, which was maintained at the seventh and final day of the experiment. Lastly, it was confirmed that wildtype P. protegens DTU9.1 produced detectable amounts of DAPG, pyoluteorin, and orfamide A after seven days of axenic cultivation in the bead system, as verified using liquid chromatography coupled to high resolution-mass spectrometry (LC-HRMS) (Fig. S1).
DTU9.1 invades a four-membered synthetic microbial community in a soil-mimicking environment Next, we sought to investigate the sensitivity of each SynCom member towards the three antimicrobial Pseudomonas -produced metabolites (DAPG, pyoluteorin, and orfamide A). R. globerulus D757 displayed sensitivity towards all three metabolites, whereas three members ( Pedobacter sp. D749, R. globerulus D757 and Chryseobacterium sp. D764) were equally susceptible to pyoluteorin with an MIC of 8 μg/mL . Stenotrophomonas indicatrix D763 was generally more resistant towards all three antimicrobial metabolites . Thus, we hypothesized that several community members would be affected upon invasion by wildtype P. protegens DTU9.1 given its ability to produce all three metabolites in the hydrogel bead system. To test this hypothesis, we cultured P. protegens DTU9.1 wildtype along with the four-membered SynCom for seven days in the hydrogel bead system. After one day of cultivation, all SynCom members were still detected , whereas the amount of cultivable P. protegens DTU9.1 cells were 100–1000 fold lower than the SynCom members, likely owing to the low inoculum size of P. protegens DTU9.1 (see Methods). However, on the fourth day, P. protegens DTU9.1 had reached high cell numbers (≈10 8 CFU/mL), which were maintained throughout the experiment and comparable to that observed in axenic cultivation (c.f. ). On the seventh day of incubation P. protegens DTU9.1 constituted the majority of the bacterial biomass, although three of the four SynCom members remained detectable with our method of colony counting . On the fourth day and onwards, colony forming units of Pedobacter sp. D749 were no longer detectable in the systems inoculated with P. protegens DTU9.1. However Pedobacter sp. D749 could be detected at higher cell titers throughout the experiment without the addition of Pseudomonas , albeit with cell counts close to the detection limit on the seventh day . We collected DNA from the hydrogel bead systems for each sampling point and performed amplicon sequencing of the 16S rRNA gene to verify the presence of Pedobacter sp . D749. The presence of the bacterium was confirmed throughout the seven days in both the control system and the system with P. protegens DTU9.1 wildtype (Fig. S2). To investigate further if the changes in community composition upon invasion by P. protegens DTU9.1 was due to production of toxic levels of the three antibiotics, we extracted metabolites from the hydrogel bead systems after seven days of incubation and analyzed the samples with LC-HRMS to verify the production and quantity of the three secondary metabolites in question. During axenic cultivation, P. protegens DTU9.1 reached similar levels of colony forming units as compared to growth alongside the SynCom . The concentrations of the secondary metabolites produced by P. protegens DTU9.1 differed markedly between the two systems. In the case of both DAPG and pyoluteorin, concentrations were elevated during cocultivation compared to axenic growth, although this was not significant for pyoluteorin ( P = 0.056). This could suggest that P. protegens DTU9.1 responded to the presence of competing microorganisms by increasing the production of its antimicrobial secondary metabolites (i.e. DAPG and pyoluteorin). However, in the case of the cyclic lipopeptide, orfamide A, the concentration was significantly lower ( P = 0.0002) when P. protegens DTU9.1 was cultivated with the SynCom compared to axenic growth . Considering that the level of orfamide A in the SynCom system is below the measured MIC values , we hypothesized that introducing an orfamide A-deficient mutant would not cause significant perturbations to the system compared to introducing wildtype P. protegens DTU9.1. Thus, to test this, we constructed an orfamide A-deficient mutant (Δ ofaA ) and introduced it into the SynCom . Expectedly, the changes to community composition over a 7-day period were similar to the system inoculated with wildtype P. protegens DTU9.1. The overall effect on community composition over time compared between the two systems (SynCom + WT and SynCom + Δ ofaA ) was evaluated with a principal coordinate analysis (PCoA) using Bray-Curtis dissimilarities . The analysis suggested that the major factor affecting abundance was the sampling time, as separate clusters representing the three sampling times appeared across the first principal component, which explained the largest portion of the variance in the data (50.3%). An overall PERMANOVA using sampling time, genotypic variant of invading P. protegens DTU9.1, and their interaction as fixed effects further confirmed the significance of sampling time on the community composition ( P = 9.99 ∙ 10 −5 , R 2 = 0.75). A growth assay of each SynCom member, as well as P. protegens DTU9.1 WT and Δ ofaA in liquid broth revealed that the invading Pseudomonas had a significantly faster doubling time than all SynCom members (Fig. S3), whereas Pedobacter sp. D749 had the longest doubling time. This could explain the observed significant effect of sampling time on community composition, as visualized by the PCoA . The levels of both DAPG and pyoluteorin were elevated during coculture of P. protegens DTU9.1 and the SynCom . Furthermore, these levels exceeded the MIC values towards several SynCom members . Thus, to investigate if all three antimicrobial metabolites played a role in the ability of P. protegens DTU9.1 to establish within the bead system co-inoculated with the SynCom, we constructed a Δ phlACB , Δ pltA , Δ ofaA (from here on referred to as ΔTriple) knockout mutant deficient in production of DAPG, pyoluteorin, and orfamide A. The ΔTriple mutant was not able to establish and reach cell titers comparable to that of P. protegens DTU9.1 wildtype . Furthermore, a PCoA and subsequent overall PERMANOVA confirmed that variant rather than sampling time was the main driver of perturbations to community composition over time when comparing the SynCom + WT and SynCom + ΔTriple systems (Fig. S4). However, introduction of mutants deficient in production of either DAPG or pyoluteorin (Δ phlACB and Δ pltA , respectively) into the SynCom caused similar perturbations to community composition over time as wildtype (Fig. S5A). Additionally, we observed that levels of DAPG and pyoluteorin increased in their respective opposite knockout mutant during axenic growth of P. protegens DTU9.1 (Fig. S5B). This could suggest that P. protegens DTU9.1 utilizes a secondary metabolite-mediated approach to establish within the SynCom, which relies on production of at least DAPG or pyoluteorin.
P. protegens DTU9.1 and SynCom in the hydrogel bead system We observed that the levels of orfamide A were significantly reduced during coculture of P. protegens DTU9.1 and the SynCom compared to axenic cultivation of Pseudomonas . Given that one of the SynCom members is sensitive towards orfamide A and that orfamide A is required for swarming behavior to colonize new niches , our observation could indicate that members of the synthetic community either inhibited production of orfamide A or affected the persistence of the metabolite over time as a potential mechanism of resistance. To further investigate this observation and the fate of orfamide A ( 1 , ) further, the MS data from the LC-HRMS analyses was subjected to Global Natural Product Social (GNPS) molecular network analysis to identify potential chemical relationships between features across the three systems (Fig. S6). This revealed a distinct molecular family displaying the presence of orfamide A ( m/z 1295.8509) during both axenic cultivation of P. protegens DTU9.1 and cocultivation with the SynCom . This molecular family also contained features with related fragmentation patterns to orfamide A that only appeared during cocultivation of P. protegens DTU9.1 and the SynCom. One feature, m/z 1313.8542, corresponded to the mass of hydrolyzed orfamide A, with the addition of 18 Da . Fragmentation analysis suggested hydrolysis of the ester bond connecting the macrocyclic ring , which created a linearized congener of orfamide A ( 2 , ) when P. protegens DTU9.1 was cultivated alongside the SynCom in the bead system. Potential degradation products of orfamide A were also present in the same molecular family . The suspected degradation products were verified in the LC-HRMS data of the extracts. Differences in the retention times of each of these features provided confirmation that they were not arising from in-source fragmentation (Fig. S7). Fragmentation patterns were subsequently analyzed, confirming degradation products emanating from the hydrolyzed ester bond (Fig. S8). A similar phenomenon was observed for orfamide B through the presence of degradation products ( m/z 1113.7369, m/z 1000.6529, and m/z 887.5683 in ). The remaining degradation products of orfamide B were present in concentrations too low to be selected for MS/MS, but were observed in the raw data (available from the raw data. See Methods). Taken together, these results demonstrate that one or multiple SynCom members were able to hydrolyze orfamide A (and orfamide B) from P. protegens DTU9.1 and degrade the linearized lipopeptide.
R. globerulus D757 and inhibits the motility of P. protegens DTU9.1 To explore the degradation of orfamide A, we turned to dual-species interactions on agar surfaces to investigate if a single community member was responsible for linearization by hydrolysis and degradation. First, P. protegens DTU9.1 was cocultivated with each of the four SynCom members individually in mixed species colonies on 0.1x TSA. MS data of the metabolites extracted from an agar plug covering the entire bacterial colony revealed that the Gram-positive R. globerulus D757 was able to hydrolyze the ester bond of 1 yielding the linearized product, 2 . Additionally, we monitored the interaction between P. protegens DTU9.1 and R. globerulus D757 by matrix-assisted laser desorption-ionization mass spectrometry imaging (MALDI-MSI) to validate that the linearization of orfamide A indeed occurs in the interface between the two bacteria. The two species were cultivated for 4 days on 0.1x TSA plate prior to matrix application and imaging. Orfamide A ( 1 ) was secreted evenly around the P. protegens DTU9.1 colony, whereas the linearized product ( 2 ) was observed only in the interface between the two bacteria . Additionally, we investigated the effect of coculturing R. globerulus D757 and P. protegens DTU9.1 on the swarming motility of the Pseudomonas . Axenic cultivation of P. protegens DTU9.1 led to a uniform faint circular spread from the point of inoculation after 48 hours of incubation . Expectedly, the Δ ofaA knockout mutant was completely impaired in the ability to swarm. Cocultivation with R. globerulus D757 significantly inhibited the swarming motility of wildtype P. protegens DTU9.1, where only a slight faint zone was observed surrounding the primary colony ( , see also Fig. S9). The presence of orfamide A ( 1 ) and its hydrolyzed congener ( 2 ) was verified by extracting metabolites from an agar plug covering the entire swarming area followed by LC-HRMS (Fig. S10A). Additionally, both variants of P. protegens DTU9.1 (WT and Δ ofaA ) were chromosomally tagged with gfp under strong constitutive expression, which allowed us to determine the abundance of the non-fluorescent R. globerulus D757 and green fluorescent P. protegens DTU9.1 in the swarming colony by flow cytometry (Fig. S10B, Supplementary methods). This analysis showed that P. protegens DTU9.1 constituted half of the microbial biomass in the coculture with R. globerulus D757 after 48 hours, although swarming was inhibited. This suggests that hydrolysis of orfamide A by R. globerulus D757 may serve as both a defensive resistance mechanism to reduce toxic levels of the antimicrobial metabolite, while simultaneously blocking the swarming-mediated motility of the metabolite-producer, P. protegens DTU9.1.
R. globerulus D757 and S. indicatrix D763 In the dual species assays conducted above we only observed linearization of orfamide A by hydrolysis, but no subsequent degradation. Thus, we hypothesized that hydrolysis was a prerequisite for one or more of the remaining three SynCom members ( Pedobacter sp. D749, S. indicatrix D763, and Chryseobacterium sp. D764) to further degrade the hydrolyzed orfamide A ( 2 ). To investigate this hypothesis, we chemically hydrolyzed pure orfamide A by prolonged incubation in an alkaline solution (see Methods) and exposed the remaining three SynCom members individually to hydrolyzed orfamide A over 24 hours in liquid broth. Subsequent extraction of metabolites from the supernatants revealed that S. indicatrix D763 could degrade the hydrolyzed lipopeptide . Here, we show the feature, m/z 1014.6702 ( 3 ), as an example of degradation. This feature corresponds to the mass of 2 excluding three amino acids from the C-terminal end . However, we did also observe several of the larger degradation products as shown in (and Fig. S11). The lack of the smaller degradation products as initially observed in the bead system could be explained by the reduced incubation time of 24 hours compared to seven days in the beads. We also observed small amounts of the two features ( m/z 901.5862 and m/z 814.5541) in the culture with Chryseobacterium sp . D764. This could indicate that Chryseobacterium sp . D764 also secretes enzymes that can break down 2 , although to a much lesser extent than S. indicatrix D763. Thus, the fate of orfamide A ( 1 ) in our experimental setup was affected by a sequential, multi-species interaction involving the initial hydrolysis by R. globerulus D757 and subsequent degradation by primarily S. indicatrix D763 .
In this study, we utilized a four-membered bacterial SynCom cultivated in an artificial hydrogel bead system to evaluate the contribution of the cyclic lipopeptide, orfamide A, from a P. protegens strain on the ability of the Pseudomonas to invade and establish within the simplified microcosm. Although the genotypic complexity of our four-membered SynCom is far from a representation of the natural soil microbial diversity , the simplified system applied in this study allowed for the systematic analysis of community-level interactions affecting the secondary metabolome of P. protegens DTU9.1. We demonstrated that P. protegens readily invaded and altered the community composition. Unexpectedly, the knockout mutant unable to produce orfamide A invaded the community as efficiently as the wildtype did and caused similar perturbations in the composition of the synthetic community , despite the observation that one SynCom member was sensitive to orfamide A . This was not a result of an inability to produce non-toxic levels of orfamide A, as levels of orfamide A exceeded the MIC value towards at least one SynCom member during axenic cultivation of P. protegens DTU9.1 in the hydrogel bead system . Similarly, previous studies reported the lack of advert effects of the biocontrol strain P. protegens CHA0 on indigenous microorganisms in situ, despite large fractions of subsequently isolated rhizobacteria displaying sensitivity towards antimicrobial metabolites produced by the biocontrol strain in vitro . This could suggest that sensitive bacteria either colonize distinct niches separated from the antibiotic-producing biocontrol strain in situ or that naturally co-occurring microorganisms sustain toxic levels of antimicrobial metabolites via community-level tolerance mechanisms, such as multi-species biofilm formation or enzymatic inactivation of antimicrobial metabolites . Microbial interactions among co-existing microorganisms have been studied extensively over the past decades to identify and understand the diverse means by which these microbes communicate and compete. Foster and Bell reported that interspecies competition among bacteria isolated from the same sample site was the far most dominant type of interaction . This could suggest that bacteria have evolved intricate sensing mechanisms to respond to danger cues secreted by competing organisms . In our study, we found that levels of the antimicrobial metabolites, DAPG and pyoluteorin, produced by the invading P. protegens were elevated during cocultivation with the four-membered SynCom . This suggests that the biosynthesis of these metabolites is induced in P. protegens as a response to cues secreted by one or more competing members of the SynCom, which we have observed previously in the case of pyoluteorin biosynthesis . The concentration of both DAPG and pyoluteorin after seven days of cultivation exceeded the minimal inhibitory concentration towards several SynCom members . We further discovered that a mutant of P. protegens DTU9.1 deficient in production of DAPG, pyoluteorin, and orfamide A (ΔTriple) was unable to establish and reach cell titers comparable to its wildtype derivative during coculture with the SynCom . However, knockout mutants deficient in either DAPG or pyoluteorin production invaded the SynCom as efficiently as wildtype P. protegens (Fig. S5A). Biosynthesis of DAPG and pyoluteorin is intricately regulated in P. protegens and is interlinked due to its shared dependency on the precursor, phloroglucinol . Thus, one could imagine a compensatory effect upon generation of individual knockout mutants (DAPG levels would increase as a result of knocking out pltA , and vice versa). This was consistent with what we observed during axenic cultivation of the Δ phlACB and Δ pltA mutants (Fig. S5B), which may explain the absence of any noticeable effect on community composition over time when the mutants were introduced into the SynCom. Our results suggest that P. protegens DTU9.1 utilizes its secreted antimicrobial secondary metabolites to establish within the reduced four-membered microbiome. Specifically, we show that DAPG and pyoluteorin are required for establishment, whereas we cannot exclude that orfamide A also play a role in colonization. However, in our hydrogel bead setup orfamide A is inactivated and catabolized by SynCom members, and thus never reaches levels expected to have a significant impact. Lipopeptides from fluorescent Pseudomonas are versatile metabolites known for their antimicrobial properties and involvement in bacterial motility . Particularly, Gram-positive bacteria have been associated with increased susceptibility towards lipopeptides, due to the lack of a protective cell wall , thus it is not surprising that some Gram-positive Actinobacteria have evolved resistance mechanisms towards lipopeptides involving enzymatic inactivation . In our study we discovered that the Gram-positive R. globerulus D757 inactivated orfamide A by hydrolysis of the thermodynamically sensitive ester bond connecting the macrocyclic ring. According to D’Costa and colleagues, hydrolysis of the ester bond is the most common resistance mechanism by which Actinobacteria enzymatically inactivated the cyclic lipopeptide, daptomycin . Although enzymatic inactivation of orfamide A is likely a resistance mechanism in R. globerulus D757, we also demonstrated that hydrolysis caused a significant inhibition of the swarming motility of P. protegens DTU9.1. This indicates that the ability of R. globerulus D757 to hydrolyze the cyclic lipopeptide prevents swarming-mediated spread of P. protegens . We also observed that biotransformation of orfamide A by R. globerulus D757 was a prerequisite for subsequent degradation by S. indicatrix D763 . Two recent studies have similarly demonstrated how biotransformation of Pseudomonas -produced cyclic lipopeptides can have dramatic effects on their chemical and ecological properties . The first found that coculture between a Pseudomonas and a Paenibacillus strain led to the enzymatic modification of syringafactin, thus changing the chemical structure of the lipopeptide resulting in an amoebicidal byproduct . The second study demonstrated that Gram-positive Mycetocola strains could disarm the activity of the mushroom pathogen, Pseudomonas tolaasii , by hydrolysis of the ester bonds in the two Pseudomonas -produced cyclic lipopeptides, tolaasin I and pseudodesmin A, thus preventing pathogenesis . In our study, we identified catabolism of hydrolyzed orfamide A ( and ), which suggests that S. indicatrix D763 might utilize the hydrolyzed orfamide A as an alternative nutrient source. A prior study has demonstrated the ability of isolated soil bacteria to utilize penicillin as sole carbon source to support growth by enzymatically catabolizing the antibiotic . In our hydrogel bead setup, we did not observe any noticeable growth benefit for S. indicatrix D763 resulting from orfamide A catabolism when comparing the two systems (SynCom + WT and SynCom + Δ ofaA ). This is perhaps not surprising considering the availability of nutrients in the supplemented 0.1x TSB, even after seven days. Further investigation of the impact of orfamide A catabolism will be the subject for future studies. Previous research showed that Pseudomonas -produced cyclic lipopeptides are rapidly degraded when added exogenously to non-sterile soil yet remain stable in sterilized soil, which clearly suggests that unknown members of the indigenous soil microbiome possess the ability to transform and/or degrade metabolites with activities relevant for biocontrol . Our discovery and characterization of community-level inactivation and degradation of orfamide A involving two co-occurring soil bacteria in our SynCom provides a mechanistic explanation for this observation. Although the prevalence of such processes in different soil communities is currently unknown, we suggest that biotransformation processes of cyclic lipopeptides and other biocontrol metabolites may contribute to variations in efficiency in biocontrol applications. Collectively, our results illustrate the usefulness of synthetic communities to systematically investigate how microbial communities respond to antibiotics to enhance their resilience towards microbial invasion, and to identify processes that determines the turnover and “fate” of biocontrol metabolites. Improved knowledge of potential constraints in efficient biocontrol is a prerequisite for development of efficient biocontrol products or for development of measures to counteract the community processes responsible for modification and degradation of biocontrol metabolites.
Supplementary_clean_wrae105
|
The use of methylprednisolone after third molar surgery. A systematic review and meta-analysis of randomized controlled trials | c170f51f-f9b7-425b-ad96-6284c62d1fb8 | 11801684 | Dentistry[mh] | Third molar extractions are typically a complex procedure that requires a significant amount of time, experience, and knowledge from the professional. The location of third molars is anatomically complex due to the innervation and vascularization of these teeth. Then, post-operative sequelae, such as edema, pain, and trismus, are expected to appear immediately after surgery. Following the procedure, the worst-case scenario could include: infection, dry socket, paresthesia, and fracture . Usually, the referral for extraction is based on these associated pathologies. In the case of asymptomatic third molars without pathology, it is possible to monitor the patient according to the risk-benefit ratio evaluation . Pain, trismus, and swelling are the most common postoperative complications in this type of surgery. Therefore, understanding that it is an invasive procedure, the professional must provide the patients with the most appropriate pre-, intra-, and post-operative care . The complications are also associated with other factors, such as the degree of tooth impaction, patient's age and health status, surgeon's experience, smoking habits, use of contraceptives, and technique used . Thus, there are different ways of approaching the complications depending on the type of surgery and patient. Corticosteroids are one of the most common medications given for post-operative complications. Corticosteroids are prescribed for various conditions and have a wide range of effects on the human body. They are synthetic analogs of natural steroid hormones produced by the adrenal cortex. Their function and objective is to reduce inflammation by suppressing the immune system , decreasing cellular permeability and capillary dilatation by inhibiting the production of vasoactive substances and diminishing the amount of cytokines . Furthermore, corticosteroids repress the generation of prostaglandins, obtaining an analgesic effect . Corticosteroids can be divided into two groups: glucocorticoids and mineralocorticoids. Glucocorticoids have anti-inflammatory properties with minimal or no influence on the fluid or electrolyte balance . Regarding biodistribution, corticosteroids are immediately absorbed in the gastrointestinal tract, where they vigorously bind to proteins, undergoing hepatic metabolism and renal excretion. These drugs are available in oral, intramuscular, intravenous, intra-articular, topical, and aerosols for inhalation . These drugs are also classified according to the duration of their action. Short-acting corticosteroids contain cortisone and cortisol (hydrocortisone), with an action time of less than 12 hours and an anti-inflammatory potency of one. Intermediate-acting has an action time between 12 and 36 hours; these contain prednisone and prednisolone with an anti-inflammatory potency of 4, and 6-methylprednisolone and triamcinolone have a potency of 5 and are examples of intermediate-acting drugs. Betamethasone and dexamethasone are long-acting glucocorticoids with a time of action greater than 36 hours and a potency of the anti-inflammatory character of 25 hours . Methylprednisolone and dexamethasone are the most often corticosteroids used after impacted third molar surgery to decrease the initial inflammatory response. These can be administered by injection into the surgical area or systemically . Another previous systematic study made the comparison between both drugs (methylprednisolone and dexamethasone); therefore, it restricted the study to those medications only. Then, the primary goal of this systematic study was to evaluate the use of methylprednisolone in the postoperative period of impacted third molars in relation to its efficacy in postoperative pain and edema, dosage regimens, forms of administration, and adverse effects. Secondarily, the aim was to analyze the impact of methylprednisolone after the extraction of impacted teeth compared to no medication use or the use of different drugs from the same pharmacotherapeutic group. This systematic review followed the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement . This study was registered in PROSPERO (CRD42024512561). The PICO (Population/Intervention/Comparison/Outcome) strategy was used to formulate the clinical focus question: “In patients who underwent impacted third molar extraction (P), what was the effect of methylprednisolone used postoperatively (I) compared to non-use or the use of other medications within the same pharmacotherapeutic group (C) to reduce inflammation and pain (O)?” The outcomes observed were: (a) reduction of postoperative pain, (b) postoperative edema, (c) other complications of inflammatory origin, (d) trismus, (e) need for parallel analgesia, and (f) quality of life after extraction. - Eligibility Criteria The inclusion criteria were: 1. Randomized controlled trials (RCTs), 2. published within less than 10 years (2013-2023), 3. with a minimum of 10 patients, 4. published in English, and 5. reported the use of methylprednisolone in the postoperative period of extraction of impacted teeth. The exclusion were: 1. controlled clinical trials (CCTs), clinical studies, reviews, in vivo , in vitro studies, case series, and case reports, 2. studies that ignored the chronic use of medications that could interfere with the results, 3. studies that included patients with non-controllable systemic disease, 4. studies that included smokers, and 5. studies with a lack of information or detail and missing follow-up descriptions. - Search strategy and Data extraction Two independent investigators (HL and BL-A) performed an electronic search in the following databases: PubMed/MEDLINE and Scopus to find articles about methylprednisolone that were related to the extraction of third molars. A manual search was also performed. The combination of specific keywords was applied in each database, associated with Boolean operators: 1. PubMed/MEDLINE: (methylprednisolone) OR (corticosteroids) OR (steroids) AND ((third molar) OR (impacted teeth) OR (impacted tooth)); it was applied filters to adjust the result; 2. Scopus: (methylprednisolone) OR (corticosteroids) OR (steroids) AND ((third molar) OR (impacted teeth) OR (impacted tooth)); with filters: Limited to dentistry/limited article, clinical studies, RCT, CCT, Review, English, and 2013-2023. The agreeability inter-reviewer was assessed using Cohen’s Kappa test. The data were extracted based on the general study design, year of publication, type of study, number of patients included, gender, age, follow-up, detail of the surgeries, dosage of the medication, drug administration route, method of swelling, maximum mouth opening (MMO) evaluation, and pain assessment. - Quality assessment and Statistical analysis Two reviewers (HL and BL-A) independently assessed the quality of the study; in case of disagreement, a third author (TB) was consulted. The risk of bias for the RCTs included was performed using a revised Cochrane risk-of-bias tool for randomized trials (RoB2). The following domains were observed: the randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of reported outcomes. The low risk of bias was represented by the green color and indicated that the study implemented robust measures, minimizing the risk of bias in the respective domain; uncertain risk of bias (yellow) suggested that the study lacks clear information or details, making it difficult to determine the risk of bias; high risk (red) showed the study had significant limitations or flaws in the design or execution, increasing the risk of bias. Suppose all parameters were filled with low risk (green) or until two unclear (yellow), the overall result was a low risk of bias (green). For results with only one high risk (red) and up to two unclear (yellow), the result was a moderate risk of bias. Whereas if filled with 2 or more high risks (red) and/or more than 2 unclear risks present (yellow), the overall result will be a high risk of bias . The random effect model was used for the meta-analysis to evaluate the variables. Heterogeneity was analyzed using Cochran's Q test and Higgins' I2 statistics. Standardized mean differences were used to measure the effect. The statistical analyses were carried out using Review Manager 5.4 software. A comparative analysis of the studies' results was used to analyze variables where meta-analysis was not possible. The search queries identified 158 studies; 149 were excluded due to duplicates and/or did not meet the predefined eligibility criteria. These were excluded after the initial evaluation of the title and abstract. Nine articles were included for full-text analysis, resulting in the nine articles included in this systematic review that met the eligibility criteria (k=0.92) (Fig. ). - Study characteristics and General assessment The data was collected and exposed in . The 9 RCTs included evaluated 354 patients, with a mean age of 25.53 years; 475 surgeries were developed. Except for Chugh et al .’s and Selvaraj et al .’s studies, female patients prevailed. The route of drug administration varied through oral, muscle, submucosal, or intravenous . There was great variability in sample size, with one study having a sample of 10 patients and another with 65 patients, giving a mean value of 39.33 ± 21.18 patients per study. The average age of the patients is relatively close between the studies, except for Alcantara et al .’s study, which has the lowest average value, 20.3, and Koçer et al .’s study, which has the highest average value, 29.6. The number of surgeries reported in the studies varied between 20 and 104 (mean of 52.78 ± 24.06), which reveals a disparity in the number of surgeries analyzed in the studies. Five studies compared methylprednisolone and placebo results in the postoperative period after impacted third molars extraction. They assessed the postoperative trismus, examined the patients' range of mouth opening after impacted third molar extraction surgery, and noted any restriction in mandibular movement (subjective appraisal). In addition, those authors also evaluated the postoperative pain, measuring the intensity and duration reported. Standard pain assessment scales were used, allowing a more in-depth understanding of the effects of methylprednisolone on postoperative pain relief compared to the control/placebo group. Furthermore, the same studies also evaluated postoperative swelling, which permitted observing the facial swelling levels after extraction. The soft tissues around the surgical region were examined, noting any noticeable increase. Five studies compared methylprednisolone and dexamethasone in the postoperative period. Postoperative pain was assessed exclusively using the Visual Analog Scale (VAS); edema was evaluated through the presence and magnitude of its dimension; and trismus was measured through the mouth opening amplitude. These measurements were carried out using standardized assessment instruments, recording the distance between anatomical reference points in the mandible and maxilla. This assessment method allowed a quantitative analysis of trismus, providing objective data on the extent of mouth opening limitation following impacted third molar extraction. Regarding the routes of administration for methylprednisolone in the postoperative period, three studies performed the comparison. Koçer et al . evaluated the oral, intravenous, and masseter routes; and Gholami et al . and Selvaraj et al . investigated the gluteal and masseter muscle routes. This comprehensive analysis of these studies aims to examine the relative efficacy of methylprednisolone administered by different vias, offering valuable insights for clinical practice by identifying the most effective route of administration to optimize the control of postoperative pain, edema, and trismus. - Clinical outcomes shows the details of the parameters analyzed. summarizes the inter-incisal reduction/MMO and shows the outcomes obtained comparing methylprednisolone and the control group. Koçer et al .’s and Chugh's results show that the experimental group had less reduction and decreased over time. On the other hand, in the 7-day results of Chugh et al . and Darawade et al . , the average reduction was higher in the experimental group. summarizes the studies that reported results on inter-incisal reduction using only methylprednisolone (single group). Comparing methylprednisolone and dexamethasone , the pain increased over time in both groups in the Alcantara et al . study ; in the Srivastava et al . and Darawade et al . studies, it decreased in the methylprednisolone group and remained the same in the dexamethasone group over time. For the swelling analysis, show the results for Tragus-Commissure and Canthus-Gnathion edema, respectively, reported in Koçer et al .'s study . For Tragus-Commissure, the experimental group has lower average values than the control group in all three studies, and in both groups, the results decrease over time. For Canthus-Gnathion, the experimental group showed higher results on average than the control group in all three studies; in both groups, the results decreased over time. - Quality assessment and Statistical analysis Fig. presents the risk of bias for the included RCTs. Three studies had a low risk of bias, 2 had a moderate risk, and 4 had a high risk. Due to limitations in the included literature, the meta-analysis was only done for specific parameters. The careful selection of parameters was based on the availability and comparability of data, guaranteeing the robustness and validity of the results obtained. Fig. shows the pain results after 24 hours and 7 days, comparing the result of methylprednisolone with placebo (control). After 24 hours, Cochran's Q ( p =0.15>0.05) and I2=47%, it was observed moderate heterogeneity; the forest plot shows that the effect of the meta-analysis 0.05 (95%CI [-0.45;0.55]) was not statistically significant ( p =0.85). After 7 days, the forest plot showed the effect of the meta-analysis -0.08 (95%CI [-0.58;0.42]), which was also not statistically significant ( p =0.85). Analyzing the data for the comparison between methylprednisolone and dexamethasone, Cochran's Q ( p =0.31>0.05) and I2=3%, it was concluded there was homogeneity between studies after 24 hours; the forest plot shows that the effect of the meta-analysis 0.37 (95%CI [-0.14;0.88]) was not statistically significant ( p =0.16). Fig. shows the inter-incisal/MMO reduction results at 48 hours and 7 days. Given the results, Cochran's Q ( p =0.96>0.05) and I2=0% (A) and Q ( p =0.37>0.05) and I2=0% (B), it was concluded that there was homogeneity between studies about inter-incisal/MMO reduction at 48 hours and 7 days. The forest plot presented the most significant reduction occurred in the methylprednisolone group, and the effect was 0.65 (95%CI [0.29;1.01]) (A) and 0.44 (95%CI [0.09;0.80]) (B), respectively, statistically significant result ( p <0.01 and p <0.05). The aim of this systematic review was to assess the effect of methylprednisolone in patients undergoing third molar surgery, with a specific focus on post-operative trismus, pain, and edema. This analysis was carried out considering the predefined criteria, grouping all available randomized controlled trials (RCTs) that investigated the use of methylprednisolone in comparison with a control group, as well as in comparison with dexamethasone. For the dosage of corticosteroids to be effective, it must exceed the physiological production of the human body (5-30mg/day) . The doses used in this review were 20 mg and 40 mg. The results of this review showed that methylprednisolone was efficient in reducing pain, edema, and trismus. It is similarly observed in the literature, where the corticosteroids, e.g., dexamethasone and methylprednisolone, showed the ability to inhibit the immune system. It is an anti-inflammatory with analgesic and anti-allergic potential. Moreover, the toxicology proved that both are safe at low dosages via oral or other use routes . Therefore, careful instructions should be passed to the patients when using glucocorticoids. The literature has shown that methylprednisolone, at standard doses, was considered more effective than dexamethasone and prednisolone for treatments and five times more potent than hydrocortisone . It suggests that methylation of prednisolone makes it more potent by aiding interaction with a cellular target rather than increasing its stability. Otherwise, the long-term use of these medications and when used in high dosages, may cause some undesired side effects (steroid osteoporosis, steroid diabetes, delayed wound healing, and are closely associated with mortality) . Methylprednisolone exhibited linear plasma protein binding (average of 77%) and was moderately distributed into tissue spaces; however, the volume of distribution of prednisolone is only one-half that of methylprednisolone, proving the superiority of this drug . Within this context, the results of the present review revealed after 48 hours and 7 days that dexamethasone had lower inter-incisal/MMO reduction compared to methylprednisolone (respectively, p <0.01 and p <0.05). Agrawal et al .'s results agree with the evidence presented herein, reinforcing the efficacy of corticosteroids. This consistency strengthens the evidence of the positive potential of corticosteroids in mitigating postoperative trismus, providing additional support for their effectiveness in promoting a more comforTable and effective recovery after third molar surgery . Regarding pain, after 24 hours, there was no significant effect between methylprednisolone and dexamethasone ( p =0.16). This highlights the complexity of post-operative pain, which is influenced by different factors such as surgical trauma and the individual inflammatory response and threshold. Variability in pain administration and measurement methods between studies may also have contributed to these results. Furthermore, the prophylactic oral intake of dexamethasone 8 mg was compared to 4 mg to verify the control of the postsurgical edema after third molar extractions; 8 mg presented greater efficacy in the control of postsurgical edema . The literature also suggests other medications, such as the postoperative administration of 30 mg prednisolone; it demonstrated relief for trismus, swelling, and pain compared to non-steroidal anti-inflammatory (NSAI) (control group); this fact shows the potential of prednisolone. There was a significant reduction of swelling ( p <0.05) and higher MMO and visual analogue scale (VAS) for the use of prednisolone, without any disturbance of wound healing or other corticosteroid-related complications . In addition, another medication typically prescribed is ibuprofen (NSAI). Ibuprofen and methylprednisolone are the two drugs largely used in this type of surgery (third molar extraction) ; Schultze-Mosgau et al . showed that the combination of both medications allowed to obtain better analgesic and anti-inflammatory effects than the separate use. López Carriches et al . compared methylprednisolone and diclofenac in reducing trismus . Similarly, Bamgbose et al . also verified the absence of differences in the control of trismus when they compared a group medicated with dexamethasone and diclofenac versus using only diclofenac . On the other hand, the results obtained by Troullos et al . supported the existence of differences between the tested groups, with considerably lower trismus values in patients medicated with methylprednisolone compared to those medicated with ibuprofen . Comparing the methylprednisolone with the control groups included, the pain parameter was observed after 24 hours and 7 days; in both periods, no significant effect was observed in favor of methylprednisolone ( p =0.85). This suggests that methylprednisolone may not be more effective than the control assessed in reducing pain at these time intervals. However, it is essential to consider that postoperative pain is a subjective experience and can be influenced by various factors (such as the type of surgery performed, the patient's psychological state, the presence of pre-existing conditions, the quality of pain management, and the level of social support) . Subgroup analyses may provide additional insights into the efficacy of methylprednisolone in different populations and surgical contexts . Another study also reported no significant differences in pain between the groups treated with methylprednisolone and the control groups at the same period. These results highlight the need to consider postoperative pain as a multifaceted experience influenced by a variety of factors and the importance of more detailed analyses to understand the effects of methylprednisolone on postoperative pain management fully. In 2023, a systematic study recently compared methylprednisolone versus dexamethasone's efficacy in managing post-surgical pain, swelling, and trismus after surgery. Some methodological differences were observed compared to our study: 1. the present systematic study compared the results of methylprednisolone with any other medication or no medication, while the other systematic included articles that performed only the comparison between dexamethasone versus methylprednisolone; 2. the authors included other databases besides what was used in the present study, which can obtain a different inclusion; 3. the authors considered articles published without publication date or language restrictions, whereas the present study included some limitations (articles published in the last 10 years [2013-2023] and English language); and 4. there was differences in the eligibility criteria comparing both systematic studies. Otherwise, a similar result was found for dexamethasone, which had statistically significant and better results than methylprednisolone in reducing trismus. - Limitations of the Study As a limitation, it is possible to verify that a low number of RCTs have been developed on the subject. Also, many of the articles needed to have more crucial values. Several studies did not report complete measurements, and many did not include standard deviations, which made it impossible to include these studies in the meta-analysis. In addition, the divergence in the measures used between the different studies represented a significant challenge. For example, when assessing trismus, we opted to use the inter-incisal reduction measure since most of the available articles employed this specific method, providing greater consistency and comparability between studies. However, when trying to convert data from articles that measured maximum mouth opening to the inter-incisal reduction measure, we often lost the standard deviation value, making it impossible to calculate this crucial statistic for the meta-analysis accurately. Similarly, there was significant variability in the measurement methods used between the different studies in assessing edema. This methodological heterogeneity and the lack of reported standard deviations made it difficult to harmonize the data, limiting the possibility of carrying out a comprehensive and comparative meta-analysis. - Final considerations Within the limitations of this study, it was possible to verify that methylprednisolone efficiently treats patients safely after third molar extraction, reducing pain, edema, and trismus. There was a positive trend for edema reduction via masseter muscle; however, there was no significant difference or comparable data available regarding pain, the best route of administration, adverse effects, and patient post-operative comfort. Methylprednisolone achieved better inter-incisal level/MMO results than dexamethasone; otherwise, dexamethasone is preferable in minimizing postoperative trismus, presenting superior potential in this specific clinical context. All data should be carefully analyzed due to the study's limitations, the small number of studies included, heterogeneity found, and the risk of bias present. More RCT studies are recommended to confirm the findings of this study. |
Molecular Epidemiology Clinical Manifestations, Decolonization Strategies, and Treatment Options of Methicillin-Resistant | 46638bed-fa9e-4bd0-9fc1-7c5280bcaa34 | 11858580 | Biochemistry[mh] | Staphylococcus aureus ( S. aureus ) is a common pathogen in humans, initially identified by Ogston . S. aureus resistance to penicillin was first reported in the 1940s , whereas methicillin-resistant S. aureus (MRSA) was discovered in U.K. hospitals in 1961 . A case of neonatal MRSA infection was described for the first time in the U.S. in 1981 in a neonate with osteomyelitis . MRSA is a significant antibiotic-resistant strain of S. aureus that, currently, predominates in neonatal intensive care units (NICUs) globally and is associated with significant neonatal morbidity and mortality . Several NICUs have reported epidemiological data on neonatal MRSA after surveillance measures, transmission control, and decolonization policies . While the rate of methicillin-susceptible S. aureus infections was significantly reduced over the past 20 years, the rate of MRSA infections has remained consistent at about 10 per 10,000 hospitalized neonates . Moreover, late-onset MRSA infections in NICUs have significantly increased by 300%, from 0.7 to 3.1 cases/10,000 days, between 1995 and 2004 . In addition, the percentage of hospital-acquired (HA)-MRSA infections in intensive care units in the U.S. increased from 35.9% to 64.4% between 1992 and 2003 . Nearly 1.7 million HA-MRSA infections occur each year in U.S. hospitals, including more than 33,000 cases in NICUs, according to the Center for Disease Control and Prevention (CDC) . Until 2016, there had been reported more than 20 MRSA outbreaks in NICUs worldwide, most of which from the U.S. and Europe . Recent epidemiologic studies have also reported the evolution of MRSA clones, underscoring the growing resistance of MRSA to antimicrobial agents . Given these challenges, a comprehensive analysis of MRSA’s epidemiology, molecular characteristics, and toxicity is required to reduce the incidence of MRSA infections in NICUs. Our aim was to review the existing evidence and provide novel insights into the molecular characteristics, epidemiology, risk factors, clinical manifestations, decolonization strategies, and treatment options of MRSA infection in neonates. Our study is organized into (1) exploring the molecular characteristics of MRSA, (2) providing the epidemiological data of MRSA burden, (3) reviewing the risk factors and clinical manifestations of MRSA infection in neonates, (4) providing a summary of existing recommendations for the decolonization strategies and treatment options of MRSA disease, and (5) discussing challenges related to MRSA infections in neonates and directions for future study . The Literature Search Strategy A literature search of Pubmed was conducted by two researchers in November 2024. Only human studies and English-language articles were considered. The terms ‘Methicillin-resistant Staphylococcus aureus ’ OR ‘MRSA’ AND ‘neonate’ OR ‘newborn’ OR ‘infant’ OR ‘Neonatal Intensive Care Unit’ OR ‘Neonatology’ were used. The retrieved studies were assessed according to their titles, abstracts, and suitability for this narrative review. A literature search of Pubmed was conducted by two researchers in November 2024. Only human studies and English-language articles were considered. The terms ‘Methicillin-resistant Staphylococcus aureus ’ OR ‘MRSA’ AND ‘neonate’ OR ‘newborn’ OR ‘infant’ OR ‘Neonatal Intensive Care Unit’ OR ‘Neonatology’ were used. The retrieved studies were assessed according to their titles, abstracts, and suitability for this narrative review. HA-MRSA is transmitted within hospital settings and was formerly responsible for the majority of MRSA infections. Until the 1990s, MRSA was considered common exclusively in healthcare environments , but for the first time in the 1990s, MRSA was discovered in patients who had not been previously hospitalized or had any history of contact with HA-MRSA carriers . Since then, epidemic outbreaks of community-acquired (CA)-MRSA infections have been reported worldwide . The genetic investigations of these strains revealed that different strains of MRSA were present in the community compared with healthcare settings . However, recent reports of neonatal MRSA infections revealed that 15–21 detected MRSA strains shared common microbiological traits with strains that have surfaced in the community, indicating that CA-MRSA has emerged in healthcare settings, including NICUs . HA-MRSA and CA-MRSA differ from one another in terms of genotype and phenotype . S. aureus genome consists of core and accessory components. The core genome includes genes that are found in every isolate and comprise nearly 75% of the S. aureus genome, whereas the accessory genome, which comprises nearly 25% of the S. aureus genome, is responsible for the large proportion of MRSA’s genetic variability. The accessory genome consists of mobile genetic elements (MGEs) that are transferred between strains, including plasmids, chromosomal cassettes, transposons, bacteriophages, and pathogenicity islands, and contains virulence, immune-evasion, and drug resistance mediators. Therefore, the accessory genome is frequently more strain-specific and varied compared to the core genome . As MRSA is both a commensal and a pathogen, there is great interest in determining whether identifying MRSA colonization and attempting to eradicate carriage will lower the risk of recurrent infection. The pulsed-field gel electrophoresis (PFGE) typing system can be used for the molecular classification of MRSA strains, especially during outbreaks . PFGE uses a restriction enzyme to break the bacterial DNA, which is then subjected to an electrical gradient and produces a distinctive banding pattern . PFGE is extremely discriminatory and widely accessible; however, interlaboratory variability, technical demand, and difficulties in long-term epidemiology limit its use . According to the Atlanta CDC typing scheme, most CA-MRSA infections in the U.S. have been associated with two types of PFGE, USA300 and USA400, which differ from the commonly found HA-MRSA genotypes . USA300 is the most common strain related to MRSA infections in previously healthy neonates in the U.S. and Europe, and it is produced when an ancestral sequence type (ST) 8 strain absorbs an element that catabolizes arginine, the staphylococcal cassette chromosome (SCCmec) type IV, and a Panton-Valentine leukocidin (PVL)-encoding locus . The arginine-catabolic element increases the ability of the USA300 MRSA strain to elude the immune system and survive inside the host, while SCCmec confers antibiotic resistance and PVL increases invasiveness . There have been prior reports of USA400-caused outbreaks of skin and soft tissue infections (SSTIs) in term neonates associated with transmission in neonatal nurseries and postnatal wards . Finally, the gene locus sasX has been recently detected in MRSA clones such as ST5 and may be involved in nasal colonization, lung infections, and the development of abscesses . The supplementation of PFGE with multilocus sequence typing (MLST) provides additional comparisons with sequences described in available databases. MLST is highly repeatable, discriminate, and appropriate for long-term worldwide epidemiology. A standardized MLST database defines each strain’s alleles, while allele combinations identify strains . Spa typing is based on sequence-based analysis of the sequence region of the spa gene of polymorphism X. It is rapid and ideal for outbreak investigations. References to extensive foreign databases are made to the sequences . The multilocus variable number of tandem repeat analysis evaluates the variation in the number of repeat DNA sequences, and it is high-throughput and low-cost . SCCmec typing is based on PCR and assigns distinct SCCmec types to specific allotypes of the mecA and ccr genes. Compared to whole-genome sequencing or MLST, it is less expensive. Divergent and developing SCCmec types that are not detectable by present approaches have been found using SCCmec typing to MRSA . Types that are more prevalent in hospital or community settings can be identified by using the SCCmec typing . HA-MRSA strains have been associated with SCCmecA types I–III, while CA-MRSA strains have been associated with SCCmecA types IV, V, and VII . These different SCCmecA types carried by HA-MRSA or CA-MRSA strains provide distinct patterns of antibiotic resistance by encoding the penicillin-binding protein 2A (PBP2A) . Unlike the HA-MRSA phenotype, CA-MRSA is susceptible to various antibiotics, including clindamycin, quinolones, and trimethoprim-sulfamethoxazole . Repetitive element palindromic PCR is a genotyping method that identifies repetitive DNA sequences dispersed across the MRSA genome. STAR gene restriction profile analysis is based on PCR amplification and restriction enzyme digestion and generates restriction profiles that differ according to the intergenic regions’ sequence within the PCR product . Finally, whole-genome sequencing analyzes the entire genome sequence for single-nucleotide variants . The resolution offered by whole-genome sequencing has allowed us to determine that individuals who were colonized by circulating community strains later introduced those strains into hospitals, which ultimately led to the intermixing of CA-MRSA and HA-MRSA strains . MGEs and essential genome components were discovered in later efforts to define the elements that contributed to those strains’ success. Differences in phenotypic characteristics between HA-MRSA and CA-MRSA are also underscored by the carriage of PVL and the greater expression of additional toxins in CA-MRSA compared to HA-MRSA strains . Epidemiologic studies during the last decade revealed that specific clonal lineages, namely ST 1, 5, 8, and 22, are responsible for most neonatal MRSA infections globally . Many of these clones, such as ST8 and ST22, the two most prevalent MRSA clones in the U.S. and Europe , are characterized by genetic plasticity and hold the ability to produce toxins and biofilms . Collagen-binding proteins, fibronectin, elastin, and clumping factors all contribute to MRSA’s ability to aggregate on host tissues and indwelling devices . Among several MRSA-produced toxins, PVL leukotoxins, LukED, LukAB/DH, hemolysins, exfoliative toxins, enterotoxins, phenol-soluble modulins, and toxic-shock syndrome toxin-1 are the most important . While it remains unclear how the neonatal immune system reacts to MRSA infection , the production of toxins has been related to the generation of proinflammatory cytokines leading to hyperinflammation and tissue injury . This unbalanced inflammation is also considered the major underlying mechanism associated with long-term neonatal morbidities, such as cerebral palsy, retinopathy of prematurity, necrotizing enterocolitis, and bronchopulmonary dysplasia . Antimicrobial Resistance Clinical diagnosis now relies on the availability of sensitive and precise techniques for accurately identifying antibiotic resistance in MRSA. Molecular typing is warranted because phenotypic typing techniques are extremely reliant on growth conditions and are not capable of reliable discrimination . MRSA has acquired MGEs, including insertion sequences, transposons, and, occasionally, plasmids containing genes for antibiotic resistance to penicillin ( blaZ ), erythromycin ( ermC ), clindamycin ( ermC ), trimethoprim ( dfrA and dfrK ), and tetracyclines ( tetK and tetL ) . The mecA gene, which codes for PBP2A, may be a useful molecular marker of MRSA . The resistance to methicillin is mainly attributed to the overexpression of PBP2A, which has a low affinity for β-lactam antibiotics; nevertheless, other mechanisms, such as efflux pumps, are also associated with methicillin resistance . Antibiotic resistance in HA-MRSA strains is also genetically associated with resistance in disinfectants or heavy metals such as quaternary ammonium, mercury, or cadmium, probably reflecting the high selection pressures present in the hospital setting . Vancomycin-resistant S. aureus (VRSA) was initially detected in Japan in 1996, but numerous reports thereafter indicate that it has since spread. Given the extensive use of vancomycin to treat MRSA infections, the most worrisome genetic adaptation in S. aureus to date is the development of resistance to this antibiotic. There are two types of vancomycin resistance in S. aureus . Long or multiple courses of vancomycin often result in the emergence of vancomycin-intermediate S. aureus (VISA) strains. Several distinct mutations within a population result in varying levels of vancomycin resistance. The majority of mutations found in VISA isolates change essential genomic elements involved in cell wall production and autolysis. Cross-resistance to daptomycin is also conferred by a number of these mutations, such as those in yycH , mprF , and dltA . Unlike VISA, it has been demonstrated that plasmid transfer of the vanA operon from vancomycin-resistant Enterococcus faecalis results in VRSA . Complete resistance to vancomycin is achieved when vancomycin molecules are trapped due to the thicker cell wall and obstructing the peptidoglycan plexus, which serves as a physical barrier against vancomycin molecules. Clinical diagnosis now relies on the availability of sensitive and precise techniques for accurately identifying antibiotic resistance in MRSA. Molecular typing is warranted because phenotypic typing techniques are extremely reliant on growth conditions and are not capable of reliable discrimination . MRSA has acquired MGEs, including insertion sequences, transposons, and, occasionally, plasmids containing genes for antibiotic resistance to penicillin ( blaZ ), erythromycin ( ermC ), clindamycin ( ermC ), trimethoprim ( dfrA and dfrK ), and tetracyclines ( tetK and tetL ) . The mecA gene, which codes for PBP2A, may be a useful molecular marker of MRSA . The resistance to methicillin is mainly attributed to the overexpression of PBP2A, which has a low affinity for β-lactam antibiotics; nevertheless, other mechanisms, such as efflux pumps, are also associated with methicillin resistance . Antibiotic resistance in HA-MRSA strains is also genetically associated with resistance in disinfectants or heavy metals such as quaternary ammonium, mercury, or cadmium, probably reflecting the high selection pressures present in the hospital setting . Vancomycin-resistant S. aureus (VRSA) was initially detected in Japan in 1996, but numerous reports thereafter indicate that it has since spread. Given the extensive use of vancomycin to treat MRSA infections, the most worrisome genetic adaptation in S. aureus to date is the development of resistance to this antibiotic. There are two types of vancomycin resistance in S. aureus . Long or multiple courses of vancomycin often result in the emergence of vancomycin-intermediate S. aureus (VISA) strains. Several distinct mutations within a population result in varying levels of vancomycin resistance. The majority of mutations found in VISA isolates change essential genomic elements involved in cell wall production and autolysis. Cross-resistance to daptomycin is also conferred by a number of these mutations, such as those in yycH , mprF , and dltA . Unlike VISA, it has been demonstrated that plasmid transfer of the vanA operon from vancomycin-resistant Enterococcus faecalis results in VRSA . Complete resistance to vancomycin is achieved when vancomycin molecules are trapped due to the thicker cell wall and obstructing the peptidoglycan plexus, which serves as a physical barrier against vancomycin molecules. Over the past few decades, there has been a conceptual evolution in the epidemiology of S. aureus . Hospitalized neonates have a higher rate of MRSA colonization than the general neonatal population, with rates varying from 0.3% to 32% among institutions , explained by the 2–3 times higher MRSA carriage rate among healthcare workers compared with the general population . In a previous systematic review based on studies from high-income countries in Europe and the Western Pacific Region, the pooled prevalence of MRSA carriage was 9.5% among healthcare providers . Similarly, another systematic review in South Asia reported an MRSA carriage rate in healthcare providers at 9.23% . Of note, a systematic review of 22 studies in Iran reported a much higher rate of 32.8% of nasal MRSA carriage among healthcare providers . Individual studies reported a prevalence of MRSA colonization in healthcare providers of 27.91% in Jordan , 26.47% in India , 17.65% in Nigeria , and 16.22% in Greece . The prevalence of nasal MRSA colonization in neonates usually ranges from 2% to 4%, but it might reach up to 8% during an MRSA epidemic investigation . In addition, previous studies have found that between 0.6% and 8.4% of neonates were colonized or infected with MRSA during respective periods . According to Huang et al., MRSA colonized 5.2%, methicillin-sensitive S. aureus colonized 12%, and S. aureus colonized 17% of the neonates’ nasal cavity overall . In a previous study in China that evaluated the burden of S. aureus at the time of admission to the NICU, 17% of neonates had nasal colonization with MRSA or methicillin-sensitive S. aureus at the time of admission compared to 13% of neonates in a NICU in Taiwan (13%) , 10% in Japan , and 3.8% in the U.S. . According to a systematic review of 62 studies conducted between 2001 and 2023 in the U.S., Japan, South Korea, Brazil, Taiwan, and other countries, the Western Pacific region had the highest rate of MRSA colonization (19.8%), whereas America had the lowest (3.1%) . A cumulative incidence of 1.4% to 3% was observed for CA-MRSA infections, while a cumulative incidence of 9.5% to 9.8% was observed for HA-MRSA infections. There were regional differences, with Taiwan having the highest prevalence at 23.8% and Brazil having the lowest at 0.9%. Compared to the U.S., South Korea had a greater HA-MRSA (21.9% compared to 2.9%) and CA-MRSA incidence (8.5% compared to 1.5%) . The yearly incidence rate of MRSA colonization in China varied between 5.66 and 7.66 cases per 1000 admissions . According to data from 33 centers across 11 Latin American nations for the Tigecycline Evaluation and Surveillance Trial, the total prevalence of MRSA among S. aureus isolates was 48.3% between 2004 and 2007 . The SENTRY Antimicrobial Surveillance Program in Latin America also revealed that the frequency of MRSA among staphylococcal infections in medical centers increased from 33.8% in 1997 to 40.2% in 2006 . Finally, a recent meta-analysis by Zervou et al. reported that the pooled prevalence of MRSA colonization was 2.3%, based on 11 studies that were conducted in the U.S., compared to 1.3% based on studies conducted in Asia . The prevalence of MRSA colonization on admission was 1.5%; interestingly, the prevalence of MRSA colonization in outborn neonates was 5.8% compared with 0.2% in inborn neonates. The incidence of invasive S. aureus infections was more than 25% in 8 out of 30 European countries . Romania, Malta, Portugal, Cyprus, Greece, Italy, Slovakia, and Spain reported an incidence of >25%, while Hungary, Croatia, and Ireland had an incidence higher than the European Union population-weighted mean of 16.8% . According to previous reports, MRSA was responsible for 33–67% of S. aureus infections in neonates, and among neonates with MRSA colonization, one-fourth developed MRSA infections . During a 20-year study period in Western Australia, S. aureus sepsis was responsible for about 4% of blood culture-positive infections in neonates, with an overall incidence of 0.10/1000 live births. Infants born before 32 weeks of gestational age had a much greater incidence of S. aureus sepsis (6.87/1000 live births) than infants born after 32 weeks of gestational age (0.03/1000). Between 2001 and 2010, the frequency of S. aureus sepsis was 0.13 per 1000 live newborns, and between 2011 and 2020, it was 0.07 per 1000 live births. MRSA was responsible for 26% of cases, whereas methicillin-sensitive S. aureus was responsible for 74% . 5.1. Colonization Neonates in the NICU are a particularly vulnerable population. Specific innative and environmental factors, such as the immature neonatal immune system, exposure to multiple invasive procedures, prolonged hospitalization, and close contact with healthcare providers, predispose neonates to MRSA colonization and infection . Neonates may acquire S. aureus through the birth canal , while a concurrent maternal infection may be present in up to 20% of newborns with MRSA infection . The rate of vaginal MRSA colonization among pregnant women has been estimated to be 2.8% in previous reports , and vaginal delivery has been associated with increased risk of S. aureus neonatal transmission . In a previous systematic review, the pooled proportion of MRSA carriage among neonate mothers was 2.1% , with studies from Jordan reporting a prevalence of 9.72% MRSA colonization , Egypt of 1.69% , Brazil of 1.39% , and Germany of 0.51% . Vertical transmission of MRSA has also been indicated by the association of maternal MRSA chorioamnionitis with neonatal MRSA sepsis . Female sex and multiple gestation are additional risk factors for MRSA colonization and infection , whereas antibiotic administration in the week before delivery has been associated with a lower risk of MRSA transmission . After birth, neonates are exposed to S. aureus following contact with adult skin . The carriage of S. aureus among adults ranges from 30% to 70%. MRSA has been shown to spread horizontally through contact with healthcare providers or the hospital setting . Other factors, including NICU overcrowding and understaffing, have been associated with a higher risk of colonization and transmission and could result in MRSA outbreaks . It has also been demonstrated that mothers can vertically transmit MRSA to their infants through breast milk , while fathers can through direct contact with their infants . Controlling transmission is more challenging in intensive care units because S. aureus can persist on ambient surfaces for extended periods . Among neonatal factors, prematurity and low birth weight are the main risk factors for MRSA colonization and infection . Numerous studies have shown that low birth weight was associated with an increased risk of MRSA colonization and/or infection . The prevalence of MRSA infection in extremely low-birth-weight neonates was estimated at 53.4 per 10,000 infants, which was much higher compared to 23.2, 7.9, and 5.0 per 10,000 infants in very-low-birth-weight, low-birth-weight, and appropriate-birth-weight neonates, respectively . Neonates of lower gestational age were also more susceptible to being positive in nasal or both nasal and groin MRSA colonization, compared to neonates of higher gestational age who were positive on groin swabs only . Long-term ventilation support, intravascular catheters, antibiotic administration, total parenteral nutrition, and surgical interventions are additional risk factors for MRSA infections . In the NICU, neonates frequently need several procedures during their hospital stay, including endotracheal tube insertion, mechanical ventilation, central vascular catheterization, and surgery . A previous study demonstrated that compared with non-MRSA-colonized infants, MRSA-colonized infants who experienced a greater incidence of late-onset sepsis were more likely to be intubated or mechanically ventilated . A recent review of S. aureus outbreaks in neonatal intensive care units conducted in Leeds, U.K., identified that MRSA bacteremia was more likely to occur in infants with discharge skin lesions, prior abdominal surgery, current MRSA colonization, and Broviac or peripherally implanted central catheter lines . An increased risk of MRSA infection has also been associated with feeding practices such as parenteral nutrition and gavage feeding . Moreover, longer hospital stays and kangaroo care were other independent risk factors for MRSA infection. Among all risk factors, MRSA colonization is the most significant risk factor for developing MRSA infection in neonates. According to Huang et al., MRSA-colonized neonates had a considerably greater rate of MRSA infection (26%) than non-colonized neonates (2%) , whereas, as suggested by the metanalysis by Zervou et al., colonized neonates have a 24.2-fold higher chance of contracting an MRSA infection while in the NICU compared with non-colonized newborns . 5.2. Clinical Manifestations Neonates may be colonized with MRSA within a median of 9 days from admission, with a range of 1–91 days . In addition, the median interval between MRSA colonization and infection is 4 to 9 days . In a previous report from a New York NICU, MRSA colonization was detected at a median of 17 days, with a range of 4 to 159, whereas nearly two-thirds of the neonates developed colonization during the first 3 weeks of life . MRSA infection has a significant clinical impact. The most common manifestations of MRSA infections in neonates are SSTIs; however, invasive diseases have also been reported . Most cases of invasive MRSA disease (75%) are associated with bacteremia . Late-onset sepsis is a common clinical manifestation of MRSA infection and can range from a moderate focal infection to severe invasive disease . According to reports, late-onset newborn sepsis can raise mortality from 7% to 18% and lengthen the inpatient stay by three weeks . Among neonates with S. aureus bacteremia in a ten-year retrospective research in the U.K. between 1993–2003, MRSA was detected in nearly one-third of neonates . Similarly, among neonates with S. aureus bacteremia in the U.S., MRSA was found in 47% of the cases, indicating that MRSA has emerged as a major cause of neonatal sepsis . Finally, Dolapo et al. reported that the prevalence of MRSA bloodstream infections in neonates increased from 24% to 55% between 2000 and 2009 . Infectious endocarditis, abscesses in the myocardium, liver, spleen, or kidneys, necrotizing pneumonia, osteomyelitis, myositis, meningitis, toxic shock syndrome, septic thrombophlebitis, venous thrombosis, sustained bacteremia, ocular infections, and Waterhouse-Friderichsen syndrome are just a few of the numerous MRSA manifestations that have been reported . CA-MRSA infections typically manifest as SSTIs, in contrast to HA-MRSA infections, although more severe invasive manifestations can also occur . In comparison with HA-MRSA, CA-MRSA contains the virulence genes lukS-PV/lukf-PV that generate PVL, and produce a pore-forming cytotoxin that causes leukocyte death and tissue necrosis . In the U.S., CA-MRSA was the most common cause of SSTIs . The clinical manifestations of SSTIs can vary from cellulitis or a simple abscess to more serious soft-tissue infections such as necrotizing fasciitis, pyomyositis, and mediastinitis as a consequence of retropharyngeal abscess . When term neonates have localized only pustulosis with no signs or symptoms of sepsis, lumbar puncture is not required . Careful patient monitoring and prompt access to microbiological and laboratory tests are crucial because the clinical symptoms and indicators at the beginning of MRSA infections can be non-specific . In addition, MRSA-infected newborns may have a higher readmission rate and a longer course of infection than methicillin-susceptible S. aureus cases ; however, there appears to be no difference between MRSA and methicillin-susceptible S. aureus in terms of clinical presentation and mortality . In very immature preterm neonates, MRSA infections increase the risk of unfavorable short- and long-term outcomes, as well as mortality . The mortality rate of MRSA infections ranges from 2.9% to 28%, with significant variation across institutions . According to earlier research, the case fatality risk of neonatal MRSA sepsis ranged between 9.5% and 55% . A previous study also reported that among MRSA infections, sepsis had a mortality rate of 16%, pneumonia of 32.1%, and necrotizing enterocolitis of 27.3% . Neonates in the NICU are a particularly vulnerable population. Specific innative and environmental factors, such as the immature neonatal immune system, exposure to multiple invasive procedures, prolonged hospitalization, and close contact with healthcare providers, predispose neonates to MRSA colonization and infection . Neonates may acquire S. aureus through the birth canal , while a concurrent maternal infection may be present in up to 20% of newborns with MRSA infection . The rate of vaginal MRSA colonization among pregnant women has been estimated to be 2.8% in previous reports , and vaginal delivery has been associated with increased risk of S. aureus neonatal transmission . In a previous systematic review, the pooled proportion of MRSA carriage among neonate mothers was 2.1% , with studies from Jordan reporting a prevalence of 9.72% MRSA colonization , Egypt of 1.69% , Brazil of 1.39% , and Germany of 0.51% . Vertical transmission of MRSA has also been indicated by the association of maternal MRSA chorioamnionitis with neonatal MRSA sepsis . Female sex and multiple gestation are additional risk factors for MRSA colonization and infection , whereas antibiotic administration in the week before delivery has been associated with a lower risk of MRSA transmission . After birth, neonates are exposed to S. aureus following contact with adult skin . The carriage of S. aureus among adults ranges from 30% to 70%. MRSA has been shown to spread horizontally through contact with healthcare providers or the hospital setting . Other factors, including NICU overcrowding and understaffing, have been associated with a higher risk of colonization and transmission and could result in MRSA outbreaks . It has also been demonstrated that mothers can vertically transmit MRSA to their infants through breast milk , while fathers can through direct contact with their infants . Controlling transmission is more challenging in intensive care units because S. aureus can persist on ambient surfaces for extended periods . Among neonatal factors, prematurity and low birth weight are the main risk factors for MRSA colonization and infection . Numerous studies have shown that low birth weight was associated with an increased risk of MRSA colonization and/or infection . The prevalence of MRSA infection in extremely low-birth-weight neonates was estimated at 53.4 per 10,000 infants, which was much higher compared to 23.2, 7.9, and 5.0 per 10,000 infants in very-low-birth-weight, low-birth-weight, and appropriate-birth-weight neonates, respectively . Neonates of lower gestational age were also more susceptible to being positive in nasal or both nasal and groin MRSA colonization, compared to neonates of higher gestational age who were positive on groin swabs only . Long-term ventilation support, intravascular catheters, antibiotic administration, total parenteral nutrition, and surgical interventions are additional risk factors for MRSA infections . In the NICU, neonates frequently need several procedures during their hospital stay, including endotracheal tube insertion, mechanical ventilation, central vascular catheterization, and surgery . A previous study demonstrated that compared with non-MRSA-colonized infants, MRSA-colonized infants who experienced a greater incidence of late-onset sepsis were more likely to be intubated or mechanically ventilated . A recent review of S. aureus outbreaks in neonatal intensive care units conducted in Leeds, U.K., identified that MRSA bacteremia was more likely to occur in infants with discharge skin lesions, prior abdominal surgery, current MRSA colonization, and Broviac or peripherally implanted central catheter lines . An increased risk of MRSA infection has also been associated with feeding practices such as parenteral nutrition and gavage feeding . Moreover, longer hospital stays and kangaroo care were other independent risk factors for MRSA infection. Among all risk factors, MRSA colonization is the most significant risk factor for developing MRSA infection in neonates. According to Huang et al., MRSA-colonized neonates had a considerably greater rate of MRSA infection (26%) than non-colonized neonates (2%) , whereas, as suggested by the metanalysis by Zervou et al., colonized neonates have a 24.2-fold higher chance of contracting an MRSA infection while in the NICU compared with non-colonized newborns . Neonates may be colonized with MRSA within a median of 9 days from admission, with a range of 1–91 days . In addition, the median interval between MRSA colonization and infection is 4 to 9 days . In a previous report from a New York NICU, MRSA colonization was detected at a median of 17 days, with a range of 4 to 159, whereas nearly two-thirds of the neonates developed colonization during the first 3 weeks of life . MRSA infection has a significant clinical impact. The most common manifestations of MRSA infections in neonates are SSTIs; however, invasive diseases have also been reported . Most cases of invasive MRSA disease (75%) are associated with bacteremia . Late-onset sepsis is a common clinical manifestation of MRSA infection and can range from a moderate focal infection to severe invasive disease . According to reports, late-onset newborn sepsis can raise mortality from 7% to 18% and lengthen the inpatient stay by three weeks . Among neonates with S. aureus bacteremia in a ten-year retrospective research in the U.K. between 1993–2003, MRSA was detected in nearly one-third of neonates . Similarly, among neonates with S. aureus bacteremia in the U.S., MRSA was found in 47% of the cases, indicating that MRSA has emerged as a major cause of neonatal sepsis . Finally, Dolapo et al. reported that the prevalence of MRSA bloodstream infections in neonates increased from 24% to 55% between 2000 and 2009 . Infectious endocarditis, abscesses in the myocardium, liver, spleen, or kidneys, necrotizing pneumonia, osteomyelitis, myositis, meningitis, toxic shock syndrome, septic thrombophlebitis, venous thrombosis, sustained bacteremia, ocular infections, and Waterhouse-Friderichsen syndrome are just a few of the numerous MRSA manifestations that have been reported . CA-MRSA infections typically manifest as SSTIs, in contrast to HA-MRSA infections, although more severe invasive manifestations can also occur . In comparison with HA-MRSA, CA-MRSA contains the virulence genes lukS-PV/lukf-PV that generate PVL, and produce a pore-forming cytotoxin that causes leukocyte death and tissue necrosis . In the U.S., CA-MRSA was the most common cause of SSTIs . The clinical manifestations of SSTIs can vary from cellulitis or a simple abscess to more serious soft-tissue infections such as necrotizing fasciitis, pyomyositis, and mediastinitis as a consequence of retropharyngeal abscess . When term neonates have localized only pustulosis with no signs or symptoms of sepsis, lumbar puncture is not required . Careful patient monitoring and prompt access to microbiological and laboratory tests are crucial because the clinical symptoms and indicators at the beginning of MRSA infections can be non-specific . In addition, MRSA-infected newborns may have a higher readmission rate and a longer course of infection than methicillin-susceptible S. aureus cases ; however, there appears to be no difference between MRSA and methicillin-susceptible S. aureus in terms of clinical presentation and mortality . In very immature preterm neonates, MRSA infections increase the risk of unfavorable short- and long-term outcomes, as well as mortality . The mortality rate of MRSA infections ranges from 2.9% to 28%, with significant variation across institutions . According to earlier research, the case fatality risk of neonatal MRSA sepsis ranged between 9.5% and 55% . A previous study also reported that among MRSA infections, sepsis had a mortality rate of 16%, pneumonia of 32.1%, and necrotizing enterocolitis of 27.3% . 6.1. Precautions Against Colonization Neonates are colonized when passing through the maternal birth canal. Moreover, newborns who are placed on the mother’s breast as soon as possible after delivery are colonized with the maternal skin microbiome. Neisseria and Streptococcus species are two of the many bacteria that quickly colonize a newborn’s mouth. According to Fukuda et al., newborns who were breastfed exhibited a quick rise in common α or Á- Streptococcus in their mouths . Importantly, Uehara et al. showed that precolonization of neonatal mouth and nostrils with common α- and/or Á- Streptococcus prevented MRSA colonization . Additionally, distributing the mother’s breast milk over and into the mouths of extremely-low-birth-weight neonates as soon as they are admitted into the NICU can greatly reduce the colonization rate of MRSA in their mouths . The most crucial infection control measure is strict hand hygiene before and after handling neonates; however, this is one of the least followed. Hand hygiene using tap water alone can significantly reduce the risk of infection, even in the absence of a disinfectant. Nonetheless, the use of chlorhexidine gluconate and other similar disinfectants in soap is not an efficient preventive measure and is only as effective as using tap water because many strains of MRSA are resistant to these disinfectants. Research has demonstrated that the MRSA isolation rate decreases when gloves are used as an infection control method when handling neonates . An overall guidance for precautions against MRSA colonization is depicted in . 6.2. Decolonization Currently, prevention rather than treatment is the best approach to managing neonatal MRSA infections. Preventing MRSA transmission in the NICU is essential because MRSA colonization is the major risk factor for developing MRSA infection . Strict hand hygiene is crucial in preventing MRSA spread, in addition to surveillance and decolonization . Cohorting and isolating MRSA-positive patients, taking barrier precautions, educating healthcare professionals, avoiding crowded wards, and monitoring and decolonizing parents and healthcare providers are additional strategies that may prevent MRSA infections . Several NICUs have implemented detection and isolation programs to prevent the spread of MRSA and decrease infection rates. These programs employ surveillance to promptly identify affected patients, followed by cohorting and isolation using standard contact precautions. Decolonization is the key to preventing infection. NICUs have reported varying degrees of success following policies of active MRSA surveillance swabs and decolonization using nasal mupirocin with or without an antiseptic . Controlling MRSA transmission in NICUs is challenging because healthcare providers, parents, family members, and visitors are asymptomatically colonized and unintentionally act as reservoirs for transmission . Furthermore, S. aureus lives on environmental surfaces for extended periods . According to the CDC 2021 S. aureus NICU recommendations, NICU patients should at minimum have their anterior nares swabbed . The recommendation that the umbilicus and neck be expressly listed as preferred screening sites in neonates has been deleted from the recently updated National Institute of Clinical Excellence guidelines on the management of MRSA due to a lack of evidence . However, targeted MRSA decolonization techniques may have limitations. First, up to 42% of infected neonates have no previous positive MRSA screening swab, preventing any chance of decolonization, even with weekly monitoring cultures . Second, the median period between colonization and infection is only 5 days, which reduces the window of opportunity for decolonization for many neonates. Third, the effectiveness of decolonization to eliminate MRSA colonization and prevent MRSA infections may be restricted because, according to previous reports, 38% of neonates who had decolonization treatment became recolonized during their NICU stay, and 16% contracted an MRSA infection . To effectively reduce MRSA infections in neonates, some authors have suggested treating all newborns with mupirocin . Many NICUs have established protocols to identify and isolate colonized children since MRSA-colonized infants frequently act as a reservoir for transmission to other infants . Unknown is how treating all newborns, including those that are not colonized, may change the neonatal microbiome over time. It is noteworthy that, in some situations, a universal approach has led to the development of mupirocin resistance . It has been noted that controlling MRSA outbreaks in NICUs can be also challenging . Such outbreaks have only been successfully contained by the application of strict infection control measures, sometimes in conjunction with mupirocin treatment. Decolonization methods in addition to continuous reinforcement of hygienic measures should include (1) mupirocin twice a day for 5 to 10 days to decolonize the nasal cavity and (2) topical body decolonization regimens using a skin antiseptic solution, such as chlorhexidine, for 5–14 days. 6.3. Antimicrobial Therapy To obtain the best empirical antimicrobial treatments for neonates suspected of having an MRSA infection in NICUs, antibiotic susceptibility monitoring is essential. According to several studies, individuals with MRSA bacteremia may benefit from taking a beta-lactam in addition to vancomycin or daptomycin to reduce the duration of their illness and prevent recurrence . Although the majority of MRSA isolates were susceptible to trimethoprim-sulfamethoxazole, tetracycline, rifampin, linezolid, ceftaroline, chlorhexidine, and mupirocin, surveillance studies over the past decade have revealed high resistance rates to erythromycin, clindamycin, and ciprofloxacin . An overview of the treatment options for MRSA infection is depicted in . In term neonates, topical mupirocin therapy may be sufficient for minor cases of localized pustulosis . Until bacteremia is ruled out, vancomycin or clindamycin should be used in cases of localized diseases in premature or very-low-birth-weight neonates or more widespread diseases affecting many sites in term infants . The best treatment for severe MRSA infections in newborns is vancomycin, although numerous antibiotics have been explored with varying degrees of efficacy . The use of combination therapy with rifampin, gentamicin, or daptomycin in neonatal sepsis should be decided on an individual basis because there is little evidence of its possible benefits . There have been reports of VISA, and even VRSA, strains, that have acquired vanA resistance from strains of vancomycin-resistant enterococci . Since vancomycin is the empirical antibiotic of choice for neonates with sepsis and extensive skin infections, particularly in areas with high MRSA prevalence, its decreased susceptibility to MRSA presents a significant challenge . Strategies that target the virulent determinants of MRSA may show promise, although their effectiveness and safety in neonatal populations have not yet been confirmed . There is limited experience using clindamycin and linezolid for severe MRSA infections in neonates; however, these medications may be used to treat neonates with susceptible isolates who have non-endovascular infections . The U.S. Food and Drug Administration (FDA) has approved clindamycin for the treatment of severe S. aureus infections. It has gained widespread use for treating SSTIs and has been effectively used to treat invasive susceptible CA-MRSA infections in children, including osteomyelitis, septic arthritis, pneumonia, and lymphadenitis, despite not being specifically approved for the treatment of MRSA infections . It is not recommended for endovascular infections such as septic thrombophlebitis or infective endocarditis because of its bacteriostatic properties. Although its entry into the cerebrospinal fluid is restricted, clindamycin has exceptional tissue penetration, especially in bone and abscesses . Linezolid is a synthetic oxazolidinone that prevents the 50S ribosome from initiating protein synthesis. The FDA has approved this treatment for nosocomial MRSA pneumonia and SSTIs in adults and children. It is also in vitro active against VRSA and VISA . Although an outbreak of MRSA infection resistant to linezolid has been reported, linezolid resistance is uncommon . Long-term use usually results in resistance through a mutation in the 23S ribosomal RNA-binding site for linezolid or methylation of adenosine at position 2503 in 23S ribosomal RNA caused by the cfr gene . Daptomycin is an antibiotic of the lipopeptide class that causes bactericidal action in a concentration-dependent manner by interfering with the function of cell membranes through calcium-dependent binding. Research is ongoing to establish the pharmacokinetics, safety, and efficacy of daptomycin in children . Due to a lack of research on daptomycin’s effectiveness and safety, it is not frequently used in neonates, with numerous examples, despite having demonstrated the advantages and relative safety of daptomycin use in newborns . When vancomycin fails clinically, daptomycin may be considered. Due to their synergistic impact, daptomycin and beta-lactams are more successful when used in combination to treat invasive MRSA infections, including bacteremia and endocarditis . However, according to a randomized clinical trial conducted between 2015 and 2018, there was no correlation between beta-lactam use and lower treatment failure and death when used in combination with regular vancomycin or daptomycin therapy , while a meta-analysis suggested that combined treatment might enhance certain microbiological outcomes but not mortality . Rifampicin exhibits bactericidal action against S. aureus , reaches high intracellular levels, and penetrates biofilms . It should not be used as monotherapy due to the rapid development of resistance; however, in some situations, it may be used in combination with another active antibiotic. Telavancin, a parenteral lipoglycopeptide, prevents cell wall formation by attaching to peptidoglycan chain precursors and depolarizing cell membranes . MRSA, VISA, and VRSA are all susceptible to its bactericidal effects. The FDA has not approved trimethoprim-sulfamethoxazole for the treatment of staphylococcal infections. However, trimethoprim-sulfamethoxazole has emerged as a significant option for the outpatient treatment of SSTIs as 95–100% of CA-MRSA strains are sensitive in vitro . Because trimethoprim-sulfamethoxazole increases the risk of kernicterus, it is not recommended during the first few months of life. Neonates are colonized when passing through the maternal birth canal. Moreover, newborns who are placed on the mother’s breast as soon as possible after delivery are colonized with the maternal skin microbiome. Neisseria and Streptococcus species are two of the many bacteria that quickly colonize a newborn’s mouth. According to Fukuda et al., newborns who were breastfed exhibited a quick rise in common α or Á- Streptococcus in their mouths . Importantly, Uehara et al. showed that precolonization of neonatal mouth and nostrils with common α- and/or Á- Streptococcus prevented MRSA colonization . Additionally, distributing the mother’s breast milk over and into the mouths of extremely-low-birth-weight neonates as soon as they are admitted into the NICU can greatly reduce the colonization rate of MRSA in their mouths . The most crucial infection control measure is strict hand hygiene before and after handling neonates; however, this is one of the least followed. Hand hygiene using tap water alone can significantly reduce the risk of infection, even in the absence of a disinfectant. Nonetheless, the use of chlorhexidine gluconate and other similar disinfectants in soap is not an efficient preventive measure and is only as effective as using tap water because many strains of MRSA are resistant to these disinfectants. Research has demonstrated that the MRSA isolation rate decreases when gloves are used as an infection control method when handling neonates . An overall guidance for precautions against MRSA colonization is depicted in . Currently, prevention rather than treatment is the best approach to managing neonatal MRSA infections. Preventing MRSA transmission in the NICU is essential because MRSA colonization is the major risk factor for developing MRSA infection . Strict hand hygiene is crucial in preventing MRSA spread, in addition to surveillance and decolonization . Cohorting and isolating MRSA-positive patients, taking barrier precautions, educating healthcare professionals, avoiding crowded wards, and monitoring and decolonizing parents and healthcare providers are additional strategies that may prevent MRSA infections . Several NICUs have implemented detection and isolation programs to prevent the spread of MRSA and decrease infection rates. These programs employ surveillance to promptly identify affected patients, followed by cohorting and isolation using standard contact precautions. Decolonization is the key to preventing infection. NICUs have reported varying degrees of success following policies of active MRSA surveillance swabs and decolonization using nasal mupirocin with or without an antiseptic . Controlling MRSA transmission in NICUs is challenging because healthcare providers, parents, family members, and visitors are asymptomatically colonized and unintentionally act as reservoirs for transmission . Furthermore, S. aureus lives on environmental surfaces for extended periods . According to the CDC 2021 S. aureus NICU recommendations, NICU patients should at minimum have their anterior nares swabbed . The recommendation that the umbilicus and neck be expressly listed as preferred screening sites in neonates has been deleted from the recently updated National Institute of Clinical Excellence guidelines on the management of MRSA due to a lack of evidence . However, targeted MRSA decolonization techniques may have limitations. First, up to 42% of infected neonates have no previous positive MRSA screening swab, preventing any chance of decolonization, even with weekly monitoring cultures . Second, the median period between colonization and infection is only 5 days, which reduces the window of opportunity for decolonization for many neonates. Third, the effectiveness of decolonization to eliminate MRSA colonization and prevent MRSA infections may be restricted because, according to previous reports, 38% of neonates who had decolonization treatment became recolonized during their NICU stay, and 16% contracted an MRSA infection . To effectively reduce MRSA infections in neonates, some authors have suggested treating all newborns with mupirocin . Many NICUs have established protocols to identify and isolate colonized children since MRSA-colonized infants frequently act as a reservoir for transmission to other infants . Unknown is how treating all newborns, including those that are not colonized, may change the neonatal microbiome over time. It is noteworthy that, in some situations, a universal approach has led to the development of mupirocin resistance . It has been noted that controlling MRSA outbreaks in NICUs can be also challenging . Such outbreaks have only been successfully contained by the application of strict infection control measures, sometimes in conjunction with mupirocin treatment. Decolonization methods in addition to continuous reinforcement of hygienic measures should include (1) mupirocin twice a day for 5 to 10 days to decolonize the nasal cavity and (2) topical body decolonization regimens using a skin antiseptic solution, such as chlorhexidine, for 5–14 days. To obtain the best empirical antimicrobial treatments for neonates suspected of having an MRSA infection in NICUs, antibiotic susceptibility monitoring is essential. According to several studies, individuals with MRSA bacteremia may benefit from taking a beta-lactam in addition to vancomycin or daptomycin to reduce the duration of their illness and prevent recurrence . Although the majority of MRSA isolates were susceptible to trimethoprim-sulfamethoxazole, tetracycline, rifampin, linezolid, ceftaroline, chlorhexidine, and mupirocin, surveillance studies over the past decade have revealed high resistance rates to erythromycin, clindamycin, and ciprofloxacin . An overview of the treatment options for MRSA infection is depicted in . In term neonates, topical mupirocin therapy may be sufficient for minor cases of localized pustulosis . Until bacteremia is ruled out, vancomycin or clindamycin should be used in cases of localized diseases in premature or very-low-birth-weight neonates or more widespread diseases affecting many sites in term infants . The best treatment for severe MRSA infections in newborns is vancomycin, although numerous antibiotics have been explored with varying degrees of efficacy . The use of combination therapy with rifampin, gentamicin, or daptomycin in neonatal sepsis should be decided on an individual basis because there is little evidence of its possible benefits . There have been reports of VISA, and even VRSA, strains, that have acquired vanA resistance from strains of vancomycin-resistant enterococci . Since vancomycin is the empirical antibiotic of choice for neonates with sepsis and extensive skin infections, particularly in areas with high MRSA prevalence, its decreased susceptibility to MRSA presents a significant challenge . Strategies that target the virulent determinants of MRSA may show promise, although their effectiveness and safety in neonatal populations have not yet been confirmed . There is limited experience using clindamycin and linezolid for severe MRSA infections in neonates; however, these medications may be used to treat neonates with susceptible isolates who have non-endovascular infections . The U.S. Food and Drug Administration (FDA) has approved clindamycin for the treatment of severe S. aureus infections. It has gained widespread use for treating SSTIs and has been effectively used to treat invasive susceptible CA-MRSA infections in children, including osteomyelitis, septic arthritis, pneumonia, and lymphadenitis, despite not being specifically approved for the treatment of MRSA infections . It is not recommended for endovascular infections such as septic thrombophlebitis or infective endocarditis because of its bacteriostatic properties. Although its entry into the cerebrospinal fluid is restricted, clindamycin has exceptional tissue penetration, especially in bone and abscesses . Linezolid is a synthetic oxazolidinone that prevents the 50S ribosome from initiating protein synthesis. The FDA has approved this treatment for nosocomial MRSA pneumonia and SSTIs in adults and children. It is also in vitro active against VRSA and VISA . Although an outbreak of MRSA infection resistant to linezolid has been reported, linezolid resistance is uncommon . Long-term use usually results in resistance through a mutation in the 23S ribosomal RNA-binding site for linezolid or methylation of adenosine at position 2503 in 23S ribosomal RNA caused by the cfr gene . Daptomycin is an antibiotic of the lipopeptide class that causes bactericidal action in a concentration-dependent manner by interfering with the function of cell membranes through calcium-dependent binding. Research is ongoing to establish the pharmacokinetics, safety, and efficacy of daptomycin in children . Due to a lack of research on daptomycin’s effectiveness and safety, it is not frequently used in neonates, with numerous examples, despite having demonstrated the advantages and relative safety of daptomycin use in newborns . When vancomycin fails clinically, daptomycin may be considered. Due to their synergistic impact, daptomycin and beta-lactams are more successful when used in combination to treat invasive MRSA infections, including bacteremia and endocarditis . However, according to a randomized clinical trial conducted between 2015 and 2018, there was no correlation between beta-lactam use and lower treatment failure and death when used in combination with regular vancomycin or daptomycin therapy , while a meta-analysis suggested that combined treatment might enhance certain microbiological outcomes but not mortality . Rifampicin exhibits bactericidal action against S. aureus , reaches high intracellular levels, and penetrates biofilms . It should not be used as monotherapy due to the rapid development of resistance; however, in some situations, it may be used in combination with another active antibiotic. Telavancin, a parenteral lipoglycopeptide, prevents cell wall formation by attaching to peptidoglycan chain precursors and depolarizing cell membranes . MRSA, VISA, and VRSA are all susceptible to its bactericidal effects. The FDA has not approved trimethoprim-sulfamethoxazole for the treatment of staphylococcal infections. However, trimethoprim-sulfamethoxazole has emerged as a significant option for the outpatient treatment of SSTIs as 95–100% of CA-MRSA strains are sensitive in vitro . Because trimethoprim-sulfamethoxazole increases the risk of kernicterus, it is not recommended during the first few months of life. Over the past 40 years, MRSA has become a significant pathogen that has spread to hospitals and the community. It is the primary cause of HA infections, including bacteremia, endocarditis, SSTIs, and infections of the bones and joints . Although the prevalence of MRSA has decreased, it still poses a serious clinical risk; hence, special attention is required. Routine surveillance and accurate detection of MRSA strains are crucial for providing the best antibiotic therapy, comprehending the evolution of nosocomial transmission control, and implementing preventative measures. Furthermore, public health in Europe continues to prioritize S. aureus or MRSA, as evidenced by the fact that 8 out of 30 countries, including Greece, report prevalence rates of MRSA > 25% . The significant increase in MRSA colonization upon admission may support some centers’ practice of isolating their outborn population until their MRSA status is determined, even though the CDC does not list interhospital transfer of neonates as one of the clinical conditions for transmission-based precautions . Significantly, compared with non-colonized neonates, those who are MRSA carriers upon admission to the NICU have a significantly higher risk of contracting an MRSA-associated infection while in the hospital. To reduce MRSA rates and reduce disease transmission, numerous NICUs have implemented active detection and isolation programs . Attempts have been made in various healthcare settings to implement universal MRSA-targeted decolonization. Because strains that colonize neonates and cause subsequent infections are strongly associated, many NICUs have attempted either targeted or universal decolonization as a method of preventing MRSA infections . However, the results of these policies have been inconsistent, with the development of resistance being among the possible drawbacks. Moreover, previous studies have demonstrated that several MRSA strains can be detected in NICUs , with Carey et al. reporting that several strain types were detected in colonized/infected neonates over eight years, even though data from routine weekly surveillance cultures were lacking . Larger studies are required to ascertain the cytotoxicity status of S. aureus to better understand whether these are potentially useful markers to take into consideration in future decolonization programs, given recent evidence regarding the potential role of virulence ascertained using comparable in vitro assays . A shift in epidemiology, with CA SCCmec genotypes becoming more and more linked to hospital infections, was indicated by the SCCmec typing results, which showed a mix of CA-MRSA and HA-MRSA genotypes in the hospital . After Healy et al. published the first report of CA-MRSA infections in NICU patients in 2004, similar changes from HA-MRSA to CA-MRSA strains were observed in additional NICUs . Mupirocin decolonization works effectively and has little adverse effects in MRSA-colonized neonates . Parental decolonization is another tactic that has recently been assessed to reduce neonatal MRSA colonization and subsequent infection. Decolonization of S. aureus -colonized parents reduced the incidence of infants acquiring an S. aureus strain concordant with a parental strain by 57%, according to a previous randomized controlled trial conducted in the U.S. . Strict commitment to neonatal decolonization methods combined with parent decolonization may be required to decrease infant colonization and infection . To identify the best empirical antimicrobial treatments for patients with suspected infections, antibiotic susceptibility monitoring in individual NICUs is essential. Vancomycin is still the best option for treating MRSA infections, although VISA and VRSA have emerged as examples of MRSA strains that are vancomycin-resistant. The co-occurrence of MRSA and VRSA phenotype has been reported in previous studies, in Asia, Europe, and North America, raising significant concerns. Human-origin isolates showed a susceptibility trend, indicating that linezolid should be the final medication of choice for multidrug-resistant MRSA. Additionally, human infections are increasingly related to oxacillin-susceptible MRSA . Traditional susceptibility testing may mistakenly identify oxacillin-susceptible MRSA strains as methicillin-sensitive S. aureus , making it more difficult to diagnose and treat S. aureus infections, underscoring that public health should prioritize surveillance of such new pathogens. The higher survival rate of very immature preterm neonates has led to an increase in the number of newborns at risk for MRSA colonization and infection. Despite the abundance of data reported from NICUs worldwide, multicenter or population-based studies to elucidate the epidemiology and clinical features of neonatal MRSA colonization and infections are lacking. Such information is essential for precisely estimating the MRSA disease burden and supporting surveillance and preventative decision-making. MRSA transmission, colonization, and infection in the NICU are complicated issues. The significance of reducing the colonization rate in the NICU is highlighted by the 24.2 relative risk of recurrent infection among MRSA carriers compared with non-carriers. To reduce MRSA colonization, infection, and transmission in hospitalized neonates, customized approaches are required. Data from prospective randomized multicenter trials and continuous local surveillance of clinical and molecular epidemiology of MRSA must be combined to effectively control MRSA in the NICU. It is crucial to address the rapid changes in MRSA population structure and pathogenic factors; therefore, new techniques for detecting MRSA resistance are required. Because of the increasing antibiotic resistance of MRSA and the uncertainty surrounding the safety and efficacy of decolonization procedures in newborns, basic preventative measures continue to be the key to reducing neonatal MRSA infections. New strategies to limit MRSA from endangering NICU patients should be developed, including molecular analysis of the strains, shifting patterns of antibiotic susceptibility, and the existence of possible virulence factors. Further extensive research and surveillance are warranted to explore the genetic diversity and prevalence of MRSA. |
A Novel Technique of Amniotic Membrane Preparation Mimicking Limbal Epithelial Crypts Enhances the Number of Progenitor Cells upon Expansion | c0b49748-feb6-4135-9fa1-15fdb55b159c | 10001367 | Anatomy[mh] | The homeostasis of the dynamic cellular organization in the cornea mainly depends on the regenerative efficiency of the stem cells in the surrounding limbus . Tissue-specific human limbal epithelial stem cells (hLESCs) residing in the limbal epithelial crypts of the palisades of Vogt continuously compensate for the loss of superficial human corneal epithelial cells (hCECs) . The insufficient compensation of diminished hCECs in the corneal epithelium due to the lack or malfunction of hLESCs leads to severe ocular surface disease or so-called limbal epithelial stem cell deficiency (LSCD) . The hLESCs play an essential role in epithelial differentiation, angiogenesis, and extracellular matrix (ECM) organization . Diverse therapeutic approaches have been used to treat both monocular and binocular LSCD. However, cultivated limbal epithelial stem cell transplantation (CLET) of expanded autologous limbal tissue seems to be the most common method for monocular LSCD . The CLET procedure is based on isolating the limbal biopsy from the contralateral eye and treatment with a proteolytic enzyme to digest the surrounding ECM, which helps the hLESCs get released and migrate from the niche. Furthermore, the digested limbal tissue or single isolated hLESCs are harvested ex vivo in a medium containing stem-cell-supporting growth factors and supplements, achieving cell expansion and graft tissue synthesis . Upon transplantation, hLESCs reside on the damaged corneal-limbal tissue, re-creating the limbal stem cell niche that allows epithelial regeneration. Following the existing standard protocols, the reported success rate of CLET varies. A favorable morphological outcome implying stable, intact, completely epithelized and avascular corneal surface is reported as 46.7% to 80.9%, whereas success as a functional outcome such as visual acuity varies from 60.5% to 78.7% . Successful transplantation is directly dependent on the graft tissue quality and the percentage of hLESCs/early progenitor cells in the graft. For successful transplantation, at least 3% of the cells in the expanded cell culture must express the p63 marker . Therefore, establishing a protocol that provides a high percentage of the hLESCs/early progenitor cells in the transplantation graft is of high importance. Human amniotic membrane (HAM) has proven to be a very efficient therapeutic tool in many ocular surface diseases, supporting wound healing and regeneration while suppressing inflammation , angiogenesis , and fibrosis , and it possesses anti-microbial features . It is used for corneal epithelial regeneration, conjunctival reconstruction, glaucoma interventions, and the treatment of corneal melting and perforations. Importantly, it is one of the most used carriers for the ex vivo expansion of hLESCs . HAM contains stem cell niche factors that support maintenance . Generally, such maintenance depends on the inhabitance of the stem cells in a specific niche that allows their anchoring and communication with supporting cells, the release of specific growth factors and cell cycle molecules, and the involvement of evolutionary conserved molecular pathways. Within the niche, the stem cells undergo symmetric or asymmetric division to transient amplifying cells (TACs) that leave the environment and become functionally mature corneal cells . It is known that the physical cues of the cellular environment guide stem cell fate . It is also suggested that biomechanical changes in the limbal stromal niche affect hLESCs fate . No less importantly, mechanical and environmental changes in the corneal tissue have implications for some corneal diseases and pathologies . We hereby present a novel suturing preparation technique that causes the three-dimensional (3D) radial folding of the HAM, mimicking crypt-like formations. The novel approach may allow the hLESCs, upon limbal biopsy expansion and cultivation ex vivo, to reside in the undulated crypts of HAM. This may potentially maintain the putative characteristics of the expanded hLESCs and thus ensure a higher quality of the expanded graft tissue compared to the conventional state-of-the-art method. Therefore, we aimed to compare the progenitor/differentiation state of the cells cultivated in the crypt-like HAMs vs. the cells cultivated on the flat-like HAMs.
The Regional Committee for Medical and Health Research Ethics in South-Eastern Norway (No 2017/418) approved tissue harvesting and laboratory procedures, and all tissue collections complied with the Guidelines of the Helsinki Declaration. Unless stated otherwise, all reagents were purchased from Merck (Darmstadt, Germany). 2.1. Human Amniotic Membrane (HAM) A Placenta was collected after a scheduled cesarian section from a full-term pregnancy. Informed consent and institutional board review approval had previously been obtained from the patient. According to the standard protocol, the placenta was immediately transported in a sterile container and further processed under sterile conditions . Proper washing with 0.9% NaCl (Fresenius Kabi AB, Uppsala, Sweden) or 0.9% NaCl containing 100 U/mL Penicillin, 100 μg/mL Streptomycin (P4333), and 2.5 μg/mL Amphotericin B (A2942) was repeatedly performed. The HAM was then separated from the chorion by blunt dissection, washed out from residual blood, and transferred onto a nitrocellulose filter carrier, pore size 0.45 μm (111306-47-CAN, Sartorius, Gottingen, Germany) with the epithelial side up, then divided into 3 × 3 cm and 5 × 5 cm pieces. HAM pieces were cryopreserved in 50% glycerol, 48.5% DMEM/F12 (31331028, Invitrogen, Carlsbad, CA, USA), 100 U/mL Penicillin, 100 μg/mL Streptomycin and 2.5 μg/mL Amphotericin B and stored at −80 °C. 2.2. HAM Preparation for hLESC Expansion and Cultivation Before use, HAMs were thawed, warmed to room temperature, and washed three times with a medium containing DMEM/F12, 100 U/mL Penicillin, and 100 μg/mL Streptomycin. Thereafter, the HAMs were placed on polyester membrane Netwell TM inserts (3479, Corning Inc., New York, NY, USA) 24 mm in diameter, with the epithelial side up by two different techniques : (1) HAMs 3 × 3 cm ( .1A) in size were peeled from the nitrocellulose filter paper. Further, HAMs were placed and stretched on the top of the polyester membrane and sutured by the eight individual sutures ( .1B) near the edge of the polyester membrane. The HAMs were then tightly stretched on the top of the membrane, making a flat surface ( .1C). Excess HAM tissue that remained at the edge of the polyester membrane was carefully removed with a disposable sterile scalpel. (2) The other approach was to use HAMs 5 × 5 cm in size ( .2A) placed on top of the membrane and sutured so that the HAMs were loosely attached. HAMs were sutured by individual sutures ( .2B). In addition, an individual suture was placed in the center of the HAM/polyester membrane to obtain the folding of the HAM and to keep it in close contact with the membrane ( .2C). The excessive HAM tissue at the edges was again removed accordingly. The sutured HAMs on the Netwell TM inserts were immersed in DMEM/F12 medium containing 100 U/mL Penicillin and 100 μg/mL Streptomycin and kept at 37 °C, 5% CO 2 , and 95% air overnight to obtain HAM free of any glycerol remains. 2.3. Limbal Biopsies and Human LESC Harvesting Following corneal transplantation, the remaining human corneal-scleral rings from three donors (n = 3) were divided into twelve limbal biopsies of equal size and thoroughly washed with DMEM/F12 medium containing 100 U/mL Penicillin and 100 μg/mL Streptomycin. The biopsies were treated with neutral protease and Dispase II (2.4 U/mL 4942078001, Roche Diagnostics, Mannheim, Germany) for 10 min at 37 °C. The dissociation process was blocked using Fetal Bovine Serum (FBS, F2442). The limbal biopsies were then placed centrally on the top of the HAMs, with the epithelial side down and submerged in a standardly used complex medium (COM). COM consisted of DMEM/F12, Penicillin (100 U/mL), Streptomycin (100 μg/mL), Amphotericin B (2.5 μg/mL), human epidermal growth factor (2 ng/mL, E9644), insulin (5 μg/mL), sodium selenite (5 ng/mL) and transferrin (5 μg/mL, l1884), cholera toxin A subunit from Vibrio cholerae (30 ng/mL, C8180), hydrocortisone (0.03 μg/mL, H0888), 5% FBS, and 0.5% dimethyl sulfoxide (DMSO, D2650). After 2 h of incubation, the attached limbal biopsies were completely covered by COM. Further, the limbal biopsies and outgrowing LESCs were harvested and incubated at 37 °C with 5% CO 2 , and 95% air for the following three weeks. The culture medium was changed every three days. 2.4. Immunohistochemistry (IHC) and Immunofluorescence Microscopy Limbal biopsies with hLESCs cultured on HAMs were cut from the Netwell TM inserts. Samples were fixed in 4% formalin overnight at 4 °C and then processed in dehydrated graded alcohol series of 70% (10–15 min), 80% (10–15 min), 96% (2 × 10 min), and 100% ethanol (2 × 10 min) before having xylene added (3 × 10 min) and being washed with melted paraffin (3 × 10 min) and then embedded in paraffin for immunohistochemistry (IHC). Paraffinized tissue was cut into 3–4 μm thick sections using an automated microtome (HM 355S, Thermo Fisher Scientific, Waltham, MA, USA) and attached to histological slides. Deparaffinization was performed in xylene (2 × 10 min), then rehydration was performed by sinking in 100%, 96%, and 70% ethanol, and then distilled water. Hematoxylin & Eosin (H&E) staining was primarily performed. Slides were immersed in Mayers hematoxylin plus solution (01825, Histolab, Askim, Sweden) for 10 min and then rinsed with distilled water (10 min) followed by eosin staining (10 min), and then they were rehydrated in upgraded alcohol series of 70%, 96%, 100% ethanol, and xylene. Slides were further mounted using Pertex (00840, Histolab) mounting medium. For IHC, heat-induced antigen retrieval was performed in a microwave for 5 min at 900W and 15 min at 600W in a citrate buffer (pH 6, C9999) or by PT module (LabVision, Fremont, CA, USA). Blocking of non-specific binding sites with 5% Bovine Serum Albumin (BSA, A9418) dissolved in Dulbecco’s Phosphate Buffered Saline (DPBS, 14190-144, Thermo Fisher Scientific) was conducted for 20 min. Further, slides were stained with primary antibodies diluted in 1% BSA for 1 h. Slides were stained using antibodies for the following progenitor markers: tumor protein p63 alpha (p63α, rabbit polyclonal, 1:200 dilution, 4892S, Cell Signaling, Beverly, MA, USA), SRY-Box Transcription Factor 9 (SOX9, 82630, Cell Signaling, rabbit monoclonal, 1:200), quiescence marker: CCAAT/enhancer-binding protein delta (CEBPD, rabbit polyclonal, 1:200 dilution, ab198320, Abcam, Cambridge, UK), and proliferation marker Ki-67 (rabbit monoclonal, 1:200, RM-9106-S, Thermo Scientific) and the following differentiation markers: cytokeratin 3/12 (KRT3/12, mouse monoclonal, 1:100 dilution, 08691431, MP biomedicals, Santa Ana, CA, USA) and connexin-43 (CX43, rabbit polyclonal, 1:300, C6219). Then, the slides were thoroughly washed three times for 5 min with PBS-tween buffer (28352, Thermo Fisher Scientific). Incubation was continued with the appropriate animal type of secondary antibody: Cy3 ® goat anti-rabbit IgG (rabbit monoclonal, 1:500 dilution, A10520, Abcam) for samples stained with p63α, CEBPD, SOX9, Ki67, and Alexa Fluor ® 488 donkey anti-mouse IgG (1:500 dilution, mouse monoclonal, 21202, Abcam) for samples stained with KRT3/12, and Alexa Fluor ® 488 donkey anti-rabbit IgG (1:500 dilution, rabbit monoclonal, A21206, Abcam) for antibody staining CX43. The secondary antibody was incubated for 45 min and washed three times for 5 min. Nuclear staining was performed using a 4′,6-daminidino-2-phenylindole (DAPI) mounting solution (P36931, Life technologies corporation, Carlsbad, CA, USA). Further, LabVision Autostainer 360 (Lab Vision Corporation, VT) was used for staining with antibodies against adherent junction molecules such as E-cadherin (CDH1, mouse monoclonal, 1:50 dilution, n1620, DakoCytomation, Santa Clara, CA, USA) and N-cadherin (CDH2, mouse monoclonal, 1:100 dilution, m3613, DakoCytomation). Visualization was done using the standard peroxidase technique (UltravisionOne HRP system, Thermo Fisher Scientific). Primary antibody binding to an expressed antigen was recognized by a secondary antibody conjugated with peroxidase-labeled polymer with diaminobenzidine (DAB). Each staining was performed at least three times, and each sample was tested in triplicate. Negative and positive controls were performed simultaneously for all antibodies. All antibodies used for IHC in this study are summarized in . Bright-field images of H&E and DAB-stained samples were taken by a ZEISS Axio Observer Z1 microscope (ZEISS, Oberkochen, Germany). Fluorescence was recorded by a ZEISS Axio Imager M1 fluorescence microscope (ZEISS). Three independent individuals used Image J software and counted nuclear antibody positivity (p63α, CEBPD, SOX9, and Ki67). 2.5. Statistical Analysis The technical replicates from the same donor and group of three donors of the hLESC harvested in two different conditions were averaged as a percentage mean ± standard error of the mean (SEM). Prism 8.3.0 (GraphPad, San Diego, CA, USA) was used for statistical analysis. The data were counted and analyzed by two different methods: in percentages representing the ratio of the number of cells positive for a specific marker and the total number of cells (DAPI positivity), or as the number of cells positive for a specific marker per mm 2 . Further, data were tested for normal distribution (Shapiro–Wilk test), and the difference was tested using an unpaired two-sample t-test. The significance level p ≤ 0.05 was counted as significant.
A Placenta was collected after a scheduled cesarian section from a full-term pregnancy. Informed consent and institutional board review approval had previously been obtained from the patient. According to the standard protocol, the placenta was immediately transported in a sterile container and further processed under sterile conditions . Proper washing with 0.9% NaCl (Fresenius Kabi AB, Uppsala, Sweden) or 0.9% NaCl containing 100 U/mL Penicillin, 100 μg/mL Streptomycin (P4333), and 2.5 μg/mL Amphotericin B (A2942) was repeatedly performed. The HAM was then separated from the chorion by blunt dissection, washed out from residual blood, and transferred onto a nitrocellulose filter carrier, pore size 0.45 μm (111306-47-CAN, Sartorius, Gottingen, Germany) with the epithelial side up, then divided into 3 × 3 cm and 5 × 5 cm pieces. HAM pieces were cryopreserved in 50% glycerol, 48.5% DMEM/F12 (31331028, Invitrogen, Carlsbad, CA, USA), 100 U/mL Penicillin, 100 μg/mL Streptomycin and 2.5 μg/mL Amphotericin B and stored at −80 °C.
Before use, HAMs were thawed, warmed to room temperature, and washed three times with a medium containing DMEM/F12, 100 U/mL Penicillin, and 100 μg/mL Streptomycin. Thereafter, the HAMs were placed on polyester membrane Netwell TM inserts (3479, Corning Inc., New York, NY, USA) 24 mm in diameter, with the epithelial side up by two different techniques : (1) HAMs 3 × 3 cm ( .1A) in size were peeled from the nitrocellulose filter paper. Further, HAMs were placed and stretched on the top of the polyester membrane and sutured by the eight individual sutures ( .1B) near the edge of the polyester membrane. The HAMs were then tightly stretched on the top of the membrane, making a flat surface ( .1C). Excess HAM tissue that remained at the edge of the polyester membrane was carefully removed with a disposable sterile scalpel. (2) The other approach was to use HAMs 5 × 5 cm in size ( .2A) placed on top of the membrane and sutured so that the HAMs were loosely attached. HAMs were sutured by individual sutures ( .2B). In addition, an individual suture was placed in the center of the HAM/polyester membrane to obtain the folding of the HAM and to keep it in close contact with the membrane ( .2C). The excessive HAM tissue at the edges was again removed accordingly. The sutured HAMs on the Netwell TM inserts were immersed in DMEM/F12 medium containing 100 U/mL Penicillin and 100 μg/mL Streptomycin and kept at 37 °C, 5% CO 2 , and 95% air overnight to obtain HAM free of any glycerol remains.
Following corneal transplantation, the remaining human corneal-scleral rings from three donors (n = 3) were divided into twelve limbal biopsies of equal size and thoroughly washed with DMEM/F12 medium containing 100 U/mL Penicillin and 100 μg/mL Streptomycin. The biopsies were treated with neutral protease and Dispase II (2.4 U/mL 4942078001, Roche Diagnostics, Mannheim, Germany) for 10 min at 37 °C. The dissociation process was blocked using Fetal Bovine Serum (FBS, F2442). The limbal biopsies were then placed centrally on the top of the HAMs, with the epithelial side down and submerged in a standardly used complex medium (COM). COM consisted of DMEM/F12, Penicillin (100 U/mL), Streptomycin (100 μg/mL), Amphotericin B (2.5 μg/mL), human epidermal growth factor (2 ng/mL, E9644), insulin (5 μg/mL), sodium selenite (5 ng/mL) and transferrin (5 μg/mL, l1884), cholera toxin A subunit from Vibrio cholerae (30 ng/mL, C8180), hydrocortisone (0.03 μg/mL, H0888), 5% FBS, and 0.5% dimethyl sulfoxide (DMSO, D2650). After 2 h of incubation, the attached limbal biopsies were completely covered by COM. Further, the limbal biopsies and outgrowing LESCs were harvested and incubated at 37 °C with 5% CO 2 , and 95% air for the following three weeks. The culture medium was changed every three days.
Limbal biopsies with hLESCs cultured on HAMs were cut from the Netwell TM inserts. Samples were fixed in 4% formalin overnight at 4 °C and then processed in dehydrated graded alcohol series of 70% (10–15 min), 80% (10–15 min), 96% (2 × 10 min), and 100% ethanol (2 × 10 min) before having xylene added (3 × 10 min) and being washed with melted paraffin (3 × 10 min) and then embedded in paraffin for immunohistochemistry (IHC). Paraffinized tissue was cut into 3–4 μm thick sections using an automated microtome (HM 355S, Thermo Fisher Scientific, Waltham, MA, USA) and attached to histological slides. Deparaffinization was performed in xylene (2 × 10 min), then rehydration was performed by sinking in 100%, 96%, and 70% ethanol, and then distilled water. Hematoxylin & Eosin (H&E) staining was primarily performed. Slides were immersed in Mayers hematoxylin plus solution (01825, Histolab, Askim, Sweden) for 10 min and then rinsed with distilled water (10 min) followed by eosin staining (10 min), and then they were rehydrated in upgraded alcohol series of 70%, 96%, 100% ethanol, and xylene. Slides were further mounted using Pertex (00840, Histolab) mounting medium. For IHC, heat-induced antigen retrieval was performed in a microwave for 5 min at 900W and 15 min at 600W in a citrate buffer (pH 6, C9999) or by PT module (LabVision, Fremont, CA, USA). Blocking of non-specific binding sites with 5% Bovine Serum Albumin (BSA, A9418) dissolved in Dulbecco’s Phosphate Buffered Saline (DPBS, 14190-144, Thermo Fisher Scientific) was conducted for 20 min. Further, slides were stained with primary antibodies diluted in 1% BSA for 1 h. Slides were stained using antibodies for the following progenitor markers: tumor protein p63 alpha (p63α, rabbit polyclonal, 1:200 dilution, 4892S, Cell Signaling, Beverly, MA, USA), SRY-Box Transcription Factor 9 (SOX9, 82630, Cell Signaling, rabbit monoclonal, 1:200), quiescence marker: CCAAT/enhancer-binding protein delta (CEBPD, rabbit polyclonal, 1:200 dilution, ab198320, Abcam, Cambridge, UK), and proliferation marker Ki-67 (rabbit monoclonal, 1:200, RM-9106-S, Thermo Scientific) and the following differentiation markers: cytokeratin 3/12 (KRT3/12, mouse monoclonal, 1:100 dilution, 08691431, MP biomedicals, Santa Ana, CA, USA) and connexin-43 (CX43, rabbit polyclonal, 1:300, C6219). Then, the slides were thoroughly washed three times for 5 min with PBS-tween buffer (28352, Thermo Fisher Scientific). Incubation was continued with the appropriate animal type of secondary antibody: Cy3 ® goat anti-rabbit IgG (rabbit monoclonal, 1:500 dilution, A10520, Abcam) for samples stained with p63α, CEBPD, SOX9, Ki67, and Alexa Fluor ® 488 donkey anti-mouse IgG (1:500 dilution, mouse monoclonal, 21202, Abcam) for samples stained with KRT3/12, and Alexa Fluor ® 488 donkey anti-rabbit IgG (1:500 dilution, rabbit monoclonal, A21206, Abcam) for antibody staining CX43. The secondary antibody was incubated for 45 min and washed three times for 5 min. Nuclear staining was performed using a 4′,6-daminidino-2-phenylindole (DAPI) mounting solution (P36931, Life technologies corporation, Carlsbad, CA, USA). Further, LabVision Autostainer 360 (Lab Vision Corporation, VT) was used for staining with antibodies against adherent junction molecules such as E-cadherin (CDH1, mouse monoclonal, 1:50 dilution, n1620, DakoCytomation, Santa Clara, CA, USA) and N-cadherin (CDH2, mouse monoclonal, 1:100 dilution, m3613, DakoCytomation). Visualization was done using the standard peroxidase technique (UltravisionOne HRP system, Thermo Fisher Scientific). Primary antibody binding to an expressed antigen was recognized by a secondary antibody conjugated with peroxidase-labeled polymer with diaminobenzidine (DAB). Each staining was performed at least three times, and each sample was tested in triplicate. Negative and positive controls were performed simultaneously for all antibodies. All antibodies used for IHC in this study are summarized in . Bright-field images of H&E and DAB-stained samples were taken by a ZEISS Axio Observer Z1 microscope (ZEISS, Oberkochen, Germany). Fluorescence was recorded by a ZEISS Axio Imager M1 fluorescence microscope (ZEISS). Three independent individuals used Image J software and counted nuclear antibody positivity (p63α, CEBPD, SOX9, and Ki67).
The technical replicates from the same donor and group of three donors of the hLESC harvested in two different conditions were averaged as a percentage mean ± standard error of the mean (SEM). Prism 8.3.0 (GraphPad, San Diego, CA, USA) was used for statistical analysis. The data were counted and analyzed by two different methods: in percentages representing the ratio of the number of cells positive for a specific marker and the total number of cells (DAPI positivity), or as the number of cells positive for a specific marker per mm 2 . Further, data were tested for normal distribution (Shapiro–Wilk test), and the difference was tested using an unpaired two-sample t-test. The significance level p ≤ 0.05 was counted as significant.
3.1. Epithelial and Basement Membrane (BM) Morphology in Corneal-limbal Tissue and Consequent Localization of hLESCs The distribution of hLESCs was examined in the different BM compartments of the human corneal-limbal tissue in situ and compared to the hLESCs cultured on conventionally flat and alternatively, HAM sutured in a radial pattern, mimicking limbal crypts ex vivo . The human corneal epithelium had 5–7 layers on the flat BM and an avascular Bowman’s layer ( A). The anterior limbus contained 7–10 epithelial layers on the irregular BM and vascularized stroma underneath ( B). The posterior limbal epithelium was attached to the undulated BM and limbal epithelial crypts that were placed deeper and were mainly surrounded by the limbal stroma ( C). The hLESCs were smaller in size, with a high nucleo-cytoplasmic (N-: C) ratio, and could be randomly detected in the basal epithelial layer of the anterior limbus ( B, black arrows). However, the hLESCs seemed to be more present and densely packed in the basal layer of the posterior limbus and limbal epithelial crypts ( C, black arrow). 3.2. Morphology of the hLESC Cultures Expanded on Conventional, Flat-sutured HAMs vs. hLESC Cultures Expanded on the Novel, Radially-sutured HAMs HAMs sutured by the novel radial suture technique comprised of flat and crypt-like areas. Furthermore, the crypts of the HAMs sutured by the novel radial suture technique consisted of (1) undulated HAM areas with the opened surface ( E) and (2) looped HAM areas that appeared to be almost closed ( F, black asterisk). The multi-layering of the epithelial cells was noted in ex vivo expanded hLESC cultures lying on the undulated and looped HAMs compared to cultures lying on the flat HAM ( E,F, black arrows vs. D). A higher presence of columnar-like epithelial cells was noted in the cultures harvested on the undulated ( E) and looped HAMs ( F) compared to cultures harvested on the flat HAM ( D). The polygonal and squamous cells were found in the middle and superficial layers of the hLESC cultures on the flat HAMs ( D). These polygonal and squamous cells appeared to be less present in the hLESC cultures in the crypt-like HAM compartments. To better understand the structural differences in cultivated tissue on flat and crypt-like HAMs, we aimed to compare the marker fingerprint of cultures growing on flat and looped-like HAMs, as these are two morphologically distinct settings. 3.3. Distribution of the Progenitor Markers In Situ Versus In Vitro Study Conditions The progenitor marker p63α was found in some of the cells of the basal and suprabasal layers of the cultures expanded on the flat and undulated HAMs ( .1A,B). However, the HAM loops contained a statistically higher number of p63α-positive hLESCs ( .1C) than cultures on the flat HAM, quantified as percentages: p63α vs. DAPI positivity ratio (flat vs. loop, 37.56 ± 3.34% vs. 62.53 ± 3.32%, p = 0.01, Figure 5A) or as a total number per mm 2 (377.8 ± 34.17 vs. 962.9 ± 167.2, p = 0.03, Figure 5B). Regarding the epithelium in the corneal-limbal tissue, p63α was not found in any of the cells of the corneal epithelium in situ ( .2A) but was identified in some cells of the basal and suprabasal layers of the anterior limbus ( .2B, arrow). The posterior limbal epithelium with undulated BM was enriched with p63α-positive cells in the basal and suprabasal layers ( .2C). Basal and suprabasal cells in the cultures expanded on flat ( .1D), undulated ( .1E), and looped ( .1F) HAMs expressed the SOX9 progenitor marker. However, the SOX9 progenitor marker positivity was significantly higher in the cultures expanded on crypt-like HAMs forming loops than in the cultures growing on flat HAMs, quantified as percentages (35.53 ± 0.96% vs. 43.23 ± 2.32%, p = 0.04, Figure 5A) or as a total number per mm 2 (442.3 ± 62.31 vs. 728.1 ± 65.97, p = 0.03, Figure 5B). In situ, the progenitor marker SOX9 was exclusive for the limbal basal epithelium ( .2E). In particular, the limbal epithelial crypts appeared to be enriched for this marker ( .2F). 3.4. Expression Profile of the Proliferation and Quiescence Markers in the Corneal-limbal Epithelial Tissue Versus In Vitro Study Conditions The expression distribution of the CEBPD marker was similar to p63α and present in the cultures expanded on flat ( .1A), undulated ( .1B), and looped HAMs ( .1C)—mainly in basal and suprabasal layers. In addition, CEBPD was found in the basal epithelial cells of both tissues in situ, the cornea ( .2D) and the anterior and posterior limbus ( .2B,C), accordingly. However, no statistical significance was noted in the number of CEBPD-positive cells expanded on HAM loops compared to cells expanded on flat HAMs, quantified as percentages (22.99 ± 2.96% vs. 30.49 ± 3.33 %, p = 0.17, A) or as a total number per mm 2 (243.00 ± 35.19 vs. 474.1 ± 138.5, p = 0.18, B). Many of the hLESCs expanded on a flat ( .1D, white arrow), undulated ( .1E), and looped HAM ( .1F) were found in the proliferation state. However, some sectors of the epithelial tissue on the flat HAM contained no Ki-67-positive cells ( .1D), while sections of the epithelial tissue on the undulated ( .1E) and/or looped HAM ( .1F) persistently maintained Ki-67-positive cells. Proliferation was significantly higher in cultures expanded on looped HAMs compared to the cultures on flat HAMs, quantified as percentages (8.43 ± 0.38 % vs. 22.38 ± 1.95 %, p = 0.002, A) or as a total number per mm 2 (100.7 ± 10.69 vs. 276.2 ± 33.34, p = 0.01, B). For comparison to the in situ state, proliferation marker Ki-67 was sporadically found in the suprabasal cells of the anterior ( .2E) and posterior ( .2F) limbal epithelium, whereas it was absent in the central corneal epithelium, and only sparsely present in some basal cells of the posterior cornea ( .2D). 3.5. Differentiation Marker profile in the Epithelium of the Corneal-limbal Tissue, and hLESC Cultures on the Flat and Crypt-like HAMs CX43 was uniformly distributed in all ex vivo expanded cells ( .1A–C), a finding similar to the CX43 pattern in the corneal epithelium in situ ( .2A). However, some basal cells in the anterior ( .2B) and posterior limbal epithelium and limbal epithelial crypts ( .2C) appeared to lack the CX43 marker. In the expanded hLESC cultures growing on flat HAMs, the differentiation marker KRT3/12 was present in the polygonal and squamous cells, mainly in the middle and top layers ( .1D). Less KRT3/12 presence could be noted in the cultures expanded on undulated ( .1E) and loop HAMs ( .1F), whereas this marker was almost absent in cells growing in small HAM loops. Regarding the corneal-limbal tissue, KRT3/12 was present in all corneal epithelial cells ( .2D). In the limbal epithelium, the majority of the cells were stained positive for the KRT3/12, whereas the cells in the lowest layers attached to the BM were devoid of this marker ( .2E,F). 3.6. Presentation of Cell Adhesion Molecules in the Corneal-Limbal Epithelium and Expanded hLESC Cultures on Flat and Crypt-like HAMs The transmembrane protein E-cadherin was present in most of the ex vivo expanded epithelial cells ( .1A–C). The same applied to the corneal-limbal tissue in situ ( .2A–C). In hLESC cultures, N-cadherin was present in a few cells of the basal layer on the flat HAMs ( .1D). On the other side, more cells in the basal layer of the crypt-like HAMs seemed to express N-cadherin since the surface of the basal layer of the cultivated tissue appeared enlarged in those crypts compared to the cultures on the flat HAMs ( .1E,F). N-cadherin was exclusive for the limbal basal epithelium in situ. Only a few basal cells in the epithelium of the anterior limbus expressed N-cadherin ( .2E). In contrast, almost all cells of the limbal basal epithelium in the posterior limbus expressed N-cadherin ( .2F).
The distribution of hLESCs was examined in the different BM compartments of the human corneal-limbal tissue in situ and compared to the hLESCs cultured on conventionally flat and alternatively, HAM sutured in a radial pattern, mimicking limbal crypts ex vivo . The human corneal epithelium had 5–7 layers on the flat BM and an avascular Bowman’s layer ( A). The anterior limbus contained 7–10 epithelial layers on the irregular BM and vascularized stroma underneath ( B). The posterior limbal epithelium was attached to the undulated BM and limbal epithelial crypts that were placed deeper and were mainly surrounded by the limbal stroma ( C). The hLESCs were smaller in size, with a high nucleo-cytoplasmic (N-: C) ratio, and could be randomly detected in the basal epithelial layer of the anterior limbus ( B, black arrows). However, the hLESCs seemed to be more present and densely packed in the basal layer of the posterior limbus and limbal epithelial crypts ( C, black arrow).
HAMs sutured by the novel radial suture technique comprised of flat and crypt-like areas. Furthermore, the crypts of the HAMs sutured by the novel radial suture technique consisted of (1) undulated HAM areas with the opened surface ( E) and (2) looped HAM areas that appeared to be almost closed ( F, black asterisk). The multi-layering of the epithelial cells was noted in ex vivo expanded hLESC cultures lying on the undulated and looped HAMs compared to cultures lying on the flat HAM ( E,F, black arrows vs. D). A higher presence of columnar-like epithelial cells was noted in the cultures harvested on the undulated ( E) and looped HAMs ( F) compared to cultures harvested on the flat HAM ( D). The polygonal and squamous cells were found in the middle and superficial layers of the hLESC cultures on the flat HAMs ( D). These polygonal and squamous cells appeared to be less present in the hLESC cultures in the crypt-like HAM compartments. To better understand the structural differences in cultivated tissue on flat and crypt-like HAMs, we aimed to compare the marker fingerprint of cultures growing on flat and looped-like HAMs, as these are two morphologically distinct settings.
The progenitor marker p63α was found in some of the cells of the basal and suprabasal layers of the cultures expanded on the flat and undulated HAMs ( .1A,B). However, the HAM loops contained a statistically higher number of p63α-positive hLESCs ( .1C) than cultures on the flat HAM, quantified as percentages: p63α vs. DAPI positivity ratio (flat vs. loop, 37.56 ± 3.34% vs. 62.53 ± 3.32%, p = 0.01, Figure 5A) or as a total number per mm 2 (377.8 ± 34.17 vs. 962.9 ± 167.2, p = 0.03, Figure 5B). Regarding the epithelium in the corneal-limbal tissue, p63α was not found in any of the cells of the corneal epithelium in situ ( .2A) but was identified in some cells of the basal and suprabasal layers of the anterior limbus ( .2B, arrow). The posterior limbal epithelium with undulated BM was enriched with p63α-positive cells in the basal and suprabasal layers ( .2C). Basal and suprabasal cells in the cultures expanded on flat ( .1D), undulated ( .1E), and looped ( .1F) HAMs expressed the SOX9 progenitor marker. However, the SOX9 progenitor marker positivity was significantly higher in the cultures expanded on crypt-like HAMs forming loops than in the cultures growing on flat HAMs, quantified as percentages (35.53 ± 0.96% vs. 43.23 ± 2.32%, p = 0.04, Figure 5A) or as a total number per mm 2 (442.3 ± 62.31 vs. 728.1 ± 65.97, p = 0.03, Figure 5B). In situ, the progenitor marker SOX9 was exclusive for the limbal basal epithelium ( .2E). In particular, the limbal epithelial crypts appeared to be enriched for this marker ( .2F).
The expression distribution of the CEBPD marker was similar to p63α and present in the cultures expanded on flat ( .1A), undulated ( .1B), and looped HAMs ( .1C)—mainly in basal and suprabasal layers. In addition, CEBPD was found in the basal epithelial cells of both tissues in situ, the cornea ( .2D) and the anterior and posterior limbus ( .2B,C), accordingly. However, no statistical significance was noted in the number of CEBPD-positive cells expanded on HAM loops compared to cells expanded on flat HAMs, quantified as percentages (22.99 ± 2.96% vs. 30.49 ± 3.33 %, p = 0.17, A) or as a total number per mm 2 (243.00 ± 35.19 vs. 474.1 ± 138.5, p = 0.18, B). Many of the hLESCs expanded on a flat ( .1D, white arrow), undulated ( .1E), and looped HAM ( .1F) were found in the proliferation state. However, some sectors of the epithelial tissue on the flat HAM contained no Ki-67-positive cells ( .1D), while sections of the epithelial tissue on the undulated ( .1E) and/or looped HAM ( .1F) persistently maintained Ki-67-positive cells. Proliferation was significantly higher in cultures expanded on looped HAMs compared to the cultures on flat HAMs, quantified as percentages (8.43 ± 0.38 % vs. 22.38 ± 1.95 %, p = 0.002, A) or as a total number per mm 2 (100.7 ± 10.69 vs. 276.2 ± 33.34, p = 0.01, B). For comparison to the in situ state, proliferation marker Ki-67 was sporadically found in the suprabasal cells of the anterior ( .2E) and posterior ( .2F) limbal epithelium, whereas it was absent in the central corneal epithelium, and only sparsely present in some basal cells of the posterior cornea ( .2D).
CX43 was uniformly distributed in all ex vivo expanded cells ( .1A–C), a finding similar to the CX43 pattern in the corneal epithelium in situ ( .2A). However, some basal cells in the anterior ( .2B) and posterior limbal epithelium and limbal epithelial crypts ( .2C) appeared to lack the CX43 marker. In the expanded hLESC cultures growing on flat HAMs, the differentiation marker KRT3/12 was present in the polygonal and squamous cells, mainly in the middle and top layers ( .1D). Less KRT3/12 presence could be noted in the cultures expanded on undulated ( .1E) and loop HAMs ( .1F), whereas this marker was almost absent in cells growing in small HAM loops. Regarding the corneal-limbal tissue, KRT3/12 was present in all corneal epithelial cells ( .2D). In the limbal epithelium, the majority of the cells were stained positive for the KRT3/12, whereas the cells in the lowest layers attached to the BM were devoid of this marker ( .2E,F).
The transmembrane protein E-cadherin was present in most of the ex vivo expanded epithelial cells ( .1A–C). The same applied to the corneal-limbal tissue in situ ( .2A–C). In hLESC cultures, N-cadherin was present in a few cells of the basal layer on the flat HAMs ( .1D). On the other side, more cells in the basal layer of the crypt-like HAMs seemed to express N-cadherin since the surface of the basal layer of the cultivated tissue appeared enlarged in those crypts compared to the cultures on the flat HAMs ( .1E,F). N-cadherin was exclusive for the limbal basal epithelium in situ. Only a few basal cells in the epithelium of the anterior limbus expressed N-cadherin ( .2E). In contrast, almost all cells of the limbal basal epithelium in the posterior limbus expressed N-cadherin ( .2F).
Different techniques of HAM suturing to the corneal surface have been used thus far. A HAM can be sutured as a graft (inlay) or as a patch (overlay). While used as a graft, a HAM is placed on the defect with the stromal face down and acts as a BM, allowing the epithelium to proliferate and regenerate over it. It can be used as a single or multilayered graft with a lamellar sac, filling, or roll-filling technique, mostly depending on the depth of the corneal defect. The patch technique is mostly used for epithelial defects without perforations. The epithelial side of the HAM is placed down towards the defect, and the HAM serves as a biological compressive bandage . Graft alone or as part of a sandwich technique, which is a combination of both graft and patch techniques, has been standardly used for HAMs carrying cultivated hLESCs and limbal explants . However, this is the first study to propose the manipulation of a HAM prior to the expansion of the hLESCs to ensure a better quality of the transplanted tissue as an adjuvant technique to the previously used HAM suturing techniques. Stem cell niches vary in size and functional organization in mammals . Stem cells can be found as individual structures under the BM of the skeletal muscle , or grouped as epithelial stem cells in the hair follicle bulges and neural stem cells in the forebrain subventricular zone in mammals . In this study, we provided a more optimal microenvironment for the expansion of hLESCs ex vivo by mimicking the BM folding in niches residing in the posterior limbus and limbal epithelial crypts. Generally, the maintenance of stem cells, including hLESCs, depends on a functional niche characteristic. These niches provide cell anchoring, mechanical protection, communication with underlying stroma and vasculature, the release of specific growth factors, cell cycle molecules, and the involvement of evolutionary conserved molecular pathways. Such 3D microenvironments allow the stem cells to hold the quiescence, maintain stemness, and undergo asymmetric or symmetric proliferation when needed . Our study supports earlier findings that most hLESC/early progenitor cells reside on the bottom of the limbal epithelial crypts, which are deep epithelial protrusions directly surrounded by a loose stromal matrix . When an epithelial stem cell niche is established along the stiff BM, it maintains its regular morphology. However, when the epithelial stem cell niche forms along a flexible and extensible BM, it may arrange in the form of finger-like protrusions, enabling a higher surface for stem cells to allocate, thus providing the protection and preservation of the putative stem cell characteristics . Stem cell progenies acquire differentiation properties by leaving the stem cell niche towards the more rigid and flat ground, such as the Bowman membrane, to eventually terminally differentiate and undergo apoptosis (26). A HAM is a desirable elastic and adaptable scaffold for creating 3D protrusions that can physically mimic limbal crypts ex vivo. It is a widely available natural semi-transparent and permeable membrane. Its mechanical and functional characteristics are desirable for the migration, adhesion, and growth of epithelial cells on the ocular surface. It possesses high elasticity, low stiffness, and high tensile strength properties , and it also resembles the cornea and conjunctiva in regards to the collagen arrangement . The stiffness should be similar between flat vs. crypt-like HAMs, as we used pieces from the same donor. Even though there might be some local differences in stiffness within the same HAM, we used nine pieces of both flat and crypt-like HAM, and all the pieces showed significant changes related to the suturing method. There was only one extra suture on the crypt-like vs. flat HAMs to enable the folding, so this should not have affected the overall stiffness of the crypt-like vs. flat HAMs. Functionally, HAM is immunotolerant and has low antigenicity, even though some immunomodulatory effects have been reported; it has an anti-fibrotic impact, mainly due to the TGF-β inhibition. It secretes a wide range of growth factors, such as EGF, bFGF, HGF, KGF and KGF receptors, TGFα, and TGFβ 1,2,3 isoforms, sharing some common features with stem cell niche composition . However, not all HAM properties seem beneficial for stem cell maintenance. An intact HAM promotes the epithelial differentiation of explanted limbal cultures. Therefore, removing the epithelium from the HAM upon preparation has been used by some authors to maintain progenitor properties, postpone differentiation, and thus, improve the quality of the explanted tissue . Also, not all of the cells expanded on a HAM have the features of hLESCs or early progenies. As previously shown, most of the hLESC/progenies are positioned in the basal epithelial cell layers—the ones attached to the HAM. In contrast, the cells in the upper/superficial layers exhibit more differentiation properties . With our novel suturing technique, we aimed to enlarge the surface area of the HAM, and hence enlarge the number of cells in the basal layer attached to the HAM, maintaining the more undifferentiated state. In addition, the expanded epithelial tissue appeared multilayered in the crypt-like HAMs, and it contained a higher number of columnar-like cells, likely indicating a higher proliferation rate. The novel suturing technique would also increase the supply of the stem-cell-supporting molecules secreted by the HAM. HAM is widely implemented in tissue engineering and regenerative medicine . However, its limited chemical and physical features, and the high cost of preserving it in a fresh condition, caused the urgency for new solutions . HAM has, so far, undergone additional adjustments to upgrade its properties for easier manipulation, duration, and utilization, and for higher resistance to microbes and broader applications, among others, in ocular surface reconstruction . For instance, AM can be used as a constituent of various composite scaffolds, in a form of extract, and as a hydrogel . Regarding the attempts of HAM modification for successful LESC transplantation, decellularized AM (dAM) conjugated with an electrospun polymer nanofiber mesh promoted LESC proliferation and adhesion in a rabbit model . An amniotic membrane in a form of extract (AME) and eye drops proved beneficial for the treatment of ocular surface disorders, injuries, and the in vivo cultivation of hLESCs . In addition, AME, as an animal-free product, was suggested as a suitable replacement for FBS upon LSC transplantation to avoid the risk of disease transmission and accumulation of bovine antigens . However, all the above-modified HAM methods require very complex processing or serve only as adjuvant therapy. Therefore, we present an easy-handling, widely available, and inexpensive method of HAM manipulation prior to hLESC expansion. As previously mentioned, the successful long-term restoration of the corneal epithelium after CLET requires more than 3% of the cells in the transplanted graft to be p63 positive . In our study, cells growing on either the flat or crypt-like HAMs were enriched with the p63α marker. However, we found a significantly larger number of cells positive for p63α in the looped regions of the crypt-like HAMs, compared to the cells growing on flat HAMs. Initially, the whole tumor protein p63 was perceived as a specific marker for hLESCs . Later studies discovered ΔNp63α to be more distinct for hLESCs and for early progenies residing in the limbal basal epithelium, while other p63 isoforms were detected in the suprabasal layers of the limbus and cornea, playing a role in corneal differentiation . Indeed, in our samples, p63α stained the particular cells in the basal limbal epithelium, while staining in the non-limbal cornea was absent. Significantly higher cell turnover was present in the cultures on the crypt-like HAMs compared to the cultures growing on flat HAMs, indicating the presence of cells with intense proliferation, such as early progenies/TACs. Compared to the in situ state, proliferation appeared to be much lower in corneal-limbal tissue, and it was noted in a few suprabasal cells of the anterior and posterior limbus and some cells in the posterior cornea as previously described . The CEBPD marker was not significantly more abundant in the cultures growing on crypt-like HAMs compared to those on flat-like HAMs. CEBPD is a quiescence marker that controls the cell cycle and inhibits the proliferation of hLESCs in ex vivo cultures . Since proliferation was significantly higher in ex vivo cultures growing on crypt-like HAMs compared to those growing on flat-like HAM, we expected these cells not to be positive for CEBPD; hLESCs do not co-express CEBPD with the Ki-67 marker in the limbus. In addition, CEBPD-positive hLESCs that co-express the ΔNp63α marker in the basal limbal epithelium in situ are the ones considered quiescent . However, it seems that the CEBPD marker is not specific for hLESCs in situ, since it is also found in the cells of the basal corneal epithelium in our samples. The transcriptional factor SOX9 plays diverse roles in the embryonal and adult development of mammals as well as in stem cell maintenance. This marker was upregulated in the cultures on crypt-like HAMs compared to those on flat-like HAMs. Its nuclear localization in TACs is essential for proliferation upon wound healing. However, SOX9 is particularly involved in the proliferation and differentiation steps of early progenies derived from hLESCs, but not in the terminal differentiation, which explains why SOX9 is absent in the cornea . This finding contributes to our conclusion that crypt-like HAMs contain numerous cells that are positive for SOX9 and are thus in a more undifferentiated state. We found no difference in the presence of the CX43 transmembrane protein between the cultures on crypt-like HAM and flat HAM. CX43 is a protein involved in the communication between mammalian cells through diverse mechanisms . Constitutionally, CX43 is present in the corneal epithelium and all suprabasal epithelial layers in the limbus, whereas it is absent in some cells of the basal limbal layer . A similar pattern of expression of CX43 applies to expanded cultures on the flat HAMs . The absence of cell interaction may be one of the mechanisms for maintaining stemness and the quiescence state . Thus, according to some authors, CX43 positivity in cells distinguishes the hLESCs from the TACs/early progenies in vivo. Since most of the cells in the basal and suprabasal layers of the cultivated epithelial tissue were CX43 positive, it seems that a very small cell fraction remains in a quiescent state ex vivo. Cytokeratin 3 (KRT3) and cytokeratin 12 (KRT12) are cornea-specific intermediate filaments, hallmarks of differentiated hCECs in the cornea and differentiated epithelial cells in the limbus . In cultured epithelial tissue, KRT3/12 has been found in the suprabasal and superficial layers, but not in the basal layer of cells cultured on flat HAM, a finding corresponding to our results . However, the KRT3/12 marker was reduced in the looped regions of the crypt-like HAMs. In some small looped regions, the respective marker was absent. The KRT3/12 marker is known to be absent from limbal basal layers—a finding that shows a more mature nature of the corneal basal cells compared to the limbal basal epithelial cells, due to different characteristics of the corresponding basal membranes . Also, KRT3/12 is absent from limbal epithelial crypts . Our study shows that unique cell fractions in the basal layers attached to the HAM are positive for the N-cadherin marker. It seems that the isolation of the cell cultures surrounded by a double HAM membrane increases cell positivity for this marker. N-cadherin is essential for maintaining the progenitor characteristics in cultured hLESCs . Differentiated corneal and limbal epithelial cells express E-cadherin, while N-cadherin is present in the hLESCs/progenitor cells in the limbal basal epithelium, a finding concomitant with ours. In particular, our basal epithelial layer in the posterior limbus appeared to be enriched with this marker. It is suggested that communication with the melanocytes is achieved via N-cadherin forming homotypic adhesions . A disadvantage of our technique may be that a larger area of HAM tissue is needed for transplantation and, evidently, a decreased transparency of the cultivated transplantation graft, in addition to any usual disadvantages of using HAM .
In conclusion, this novel HAM suturing technique increased the number of progenitor cells upon expansion and may thus increase the quality of the transplanted graft. We believe this technique can be a valuable, simple, and inexpensive tool to increase the success rate of corneal epithelial regeneration. However, the suturing technique needs to be tested in vivo to confirm its efficacy. Future clinical studies comparing conventional, flat suturing and the current suturing method are also required.
|
Enhanced alveolar ridge preservation with hyaluronic acid-enriched allografts: a comparative study of granular allografts with and without hyaluronic acid addition | 28bfa0ab-eebd-40a9-82b1-f99f2bde04a4 | 11465134 | Dentistry[mh] | Alveolar ridge preservation is a vital procedure in dentistry and oral surgery, aimed at minimizing bone resorption and maintaining the alveolar ridge’s volume and morphology following tooth extraction . Loss of alveolar bone after extraction can compromise the esthetic and functional outcomes of dental implant placement and prosthetic rehabilitation . Various techniques, including autografts, allografts, and xenografts, provide structural support and scaffolding for new bone formation, promoting osteogenesis and osteoconduction . Additionally, guided bone regeneration (GBR) techniques use barrier membranes to facilitate undisturbed bone regeneration, while growth factors like platelet-rich fibrin and bone morphogenetic proteins accelerate tissue regeneration and enhance bone formation . Despite the advancements, there is no consensus on the optimal ridge preservation technique for different clinical scenarios . Studies have highlighted significant dimensional changes in hard and soft tissues post-extraction, with substantial reductions in horizontal and vertical dimensions impacting treatment planning and prosthetic rehabilitation . Current methods, although effective, vary in outcomes, necessitating further exploration of innovative materials and approaches . Hyaluronic acid has emerged as a promising adjunct in alveolar ridge preservation due to its multifaceted benefits in tissue regeneration and wound healing . Its viscoelastic properties help maintain hydration and structural integrity within the graft site, thereby reducing graft shrinkage and promoting vertical stability. Hyaluronic acid interacts with cell surface receptors to stimulate osteoblast activity and inhibit osteoclastogenesis, leading to enhanced bone density and improved integration of the graft material. Clinical studies have demonstrated that hyaluronic acid not only modulates inflammatory responses and enhances angiogenesis but also synergistically enhances the osteoconductive properties of graft materials, making it an effective component for improving outcomes in alveolar ridge preservation . This study aimed to evaluate the clinical efficacy of ridge preservation using a hyaluronic acid-enriched allogeneic bone substitute material in patients with compromised extraction sockets. Over a 12-month period, the intervention group receiving the hyaluronic acid-enriched allograft was compared to a control group receiving an allogeneic bone substitute material without hyaluronic acid. The outcomes assessed included horizontal and vertical bone gain, volume stability, graft shrinkage, and bone mineral density. This comparison aimed to determine the potential benefits of adding hyaluronic acid to granular allografts in enhancing bone regeneration and soft tissue healing. This is the first clinical study to compare allogeneic bone substitutes with and without hyaluronic acid for alveolar ridge preservation.
Patients A total of 40 patients presenting with decayed teeth necessitating extraction were enrolled in this retrospective study. Post-extraction, radiological evaluation using cone-beam computed tomography (CBCT) revealed defects in the buccal cortical plate, consistent with the clinical presentation of compromised extraction sockets. All patients needed single-implant treatments. After four months of healing time, every patient received one titanium implant per augmented region. All patients were fully informed about the surgical procedures and treatment alternatives. The minimum extraction socket defect type for inclusion was a Type-IV bone defect, as defined by Kim et al. 2021 . Exclusion criteria consisted of a history of radiotherapy in the head and neck region, systemic disease that would contraindicate oral surgery, uncontrolled periodontal disease, bruxism, pregnancy, psychiatric problems, and/or use of medications known to alter bone healing. Prior to surgery, patients were presented with two alternative procedures for ridge preservation (allogeneic bone grafts with or without hyaluronic acid). All patients were provided with standardized information sheets regarding allogeneic ridge preservation and dental implantation. In addition, patients were informed about the optional use of hyaluronic acid. The decision to opt for this addition was left entirely to the discretion of the patients, as no conclusive evidence regarding potential added benefits could be provided at the time of surgery. In order to reduce potential sources of bias, patients were selected for each of the two study groups so that they did not differ in demographic or anamnestic characteristics (Table ). After screening the available clinical data, 19 patients with comparable demographic characteristics were allocated the “Allo” treatment group (no hyaluronic acid) and 21 patients were allocated the “AlloHya” treatment group (with hyaluronic acid). Power calculations For a statistical power calculation, we made the following assumptions: the average resorption rate for grafts materials ranged between 14.4% (SD 9.0) for allogeneic bone blocks to 33.4% (SD 3.1) for xenogeneic granular graft materials . Since the allogeneic material that we used was granular, we assumed that the resorption rate of the granular allogeneic material used in this study was similar to the resorption rate from xenogeneic granular graft materials. With these assumptions, we had statistical power of 100% to detect significant differences between the two study groups with 20 patients each and with a level of significance of 5%. Power calculations were performed with RStudio using the package “pwr”. Grafting materials Two variants of allogeneic bone substitutes in granular form were utilized. On the one hand, granular maxgraft ® was used (study group “Allo”), and on the other hand, granular maxgraft ® + Hya (study group “AlloHya”) was applied (botiss biomaterials GmbH, Berlin, Germany). maxgraft ® granules comprise allograft bone substitute derived from human donor bone, meticulously processed by the Cells + Tissuebank Austria using the specialized cleaning procedure known as the Allotec ® process. Offered in both cancellous and cortico-cancellous forms, maxgraft ® retains its natural bone structure and collagen composition. Serving as an effective scaffold, it facilitates the natural regeneration of bone tissue and holds the potential for complete assimilation into the patient’s own bone through remodeling . maxgraft ® + Hya merges the allogeneic bone grafting material maxgraft ® with the hydrophilic properties of hyaluronic acid. Utilizing the notable liquid-binding capabilities of hyaluronate, maxgraft ® + Hya transforms into a cohesive and adhesive bone material upon hydration, commonly referred to as “sticky bone”. This transformation enhances application convenience by enabling straightforward uptake and precise delivery to the target site, thereby improving procedural efficiency. Prior to application, maxgraft ® + Hya requires rehydration. According to the manufacturer’s protocol, approximately 0.8 ml of saline solution is used per 1 ml of maxgraft ® + Hya. Surgical procedure Prior to extraction blood was drawn from the patient to prepare A-PRF (advanced platelet rich fibrin) following the protocol of Choukroun with a Mectron centrifuge . After atraumatic tooth extraction, the extraction site was thoroughly mechanically cleaned to remove any remaining debris and granulation tissue (Fig. ). Once the site was prepared, allogeneic grafting material was rehydrated according to the manufacturer’s instructions with sterile saline and introduced into the socket to fill the void left by the extracted tooth (Fig. ). The bone substitute was softly condensed. After placing the grafting material, A-PRF was positioned over the allograft and fixed with resorbable sutures (Fig. ). Post-operative care instructions, including oral hygiene measures and dietary restrictions, were provided to the patient to ensure optimal healing and reduce the risk of complications. Routine post-operative care included administration of amoxicillin and clavulanic acid (625 mg, administered orally, three times a day for 4 days), ibuprofen (600 mg, administered orally, every 6 h as needed), and mouthwash (0.2% chlorhexidine, three times daily for 7 days). . After four months of healing time, CBCT scans were done to investigate the osseous healing of the socket, and one titanium implant per augmented region was inserted (Fig. ). Implant survival and success Twelve months post-implantation, patients were recalled for a follow-up evaluation to assess implant success. The assessment was conducted using the guidelines from the International Congress of Oral Implantologists (ICOI) Pisa Consensus Conference on Implant Success, Survival, and Failure. Key parameters evaluated included pain, implant mobility, bleeding on probing, and radiographic bone loss relative to the initial bone level. These parameters were used to categorize the implants according to the ICOI implant quality scale . Radiographic analyses Every patient was subjected to three-dimensional x-ray diagnostics (CBCT). In total three CBCTs were recorded for each patient, one before treatment, one directly after ridge preservation, and one after four months of healing before implantation. At each time point, the alveolar bone levels were measured in their height, width and depth at the cervical level, the middle height of the defect and at the apical level. An illustration of the measured regions is shown in Fig. . The CBCT machine used to acquire all images was the Carestream CS 9300 (Carestream Health Inc., Onex Corporation, Rochester, New York, USA). The imaging parameters were set with a dose of 120 mGy.cm2, tube current of 18.66 mAs, and a voxel size of 90 μm x 90 μm x 90 μm. The selected field of view was 5 cm x 5 cm. Data from the scans were saved in the Digital Imaging and Communications in Medicine (DICOM) format and reconstructed with the Carestream implant planning software program. All measurements were made on parasagittal sections perpendicular to the longitudinal axis of the adjacent teeth. The CBCT was oriented transversally through the teeth neighboring the defect so that the nerve canal of the tooth, which was mesial to the defect region, was visible. The nerve canal of the mesial tooth was defined as an anatomic reproducible landmark and a straight line was drawn through the middle of the defect region between the two neighboring teeth. The mesial tooth was used as a reference for apical and crestal bone levels. The distances were obtained using a software ruler. The same anatomic landmarks and distances were used for measurements on CBCT at the defined time intervals. The following measurements were taken (Fig. ): Defect height (mm): distance between the apical and crestal bone level in the middle of the defect region; represented by line “h” in Fig. . Apical defect width (mm): distance between the apical root tips of the neighboring teeth; represented by line “a” in Fig. . Defect width in the middle zone (mm): distance between the roots of the neighboring teeth in the middle of the defect height; represented by line “b” in Fig. . Cervical defect width (mm): distance between the crestal bone levels of the neighboring teeth; represented by line “c” in Fig. . Apical defect depth (mm): distance between the labial/buccal and palatal edges of the jaw crest at the level of the apical tips of the neighboring teeth, but in the middle of the defect area; represented by line “d” in Fig. . Defect depth in the middle zone (mm): distances between the labial/buccal and palatal edges of the jaw crest at the level of the middle zone; represented by lines “e” (mesial region), “f” (central region) and “g” (distal region) in Fig. . Cervical defect depth (mm): distance between the labial/buccal and palatal edges of the jaw crest at the cervical level; represented by lines “k” (mesial region), “l” (central region) and “m” (distal region) in Fig. . Bone density In order to estimate the bone density of the allograft, Hounsfield Units (HU) were used as a measurement of radiodensity. The integrated measurement module of a picture archiving and communication system software was used (PACS, DeepUnity Diagnost 1.1.1.1, DEDALUS). The regions of interest were defined in the graft’s crestal, middle and apical portions and adjusted to their size (5–20 mm 2 ) at each level. In all cases, CBCT sections with low scattered radiation were chosen. The maximum area within the allograft was selected in such a way that the bony marginal structures of the alveolar process were not touched. Both the Hounsfield Units and the area of the measured region were recorded. For the purpose of determining the average bone density within a designated region of the CBCT scan, total HUs were divided by the area of the region measured. This provided the average radiodensity within that specific area. Mathematics and statistics Based on the radiographic measurements, the graft volume was inferred as the sum of the volumes of two superimposed frustums of pyramids. The formula for obtaining the volume of one pyramid trunk is: [12pt]{minimal}
$$\:{V}_{pyramid\:trunk}=\:(B+P+)\:$$ , where “g ” is the height of the truncated pyramid, “B” is the base area and “P” is the peak area. The two pyramid trunks are depicted in Fig. . The formula for obtaining the volume of the entire defect was therefore: [12pt]{minimal}
$$} = {h 6} ( {{A_{crestal}} + {A_{middle}} + {{A_{crestal}} {A_{middle}}} } ) & + {h 6} ( {{A_{middle}} + {A_{apical}} + {{A_{middle}} {A_{apical}}} } ) }$$ In the formula above, “h” denotes the distance between the apical and crestal bone level in the middle of the defect region. “A crestal ” is the area of the cross-sectional surface of the alveolus at the crestal level. This was calculated from the width “c”, and the three horizontal, buccal-lingual depths “k” (mesial area), “l” (central area) and “m” (distal area). “A middle ” is the area of the cross-sectional surface of the alveolus at the middle height of the defect. This was calculated from the width “b”, and the three horizontal, buccal-lingual depths “e” (mesial area), “f” (central area) and “g” (distal area). “A apical ” is the area of the cross-sectional surface of the alveolus at the apical level. This was calculated from the width “a”, and the buccal-lingual depth “d”. Statistical analyses were performed with IBM SPSS (version 27; International Business Machines Corp., Armonk, NY, USA). Pearson’s chi-squared test was applied to sets of unpaired categorical data to evaluate the likelihood that any observed difference between the sets was due to chance. All metric variables were tested for normal distribution (Shapiro-Wilk test) and homogeneity of variance (Levene’s test) before parametric tests. An independent sample t -test was used when two separate sets of independent and identically distributed samples were obtained, and their population means were compared to each other. Multiple linear regression was used to try to explain an observed outcome variable (bone density) by several independent variables. The categorical variables were added to the model as factors. In the first model, all potential predictors were included. Then, all non-significant variables were removed from the multiple linear regression model in a stepwise manner. Only two-sided significance tests were used. A probability of error of p ≤ 0.05 was chosen as the threshold value. An alpha adjustment for multiple testing was not performed. The results are therefore explorative and descriptive.
A total of 40 patients presenting with decayed teeth necessitating extraction were enrolled in this retrospective study. Post-extraction, radiological evaluation using cone-beam computed tomography (CBCT) revealed defects in the buccal cortical plate, consistent with the clinical presentation of compromised extraction sockets. All patients needed single-implant treatments. After four months of healing time, every patient received one titanium implant per augmented region. All patients were fully informed about the surgical procedures and treatment alternatives. The minimum extraction socket defect type for inclusion was a Type-IV bone defect, as defined by Kim et al. 2021 . Exclusion criteria consisted of a history of radiotherapy in the head and neck region, systemic disease that would contraindicate oral surgery, uncontrolled periodontal disease, bruxism, pregnancy, psychiatric problems, and/or use of medications known to alter bone healing. Prior to surgery, patients were presented with two alternative procedures for ridge preservation (allogeneic bone grafts with or without hyaluronic acid). All patients were provided with standardized information sheets regarding allogeneic ridge preservation and dental implantation. In addition, patients were informed about the optional use of hyaluronic acid. The decision to opt for this addition was left entirely to the discretion of the patients, as no conclusive evidence regarding potential added benefits could be provided at the time of surgery. In order to reduce potential sources of bias, patients were selected for each of the two study groups so that they did not differ in demographic or anamnestic characteristics (Table ). After screening the available clinical data, 19 patients with comparable demographic characteristics were allocated the “Allo” treatment group (no hyaluronic acid) and 21 patients were allocated the “AlloHya” treatment group (with hyaluronic acid).
For a statistical power calculation, we made the following assumptions: the average resorption rate for grafts materials ranged between 14.4% (SD 9.0) for allogeneic bone blocks to 33.4% (SD 3.1) for xenogeneic granular graft materials . Since the allogeneic material that we used was granular, we assumed that the resorption rate of the granular allogeneic material used in this study was similar to the resorption rate from xenogeneic granular graft materials. With these assumptions, we had statistical power of 100% to detect significant differences between the two study groups with 20 patients each and with a level of significance of 5%. Power calculations were performed with RStudio using the package “pwr”.
Two variants of allogeneic bone substitutes in granular form were utilized. On the one hand, granular maxgraft ® was used (study group “Allo”), and on the other hand, granular maxgraft ® + Hya (study group “AlloHya”) was applied (botiss biomaterials GmbH, Berlin, Germany). maxgraft ® granules comprise allograft bone substitute derived from human donor bone, meticulously processed by the Cells + Tissuebank Austria using the specialized cleaning procedure known as the Allotec ® process. Offered in both cancellous and cortico-cancellous forms, maxgraft ® retains its natural bone structure and collagen composition. Serving as an effective scaffold, it facilitates the natural regeneration of bone tissue and holds the potential for complete assimilation into the patient’s own bone through remodeling . maxgraft ® + Hya merges the allogeneic bone grafting material maxgraft ® with the hydrophilic properties of hyaluronic acid. Utilizing the notable liquid-binding capabilities of hyaluronate, maxgraft ® + Hya transforms into a cohesive and adhesive bone material upon hydration, commonly referred to as “sticky bone”. This transformation enhances application convenience by enabling straightforward uptake and precise delivery to the target site, thereby improving procedural efficiency. Prior to application, maxgraft ® + Hya requires rehydration. According to the manufacturer’s protocol, approximately 0.8 ml of saline solution is used per 1 ml of maxgraft ® + Hya.
Prior to extraction blood was drawn from the patient to prepare A-PRF (advanced platelet rich fibrin) following the protocol of Choukroun with a Mectron centrifuge . After atraumatic tooth extraction, the extraction site was thoroughly mechanically cleaned to remove any remaining debris and granulation tissue (Fig. ). Once the site was prepared, allogeneic grafting material was rehydrated according to the manufacturer’s instructions with sterile saline and introduced into the socket to fill the void left by the extracted tooth (Fig. ). The bone substitute was softly condensed. After placing the grafting material, A-PRF was positioned over the allograft and fixed with resorbable sutures (Fig. ). Post-operative care instructions, including oral hygiene measures and dietary restrictions, were provided to the patient to ensure optimal healing and reduce the risk of complications. Routine post-operative care included administration of amoxicillin and clavulanic acid (625 mg, administered orally, three times a day for 4 days), ibuprofen (600 mg, administered orally, every 6 h as needed), and mouthwash (0.2% chlorhexidine, three times daily for 7 days). . After four months of healing time, CBCT scans were done to investigate the osseous healing of the socket, and one titanium implant per augmented region was inserted (Fig. ).
Twelve months post-implantation, patients were recalled for a follow-up evaluation to assess implant success. The assessment was conducted using the guidelines from the International Congress of Oral Implantologists (ICOI) Pisa Consensus Conference on Implant Success, Survival, and Failure. Key parameters evaluated included pain, implant mobility, bleeding on probing, and radiographic bone loss relative to the initial bone level. These parameters were used to categorize the implants according to the ICOI implant quality scale .
Every patient was subjected to three-dimensional x-ray diagnostics (CBCT). In total three CBCTs were recorded for each patient, one before treatment, one directly after ridge preservation, and one after four months of healing before implantation. At each time point, the alveolar bone levels were measured in their height, width and depth at the cervical level, the middle height of the defect and at the apical level. An illustration of the measured regions is shown in Fig. . The CBCT machine used to acquire all images was the Carestream CS 9300 (Carestream Health Inc., Onex Corporation, Rochester, New York, USA). The imaging parameters were set with a dose of 120 mGy.cm2, tube current of 18.66 mAs, and a voxel size of 90 μm x 90 μm x 90 μm. The selected field of view was 5 cm x 5 cm. Data from the scans were saved in the Digital Imaging and Communications in Medicine (DICOM) format and reconstructed with the Carestream implant planning software program. All measurements were made on parasagittal sections perpendicular to the longitudinal axis of the adjacent teeth. The CBCT was oriented transversally through the teeth neighboring the defect so that the nerve canal of the tooth, which was mesial to the defect region, was visible. The nerve canal of the mesial tooth was defined as an anatomic reproducible landmark and a straight line was drawn through the middle of the defect region between the two neighboring teeth. The mesial tooth was used as a reference for apical and crestal bone levels. The distances were obtained using a software ruler. The same anatomic landmarks and distances were used for measurements on CBCT at the defined time intervals. The following measurements were taken (Fig. ): Defect height (mm): distance between the apical and crestal bone level in the middle of the defect region; represented by line “h” in Fig. . Apical defect width (mm): distance between the apical root tips of the neighboring teeth; represented by line “a” in Fig. . Defect width in the middle zone (mm): distance between the roots of the neighboring teeth in the middle of the defect height; represented by line “b” in Fig. . Cervical defect width (mm): distance between the crestal bone levels of the neighboring teeth; represented by line “c” in Fig. . Apical defect depth (mm): distance between the labial/buccal and palatal edges of the jaw crest at the level of the apical tips of the neighboring teeth, but in the middle of the defect area; represented by line “d” in Fig. . Defect depth in the middle zone (mm): distances between the labial/buccal and palatal edges of the jaw crest at the level of the middle zone; represented by lines “e” (mesial region), “f” (central region) and “g” (distal region) in Fig. . Cervical defect depth (mm): distance between the labial/buccal and palatal edges of the jaw crest at the cervical level; represented by lines “k” (mesial region), “l” (central region) and “m” (distal region) in Fig. .
In order to estimate the bone density of the allograft, Hounsfield Units (HU) were used as a measurement of radiodensity. The integrated measurement module of a picture archiving and communication system software was used (PACS, DeepUnity Diagnost 1.1.1.1, DEDALUS). The regions of interest were defined in the graft’s crestal, middle and apical portions and adjusted to their size (5–20 mm 2 ) at each level. In all cases, CBCT sections with low scattered radiation were chosen. The maximum area within the allograft was selected in such a way that the bony marginal structures of the alveolar process were not touched. Both the Hounsfield Units and the area of the measured region were recorded. For the purpose of determining the average bone density within a designated region of the CBCT scan, total HUs were divided by the area of the region measured. This provided the average radiodensity within that specific area.
Based on the radiographic measurements, the graft volume was inferred as the sum of the volumes of two superimposed frustums of pyramids. The formula for obtaining the volume of one pyramid trunk is: [12pt]{minimal}
$$\:{V}_{pyramid\:trunk}=\:(B+P+)\:$$ , where “g ” is the height of the truncated pyramid, “B” is the base area and “P” is the peak area. The two pyramid trunks are depicted in Fig. . The formula for obtaining the volume of the entire defect was therefore: [12pt]{minimal}
$$} = {h 6} ( {{A_{crestal}} + {A_{middle}} + {{A_{crestal}} {A_{middle}}} } ) & + {h 6} ( {{A_{middle}} + {A_{apical}} + {{A_{middle}} {A_{apical}}} } ) }$$ In the formula above, “h” denotes the distance between the apical and crestal bone level in the middle of the defect region. “A crestal ” is the area of the cross-sectional surface of the alveolus at the crestal level. This was calculated from the width “c”, and the three horizontal, buccal-lingual depths “k” (mesial area), “l” (central area) and “m” (distal area). “A middle ” is the area of the cross-sectional surface of the alveolus at the middle height of the defect. This was calculated from the width “b”, and the three horizontal, buccal-lingual depths “e” (mesial area), “f” (central area) and “g” (distal area). “A apical ” is the area of the cross-sectional surface of the alveolus at the apical level. This was calculated from the width “a”, and the buccal-lingual depth “d”. Statistical analyses were performed with IBM SPSS (version 27; International Business Machines Corp., Armonk, NY, USA). Pearson’s chi-squared test was applied to sets of unpaired categorical data to evaluate the likelihood that any observed difference between the sets was due to chance. All metric variables were tested for normal distribution (Shapiro-Wilk test) and homogeneity of variance (Levene’s test) before parametric tests. An independent sample t -test was used when two separate sets of independent and identically distributed samples were obtained, and their population means were compared to each other. Multiple linear regression was used to try to explain an observed outcome variable (bone density) by several independent variables. The categorical variables were added to the model as factors. In the first model, all potential predictors were included. Then, all non-significant variables were removed from the multiple linear regression model in a stepwise manner. Only two-sided significance tests were used. A probability of error of p ≤ 0.05 was chosen as the threshold value. An alpha adjustment for multiple testing was not performed. The results are therefore explorative and descriptive.
Demographics The demographic characteristics of the study population are summarized in Table . Gender was distributed evenly between the two study groups. There was no significant difference in the distribution of treated loci between the two study groups. The average age of the patients was 51.7 ± 12.2 years. The average healing time between ridge preservation and implantation was 4.1 ± 0.3 months. Implant survival and success All patients were monitored for 12 months post-implantation. Throughout the healing period following ridge preservation, there were no indications of infection, wound dehiscence, graft exposure, or other postoperative complications. Allogeneic bone grafts were successfully integrated into the recipient sites by the time of implant placement. The grafted bone remained stable during drilling and implant placement in all patients, allowing for successful stabilization and restoration of all implants three months after placement. Each patient received a fixed implant-supported crown. One year after implantation, no patients reported pain at the implant sites, and none of the 40 implants exhibited signs of mobility. Bleeding on probing was observed in 4 patients. Radiographic analysis revealed that the majority of implants (23 out of 40) showed no bone loss. Nearly all implants (39 out of 40) were classified as “Success” (group I) according to the ICOI scheme. No statistically significant differences were observed in implant quality or success criteria between the two study groups (Table ). Vertical gain and graft stability The remaining height of the alveolar bone before tooth extraction averaged 9.7 ± 2.5 mm (Table ). After extraction and immediately following ridge preservation, the average height of the alveolar process at the extraction site was 10.1 ± 2.3 mm and did not differ between the two study groups. After four months of healing and therefore immediately before implant placement, the vertical height was the same in the two study arms. However, the vertical height loss after 4 months was significantly more pronounced in the Allo group than in the AlloHya group ( p = 0.011). Horizontal gain and graft stability To calculate the horizontal gain and loss rates, the cross-sectional areas were determined both at the crestal bone level (Table ) and the mean height of the alveolus (Table ). Before tooth extraction, the average cross-sectional area at the crestal bone level was 33.7 ± 19.2 mm 2 . Immediately after extraction and ridge preservation, the cross-sectional area at the crestal bone level amounted to 51.0 ± 24.3 mm 2 , and after four months the average values reached 44.5 ± 23.9 mm 2 . N statistically significant differences existed between Allo and AlloHya augmentations (Table ). Horizontal bone loss at crestal bone level was --6.5 ± 4.6 mm 2 and did not differ between the two groups (Table ). At the mid-height of the socket, the average cross-sectional area before tooth extraction was 50.2 ± 36.6 mm 2 , immediately after ridge preservation the average cross-sectional area measured 53.3 ± 35.7 mm 2 and after four months the average value came to 50.2 ± 36.6 mm 2 . Again, there was no difference between the two study groups (Table ). The horizontal bone loss at the middle defect height was − 3.1 ± 5.2 mm 2 and did not differ between Allo and AlloHya augmentations (Table ). Remodeling The three-dimensional volume of the defect area, i.e., the volume of the socket plus its bony margin, was 403.4 ± 321.2 mm 3 before tooth extraction, 510.9 ± 363.8 mm 3 immediately after extraction and subsequent ridge preservation and 449.9 ± 350.6 mm 3 after 4 months of healing. There were no statistically significant differences in the volumes between Allo and AlloHya augmentations (Table ). However, the augmentation volume shrank by an average of -80.7 ± 55.2 mm 3 in ridge preservation without hyaluronic acid, while the volume of the augmentation with hyaluronic acid only decreased by an average of only − 43.2 ± 39.2 mm 3 ( p = 0.017). Therefore, the graft shrinkage rate was 16.9 ± 11.5% in ridge preservations without hyaluronic acid, while the graft shrinkage rate in ridge preservations with hyaluronic acid was only 10.3 ± 7.7% ( p = 0.038). Bone density Immediately after ridge preservation, the average bone density was 159.4 ± 66.1, and four months after ridge preservation, the average bone density was 176.5 ± 70.9. Bone density increased for both Allo and AlloHya augmentations over the four-month healing period. Immediately after ridge preservation, the average bone density was 121.31 ± 49.08 for augmentations without hyaluronic acid and 189.38 ± 64.38 for augmentations with hyaluronic acid ( p < 0.01; Table ). After a four-month healing period, the average bone density was 132.66 ± 48.85 for augmentations without hyaluronic acid and 211.03 ± 67.35 for augmentations with hyaluronic acid ( p < 0.01; Table ).
The demographic characteristics of the study population are summarized in Table . Gender was distributed evenly between the two study groups. There was no significant difference in the distribution of treated loci between the two study groups. The average age of the patients was 51.7 ± 12.2 years. The average healing time between ridge preservation and implantation was 4.1 ± 0.3 months.
All patients were monitored for 12 months post-implantation. Throughout the healing period following ridge preservation, there were no indications of infection, wound dehiscence, graft exposure, or other postoperative complications. Allogeneic bone grafts were successfully integrated into the recipient sites by the time of implant placement. The grafted bone remained stable during drilling and implant placement in all patients, allowing for successful stabilization and restoration of all implants three months after placement. Each patient received a fixed implant-supported crown. One year after implantation, no patients reported pain at the implant sites, and none of the 40 implants exhibited signs of mobility. Bleeding on probing was observed in 4 patients. Radiographic analysis revealed that the majority of implants (23 out of 40) showed no bone loss. Nearly all implants (39 out of 40) were classified as “Success” (group I) according to the ICOI scheme. No statistically significant differences were observed in implant quality or success criteria between the two study groups (Table ).
The remaining height of the alveolar bone before tooth extraction averaged 9.7 ± 2.5 mm (Table ). After extraction and immediately following ridge preservation, the average height of the alveolar process at the extraction site was 10.1 ± 2.3 mm and did not differ between the two study groups. After four months of healing and therefore immediately before implant placement, the vertical height was the same in the two study arms. However, the vertical height loss after 4 months was significantly more pronounced in the Allo group than in the AlloHya group ( p = 0.011).
To calculate the horizontal gain and loss rates, the cross-sectional areas were determined both at the crestal bone level (Table ) and the mean height of the alveolus (Table ). Before tooth extraction, the average cross-sectional area at the crestal bone level was 33.7 ± 19.2 mm 2 . Immediately after extraction and ridge preservation, the cross-sectional area at the crestal bone level amounted to 51.0 ± 24.3 mm 2 , and after four months the average values reached 44.5 ± 23.9 mm 2 . N statistically significant differences existed between Allo and AlloHya augmentations (Table ). Horizontal bone loss at crestal bone level was --6.5 ± 4.6 mm 2 and did not differ between the two groups (Table ). At the mid-height of the socket, the average cross-sectional area before tooth extraction was 50.2 ± 36.6 mm 2 , immediately after ridge preservation the average cross-sectional area measured 53.3 ± 35.7 mm 2 and after four months the average value came to 50.2 ± 36.6 mm 2 . Again, there was no difference between the two study groups (Table ). The horizontal bone loss at the middle defect height was − 3.1 ± 5.2 mm 2 and did not differ between Allo and AlloHya augmentations (Table ).
The three-dimensional volume of the defect area, i.e., the volume of the socket plus its bony margin, was 403.4 ± 321.2 mm 3 before tooth extraction, 510.9 ± 363.8 mm 3 immediately after extraction and subsequent ridge preservation and 449.9 ± 350.6 mm 3 after 4 months of healing. There were no statistically significant differences in the volumes between Allo and AlloHya augmentations (Table ). However, the augmentation volume shrank by an average of -80.7 ± 55.2 mm 3 in ridge preservation without hyaluronic acid, while the volume of the augmentation with hyaluronic acid only decreased by an average of only − 43.2 ± 39.2 mm 3 ( p = 0.017). Therefore, the graft shrinkage rate was 16.9 ± 11.5% in ridge preservations without hyaluronic acid, while the graft shrinkage rate in ridge preservations with hyaluronic acid was only 10.3 ± 7.7% ( p = 0.038).
Immediately after ridge preservation, the average bone density was 159.4 ± 66.1, and four months after ridge preservation, the average bone density was 176.5 ± 70.9. Bone density increased for both Allo and AlloHya augmentations over the four-month healing period. Immediately after ridge preservation, the average bone density was 121.31 ± 49.08 for augmentations without hyaluronic acid and 189.38 ± 64.38 for augmentations with hyaluronic acid ( p < 0.01; Table ). After a four-month healing period, the average bone density was 132.66 ± 48.85 for augmentations without hyaluronic acid and 211.03 ± 67.35 for augmentations with hyaluronic acid ( p < 0.01; Table ).
Clinical relevance of and potential mechanisms behind the observations The results of our study indicated that the addition of hyaluronic acid to allogeneic bone grafting material significantly improved outcomes in the preservation of compromised extraction sockets. Specifically, we observed enhanced graft stability, reduced graft resorption and increased bone density. Enhanced Graft Stability While there were no differences in the horizontal graft stability between allogeneic ridge preservations with and without hyaluronic acid, the vertical height loss after 4 months was significantly more pronounced in the Allo group (-0.82 mm) than in the AlloHya group (-0.19 mm). Hyaluronic acid is well-known for its ability to enhance tissue regeneration and wound healing . Studies have suggested that hyaluronic acid promotes cellular adhesion and proliferation, which could contribute to the formation of a robust matrix within the graft site . In an animal model, cross-linked hyaluronic acid significantly enhanced periodontal wound healing and regeneration in two-wall mandibular intrabony defects . In the context of ridge preservation, the incorporation of hyaluronic acid might have facilitated better integration of the graft material with the surrounding tissues, leading to improved stability. Reduced shrinkage rate There were no statistically significant differences in the volumes between allogeneic ridge preservations with and without hyaluronic acid. However, the graft shrinkage rate was 16.9% in allogeneic ridge preservations without hyaluronic acid, while the graft shrinkage rate in allogeneic ridge preservations with hyaluronic acid was only 10.3%. Hyaluronic acid possesses unique viscoelastic properties, which could have mitigated the shrinkage of the graft material over time. By maintaining hydration levels and supporting the structural integrity of the graft, hyaluronic acid might have minimized the volume loss typically associated with bone graft resorption. Indeed, ridge preservation with a mixture of a bovine graft material and hyaluronic acid in an animal model prevented dimensional shrinkage and improved bone formation in compromised extraction sockets . Furthermore, hyaluronic acid has been shown to modulate inflammatory responses and promote angiogenesis , which could indirectly influence graft remodeling and reduce shrinkage. In our study, the graft shrinkage after 4 months amounted for ~ 17% in the Allo group and for ~ 11% in the AlloHya group. The reduced shrinkage rate linked to the addition of hyaluronic acid to allogeneic bone material was also observed when combining hyaluronic acid with xenogeneic bone material for alveolar ridge preservation . Therefore, hyaluronic acid appears to limit the post-extractional alveolar bone resorption when either mixed with allogeneic or xenogeneic bone material. Increased bone density After a four-month healing period, the average bone density was 132.66 for allogeneic ridge preservations without hyaluronic acid and 211.03 for allogeneic ridge preservations with hyaluronic acid. Hyaluronic acid is involved in various signaling pathways that regulate osteogenesis and bone remodeling . By interacting with cell surface receptors such as CD44 and RHAMM , hyaluronic acid can stimulate osteoblast activity and mineralization processes . Additionally, hyaluronic acid has been shown to inhibit osteoclastogenesis and bone resorption, thereby preserving bone density . The addition of hyaluronic matrix to xenograft in maxillary sinus augmentation significantly increased bone surface density, suggesting enhanced bone quality . The combination of hyaluronic acid with the allogeneic bone grafting material might have synergistically enhanced these osteogenic effects, resulting in higher bone density at the graft site. Differences in baseline density might be explained by the higher viscosity of hyaluronic acid in comparison to saline. However, considering the rapid turnover of hyaluronic acid in situ, this effect is unlikely to contribute to the differences after four months, in contrast even increasing the differences between grafting and four months later. Potential synergistic effects maxgraft ® is a widely used allogeneic bone grafting material known for its biocompatibility and osteoconductive properties . The addition of hyaluronic acid could have complemented these characteristics by providing a biological component that promotes tissue regeneration and modulates the local microenvironment. Studies have suggested that hyaluronic acid can interact synergistically with other biomaterials, enhancing their therapeutic efficacy in various clinical applications . This was also demonstrated for platelet rich fibrin (PRF), which was found to be beneficial for ridge preservation surgeries, especially when combined with other graft materials . Clinical evidence supporting hyaluronic acid supplementation Previous studies investigating the use of hyaluronic acid in bone regeneration and dental procedures have reported favorable outcomes, supporting its efficacy as a therapeutic adjunct . A recent randomized clinical trial showed that adding hyaluronic acid to the coronally advanced flap procedure significantly improved complete root coverage and reduced post-operative swelling and discomfort in the treatment of gingival recessions . Furthermore, adding cross-linked hyaluronic acid to demineralized bovine bone mineral during guided bone regeneration significantly improved bone quality and quantity compared to using bovine bone material alone . Topical application of hyaluronic acid as an adjunctive treatment improved clinical outcomes in both non-surgical and surgical periodontal therapies . In summary, the use of allogeneic bone substitutes combined with hyaluronic acid in ridge preservation offers advantages such as hydrophilic properties, enhanced cell attachment, reduced inflammation, periodontal regeneration, scaffold functionality, improved bone regeneration, bacteriostatic properties and scarless wound healing. By that, without changing the individual treatment protocol, an improved patient outcome can be achieved. Strength and limitations Our study on ridge preservation boasts several significant strengths, including a robust sample size of 20 patients per group. The statistical power of 100% guarantees our ability to detect even subtle differences between the groups, adding to the reliability of our findings. Additionally, the introduction of a new biomaterial that has not been published before represents a novel contribution to the field. The homogeneity of our patient group further enhances the validity of our results, minimizing variability and potential confounding factors. However, our study is not without limitations. The method used to calculate bone density relies on grey values, which may not provide the most precise measurement. Moreover, the observation period of 12 months post-implantation may not capture the long-term outcomes and stability of the ridge preservation techniques evaluated. Evaluating bone mineral density (BMD) at implant placement sites is crucial for ensuring sufficient primary stability. Computed tomography (CT) is widely acknowledged as the standard method for BMD assessment due to its consistent display of Hounsfield units (HUs). However, CT’s high radiation dosage restricts its use in dental diagnoses. A recent systematic review examined the relationship between cone-beam computed tomography (CBCT) gray values (GVs) and CT’s HUs in assessing BMD . Converting CBCT’s linear attenuation coefficients into HUs requires applying a prediction equation model or conversion ratio to GVs. Despite limitations, both qualitative and quantitative analyses in the review revealed a positive correlation between CBCT’s GVs and CT’s HUs. Therefore, CBCT’s GVs can be utilized to quantitatively estimate bone density prior to implant-related procedures, supported by evidence indicating a positive correlation between CBCT’s GVs and CT’s HUs . Although a positive influence on initial healing pattern caused by topical use of hyaluronic acid has been published , those parameters have not been evaluated in the present study but might be subject to further research. The observation period of 12 months post-implantation can be considered less problematic given that all augmentation sites remained free of inflammation and dehiscence, and all implants demonstrated stability without any signs of peri-implantitis, bleeding on probing, or other complications.
The results of our study indicated that the addition of hyaluronic acid to allogeneic bone grafting material significantly improved outcomes in the preservation of compromised extraction sockets. Specifically, we observed enhanced graft stability, reduced graft resorption and increased bone density. Enhanced Graft Stability While there were no differences in the horizontal graft stability between allogeneic ridge preservations with and without hyaluronic acid, the vertical height loss after 4 months was significantly more pronounced in the Allo group (-0.82 mm) than in the AlloHya group (-0.19 mm). Hyaluronic acid is well-known for its ability to enhance tissue regeneration and wound healing . Studies have suggested that hyaluronic acid promotes cellular adhesion and proliferation, which could contribute to the formation of a robust matrix within the graft site . In an animal model, cross-linked hyaluronic acid significantly enhanced periodontal wound healing and regeneration in two-wall mandibular intrabony defects . In the context of ridge preservation, the incorporation of hyaluronic acid might have facilitated better integration of the graft material with the surrounding tissues, leading to improved stability. Reduced shrinkage rate There were no statistically significant differences in the volumes between allogeneic ridge preservations with and without hyaluronic acid. However, the graft shrinkage rate was 16.9% in allogeneic ridge preservations without hyaluronic acid, while the graft shrinkage rate in allogeneic ridge preservations with hyaluronic acid was only 10.3%. Hyaluronic acid possesses unique viscoelastic properties, which could have mitigated the shrinkage of the graft material over time. By maintaining hydration levels and supporting the structural integrity of the graft, hyaluronic acid might have minimized the volume loss typically associated with bone graft resorption. Indeed, ridge preservation with a mixture of a bovine graft material and hyaluronic acid in an animal model prevented dimensional shrinkage and improved bone formation in compromised extraction sockets . Furthermore, hyaluronic acid has been shown to modulate inflammatory responses and promote angiogenesis , which could indirectly influence graft remodeling and reduce shrinkage. In our study, the graft shrinkage after 4 months amounted for ~ 17% in the Allo group and for ~ 11% in the AlloHya group. The reduced shrinkage rate linked to the addition of hyaluronic acid to allogeneic bone material was also observed when combining hyaluronic acid with xenogeneic bone material for alveolar ridge preservation . Therefore, hyaluronic acid appears to limit the post-extractional alveolar bone resorption when either mixed with allogeneic or xenogeneic bone material. Increased bone density After a four-month healing period, the average bone density was 132.66 for allogeneic ridge preservations without hyaluronic acid and 211.03 for allogeneic ridge preservations with hyaluronic acid. Hyaluronic acid is involved in various signaling pathways that regulate osteogenesis and bone remodeling . By interacting with cell surface receptors such as CD44 and RHAMM , hyaluronic acid can stimulate osteoblast activity and mineralization processes . Additionally, hyaluronic acid has been shown to inhibit osteoclastogenesis and bone resorption, thereby preserving bone density . The addition of hyaluronic matrix to xenograft in maxillary sinus augmentation significantly increased bone surface density, suggesting enhanced bone quality . The combination of hyaluronic acid with the allogeneic bone grafting material might have synergistically enhanced these osteogenic effects, resulting in higher bone density at the graft site. Differences in baseline density might be explained by the higher viscosity of hyaluronic acid in comparison to saline. However, considering the rapid turnover of hyaluronic acid in situ, this effect is unlikely to contribute to the differences after four months, in contrast even increasing the differences between grafting and four months later. Potential synergistic effects maxgraft ® is a widely used allogeneic bone grafting material known for its biocompatibility and osteoconductive properties . The addition of hyaluronic acid could have complemented these characteristics by providing a biological component that promotes tissue regeneration and modulates the local microenvironment. Studies have suggested that hyaluronic acid can interact synergistically with other biomaterials, enhancing their therapeutic efficacy in various clinical applications . This was also demonstrated for platelet rich fibrin (PRF), which was found to be beneficial for ridge preservation surgeries, especially when combined with other graft materials . Clinical evidence supporting hyaluronic acid supplementation Previous studies investigating the use of hyaluronic acid in bone regeneration and dental procedures have reported favorable outcomes, supporting its efficacy as a therapeutic adjunct . A recent randomized clinical trial showed that adding hyaluronic acid to the coronally advanced flap procedure significantly improved complete root coverage and reduced post-operative swelling and discomfort in the treatment of gingival recessions . Furthermore, adding cross-linked hyaluronic acid to demineralized bovine bone mineral during guided bone regeneration significantly improved bone quality and quantity compared to using bovine bone material alone . Topical application of hyaluronic acid as an adjunctive treatment improved clinical outcomes in both non-surgical and surgical periodontal therapies . In summary, the use of allogeneic bone substitutes combined with hyaluronic acid in ridge preservation offers advantages such as hydrophilic properties, enhanced cell attachment, reduced inflammation, periodontal regeneration, scaffold functionality, improved bone regeneration, bacteriostatic properties and scarless wound healing. By that, without changing the individual treatment protocol, an improved patient outcome can be achieved.
While there were no differences in the horizontal graft stability between allogeneic ridge preservations with and without hyaluronic acid, the vertical height loss after 4 months was significantly more pronounced in the Allo group (-0.82 mm) than in the AlloHya group (-0.19 mm). Hyaluronic acid is well-known for its ability to enhance tissue regeneration and wound healing . Studies have suggested that hyaluronic acid promotes cellular adhesion and proliferation, which could contribute to the formation of a robust matrix within the graft site . In an animal model, cross-linked hyaluronic acid significantly enhanced periodontal wound healing and regeneration in two-wall mandibular intrabony defects . In the context of ridge preservation, the incorporation of hyaluronic acid might have facilitated better integration of the graft material with the surrounding tissues, leading to improved stability.
There were no statistically significant differences in the volumes between allogeneic ridge preservations with and without hyaluronic acid. However, the graft shrinkage rate was 16.9% in allogeneic ridge preservations without hyaluronic acid, while the graft shrinkage rate in allogeneic ridge preservations with hyaluronic acid was only 10.3%. Hyaluronic acid possesses unique viscoelastic properties, which could have mitigated the shrinkage of the graft material over time. By maintaining hydration levels and supporting the structural integrity of the graft, hyaluronic acid might have minimized the volume loss typically associated with bone graft resorption. Indeed, ridge preservation with a mixture of a bovine graft material and hyaluronic acid in an animal model prevented dimensional shrinkage and improved bone formation in compromised extraction sockets . Furthermore, hyaluronic acid has been shown to modulate inflammatory responses and promote angiogenesis , which could indirectly influence graft remodeling and reduce shrinkage. In our study, the graft shrinkage after 4 months amounted for ~ 17% in the Allo group and for ~ 11% in the AlloHya group. The reduced shrinkage rate linked to the addition of hyaluronic acid to allogeneic bone material was also observed when combining hyaluronic acid with xenogeneic bone material for alveolar ridge preservation . Therefore, hyaluronic acid appears to limit the post-extractional alveolar bone resorption when either mixed with allogeneic or xenogeneic bone material.
After a four-month healing period, the average bone density was 132.66 for allogeneic ridge preservations without hyaluronic acid and 211.03 for allogeneic ridge preservations with hyaluronic acid. Hyaluronic acid is involved in various signaling pathways that regulate osteogenesis and bone remodeling . By interacting with cell surface receptors such as CD44 and RHAMM , hyaluronic acid can stimulate osteoblast activity and mineralization processes . Additionally, hyaluronic acid has been shown to inhibit osteoclastogenesis and bone resorption, thereby preserving bone density . The addition of hyaluronic matrix to xenograft in maxillary sinus augmentation significantly increased bone surface density, suggesting enhanced bone quality . The combination of hyaluronic acid with the allogeneic bone grafting material might have synergistically enhanced these osteogenic effects, resulting in higher bone density at the graft site. Differences in baseline density might be explained by the higher viscosity of hyaluronic acid in comparison to saline. However, considering the rapid turnover of hyaluronic acid in situ, this effect is unlikely to contribute to the differences after four months, in contrast even increasing the differences between grafting and four months later.
maxgraft ® is a widely used allogeneic bone grafting material known for its biocompatibility and osteoconductive properties . The addition of hyaluronic acid could have complemented these characteristics by providing a biological component that promotes tissue regeneration and modulates the local microenvironment. Studies have suggested that hyaluronic acid can interact synergistically with other biomaterials, enhancing their therapeutic efficacy in various clinical applications . This was also demonstrated for platelet rich fibrin (PRF), which was found to be beneficial for ridge preservation surgeries, especially when combined with other graft materials .
Previous studies investigating the use of hyaluronic acid in bone regeneration and dental procedures have reported favorable outcomes, supporting its efficacy as a therapeutic adjunct . A recent randomized clinical trial showed that adding hyaluronic acid to the coronally advanced flap procedure significantly improved complete root coverage and reduced post-operative swelling and discomfort in the treatment of gingival recessions . Furthermore, adding cross-linked hyaluronic acid to demineralized bovine bone mineral during guided bone regeneration significantly improved bone quality and quantity compared to using bovine bone material alone . Topical application of hyaluronic acid as an adjunctive treatment improved clinical outcomes in both non-surgical and surgical periodontal therapies . In summary, the use of allogeneic bone substitutes combined with hyaluronic acid in ridge preservation offers advantages such as hydrophilic properties, enhanced cell attachment, reduced inflammation, periodontal regeneration, scaffold functionality, improved bone regeneration, bacteriostatic properties and scarless wound healing. By that, without changing the individual treatment protocol, an improved patient outcome can be achieved.
Our study on ridge preservation boasts several significant strengths, including a robust sample size of 20 patients per group. The statistical power of 100% guarantees our ability to detect even subtle differences between the groups, adding to the reliability of our findings. Additionally, the introduction of a new biomaterial that has not been published before represents a novel contribution to the field. The homogeneity of our patient group further enhances the validity of our results, minimizing variability and potential confounding factors. However, our study is not without limitations. The method used to calculate bone density relies on grey values, which may not provide the most precise measurement. Moreover, the observation period of 12 months post-implantation may not capture the long-term outcomes and stability of the ridge preservation techniques evaluated. Evaluating bone mineral density (BMD) at implant placement sites is crucial for ensuring sufficient primary stability. Computed tomography (CT) is widely acknowledged as the standard method for BMD assessment due to its consistent display of Hounsfield units (HUs). However, CT’s high radiation dosage restricts its use in dental diagnoses. A recent systematic review examined the relationship between cone-beam computed tomography (CBCT) gray values (GVs) and CT’s HUs in assessing BMD . Converting CBCT’s linear attenuation coefficients into HUs requires applying a prediction equation model or conversion ratio to GVs. Despite limitations, both qualitative and quantitative analyses in the review revealed a positive correlation between CBCT’s GVs and CT’s HUs. Therefore, CBCT’s GVs can be utilized to quantitatively estimate bone density prior to implant-related procedures, supported by evidence indicating a positive correlation between CBCT’s GVs and CT’s HUs . Although a positive influence on initial healing pattern caused by topical use of hyaluronic acid has been published , those parameters have not been evaluated in the present study but might be subject to further research. The observation period of 12 months post-implantation can be considered less problematic given that all augmentation sites remained free of inflammation and dehiscence, and all implants demonstrated stability without any signs of peri-implantitis, bleeding on probing, or other complications.
In conclusion, our study demonstrates that adding hyaluronic acid to allogeneic bone substitutes in ridge preservation leads to enhanced graft stability, reduced shrinkage rate, and increased bone density. These findings highlight the potential of hyaluronic acid to optimize ridge preservation procedures and promote successful implant integration.
|
Bibliometric Analysis of Ophthalmology Publications from Arab Countries between 2012 and 2022 | 1b1f5bcc-79f3-496e-bc4f-48a438685a41 | 10903712 | Ophthalmology[mh] | The field of biomedical research has grown tremendously over the last few decades. While a significant upward trend in publication volume has been observed globally, there have been wide variations in research productivity across different regions and countries due to differences in health-care systems, educational programs, and funding support programs. Within the Arab region, political, socioeconomic, and security dynamics have also been found to influence scientific productivity. Based on bibliometric reviews of biomedical articles published between 1988 and 2002 by Tadmouri and Bissar-Tadmouri and between 2001 and 2005 by Benamer and Benamer, the scientific production of Arab nations was found to be significantly lower compared to other countries in the world. In the field of ophthalmology alone, research productivity from Arab league countries was also found to relatively lag behind. Using the time frame 1900–2012, research output in ophthalmology from Arab countries (0.96%) represented <1% of the global research productivity in ophthalmology. While the aforementioned barriers to research activity and scientific publication likely contribute to this research disparity, it is also important to note that the accuracy of prior bibliometric analysis was affected by the lack of inclusion and indexing of many of the journals commonly used for publication by Arab-based authors within the standard databases such as ISI Web of Science, Scopus, and MEDLINE for bibliometric analyses. Given the recent indexing of several major regional journals such as the Middle East African Journal of Ophthalmology in 2020, Saudi Journal of Ophthalmology in 2020, and the Journal of the Egyptian Ophthalmological Society in 2021 within the ISI Web of Science database, we sought to re-examine the status of ophthalmology research and provide a more accurate presentation of the geographic trends of research output and scientific productivity in Arab countries. In this study, we evaluated the research output of authors from Arab-based institutions in the field of ophthalmology from 2012 to 2022. This cross-sectional study involved a bibliometric analysis of all original research and review articles published in Ophthalmology Journals by ophthalmologists, optometrists, and researchers working in vision science with an affiliation with an institution from an Arab League nation between January 1, 2012, and December 31, 2022. As the study did not involve the evaluation or management of human participants, ethics committee review and approval were waived by the local institutional review board. This study abided by the Strengthening the Reporting of Observational Studies in Epidemiology reporting guideline for cross-sectional studies. The data were extracted from the ISI Web of Science database last March 28, 2023. Using the advanced search engine of ISI Web of Science, the search was limited to the “ophthalmology” category tag. Using the countries filter, Arab-based scientific articles were identified by restricting the search to the following countries: Algeria, Bahrain, Comoros, Djibouti, Egypt, Iraq, Jordan, Kuwait, Lebanon, Oman, Libya, Mauritania, Morocco, Palestine, Qatar, Saudi Arabia, Somalia, Sudan, Syria, Tunisia, United Arab Emirates, and Yemen. To evaluate the more recent production in the field of ophthalmology, the period of analysis was restricted to articles published over the last decade between January 1, 2012, and December 31, 2022. The analysis was further limited to documents classified as articles, articles in press, and reviews. Letters, correspondences, and replies were excluded from this analysis. No other exclusion criteria for language or other publication parameters were applied. All collected data were imported into Microsoft Excel (Microsoft Corp., Redmond, WA, USA) for analysis. The number of articles was used as the indicator of quantity for scientific productivity. The countries and institutions were ranked according to the number of articles produced. For the years 2012–2022, 4292 articles published in Ophthalmology Journals by authors from Arab-based institutions were identified. The number of publications by Arab authors in journals indexed in ISI Web of Science increased steadily during the early years of this decade. During the first 2 years of the COVID-19 pandemic between 2020 and 2021, the annual output of research articles nearly quadrupled from 169 in 2012 to 621 in 2020 and 645 in 2021. In 2022, the number of publications slightly decreased to 527, which was still substantially higher than prepandemic levels . Overall, a 2.11-fold increase was observed within this decade. depicts the global distribution of ophthalmology articles published between 2012 and 2022. Both Egypt and Saudi Arabia ranked within the top 25 countries with the highest number of publications worldwide. Within the Arab League , the highest number of ophthalmology articles were published from Egypt ( n = 1653, 38.51%). This was followed by Saudi Arabia ( n = 1526, 32.74%), United Arab Emirates ( n = 338, 7.88%), Lebanon ( n = 299, 6.97%), and Tunisia ( n = 254, 5.92%). According to the institution affiliation within Arab nations , King Khaled Eye Specialist Hospital (KKESH) in Saudi Arabia ranked the highest in terms of scientific productivity with 644 articles, followed by the King Saud University in Saudi Arabia with 585 articles and the Cairo University in Egypt with 393 articles. In terms of language, the majority of the articles produced were in English ( n = 4136, 96.37%) while the rest were in French ( n = 151, 3.52%), German ( n = 3, 0.07%), and Spanish ( n = 2, 0.05%). shows the top 25 peer-reviewed journals that were used for publication by Arab-affiliated ophthalmology researchers. Clinical ophthalmology (8.11%) was the most commonly used, followed by the Saudi Journal of Ophthalmology (5.78%). Analysis of biomedical research and publications in a country or group of countries is an important tool to monitor progress and trends in research and scientific activity. Research productivity can be quantitatively measured in terms of the number of publications in peer-reviewed journals. While a number of bibliometric analyses in the Arab region have been previously published by various authors, recent indexing of major Arab-based Ophthalmology Journals within the ISI World of Science database, as well as the progress toward open research in ophthalmology, has provided the opportunity to comprehensively assess the wider breadth of research and accurately evaluate research productivity within the region. This bibliometric study of publications of research from Arab nations in the field of ophthalmic and vision research shows that research productivity has substantially increased over the last decade. Notably, a sharp spike in publication volume was observed between 2020 and 2021 during the COVID-19 pandemic. This surge in publications fueled by the COVID-19 pandemic was similarly observed across all biomedical fields. Although the total number of publications in 2022 ( n = 645) had decreased compared to 2021 ( n = 527), the annual volume in 2022 is substantially higher than prepandemic levels. Overall, a 2-fold overall increase in research productivity was observed over the last decade. This trend in research productivity follows the exponential growth in publications not only in the field of ophthalmology but in biomedical research in general. Several studies have discussed the factors that have led to the relative paucity of biomedical publications in the Arab region. While the current analysis finds that both Egypt (top 20) and Saudi Arabia (top 21) have now ranked among the top 25 countries worldwide in terms of the number of ophthalmology publications in the last decade , the rest of Arab nations still lag behind in terms of research productivity. In fact, a close review of the relative contributions of different countries in the Arab region to the total number of publications in the field of ophthalmology showed that three-quarters of the total production in the last decade was contributed by authors from only three countries including Egypt (39%), Saudi Arabia (33%), and the United Arab Emirates (8%). Conversely, low-income Arab states such as Comoros, Djibouti, Mauritania, and Somalia produced the least number of publications in ophthalmology research in the studied period. Moreover, countries affected by wars and internal conflicts including Iraq, Libya, Palestine, Somalia, Sudan, Syria, and Yemen have also fared relatively poorly in terms of research output. These findings are fairly consistent with the results of previous studies. While scientific publications are broadly recognized as the primary indicator of research productivity, certain studies have indicated that raw counts should be normalized to indicators such as population size to provide a more accurate presentation of the status within each country. When the number of publications is adjusted to each by the population size in 2022, Lebanon ranked first with 54 ophthalmic publications per million population, followed by Saudi Arabia with 42 publications per million population and the United Arab Emirates with 35 publications per million within the studied time frame. Saudi Arabia, Egypt, Lebanon, and the United Arab Emirates can, therefore, be considered the leading institutions for ophthalmic research in the Arab League. In terms of individual research institutions, seven of the top ten performers were university-based centers while the rest were hospital-based research centers. In contrast to a previous bibliometric analysis of ophthalmic publications, KKESH has currently outperformed other institutions with the highest productivity in ophthalmic research within the Arab region. Established in 1983, KKESH is one of the largest specialty eye hospitals in the world with a dedicated budget to support research-related activities. The current position of KKESH among other university-based research centers reflects how the allocation of research funds to academic settings outside the university setting can further promote ophthalmology research and increase overall scientific productivity. This study should also be viewed in the light of some limitations. First, in an effort to avoid count errors related to entry duplicates, ISI Web of Science was the only database used to identify the publications for analysis. Articles published in journals that have contributed to scientific productivity but were not indexed by the ISI Web of Science at the time of analysis were not considered. Second, while no search restrictions were applied to the type of article authors, the articles were identified under the “ophthalmology” category tag which excluded articles in basic science and general internal medicine journals. While this may have resulted in an underestimation of total output, the 116 Ophthalmology Journals included in this analysis represent the most important journals in the field of ophthalmology within the Arab region and internationally. Over the last decade, the overall productivity of research in the field of ophthalmology has significantly increased. The majority of the articles were published by authors from Egypt and Saudi Arabia with KKESH as the most prolific institution among Arab nations within the studied time frame. Financial support and sponsorship Nil. Conflicts of interest There are no conflicts of interest. Nil. There are no conflicts of interest. |
Boil water notices as health-risk communication: risk perceptions, efficacy, and compliance during winter storm Uri | 952d3739-99ef-4aa4-bb63-3a3e27859d74 | 10774435 | Health Communication[mh] | Winter Storm Uri occurred February 12–16, 2021. Uri was a coast-to-coast storm system that produced record-amounts of snow and damaging ice and caused many other weather-related issues . In several regions across Texas, Uri dropped 4–6 inches of snow. The storm system produced the coldest temperatures on record for most Texas cities. Nearby Oklahoma was also severely impacted. Oklahoma City reported 6 inches of snow and drifts of 2–4 feet at the Will Rogers Airport , . In addition to snow, ice, and wind, Uri also impacted power grids and water systems. Uri caused wide-spread power outages for over 9.7 million people in the United States and Mexico . Due to power failures, water treatment plants across Texas failed to keep water moving, pipes froze, and there were many pipe leaks , . Nearly 15 million Texans experienced disruptions to their primary source of potable water , . Four days after the beginning of Winter Storm Uri, February 16, 2021, 1.4 million Texans still lacked reliable drinking water service. More than 200,000 Texans were still without water on February 25 as snow, ice, and freezing temperatures persisted. Due to the potential of contaminated drinking water, boil water notices (BWNs) were issued across Texas and Oklahoma. According to data obtained from the Texas Commission of Environmental Quality , there were 1,105 BWNs that impacted about 14.5 million Texans from February 15–19, 2021. Overall, almost 40% (1985) of public and community water systems in Texas had a BWN during Winter Storm Uri . However, it is likely that many people under a BWN did not receive timely notification about water contamination risks. BWNs as risk communication BWNs, like those issued during Uri, are risk messages , . However, crisis communication generally warns individuals about a current threat, risk messages warn individuals about a potential threat that could occur in the future – . Since BWNs aim to inform affected populations about potential water contamination and advise them to take protective actions, they can be classified as risk messages. The Extended Parallel Process Model (EPPM) suggests efficacy plays a key role in determining whether individuals enact protective behaviors . EPPM asserts the effectiveness of crisis and risk communication depends, in part, on conveying efficacy (response + self) information and informing publics about risks in a way that incites action versus fear control , . Further, at-risk individuals tend to view ‘perceived effectiveness in protecting health’ during a water contamination emergency as “the most important correlate of protective action,” which implies that risk communicators must clearly explain how and why a protective action will protect an individual’s health (p. 887) . However, it is not clear if this communication strategy also applies in a novel disaster context that has many cascading risks, like Winter Storm Uri. Cascading risks complicate communication during disasters. Cascading risks are largely associated with “the anthropogenic domain and the vulnerability component of risk. This results in a disaster escalation process. In other words, it focuses mainly on the management of social and infrastructure nodes” (p. 2253) . Simply, cascading risk(s) occur when one hazardous event (e.g., Winter Storm Uri) triggers other events/risks, which produce even more severe consequences (e.g., BWNs, infrastructure failure). Thus, effective communication is key to protecting health and safety when a disaster has cascading risks. More particularly, clear communication of BWNs can enhance individuals’ efficacy levels by persuading them to properly boil their tap water, or use another suitable water source (e.g., bottled water) to mitigate contamination risks, especially during a disaster with cascading risks like Winter Storm Uri , – . To create effective risk messages, communicators must understand risk perception and how it influences actions of at-risk individuals. Simply, risk perception is “the belief that one is vulnerable” to some future risk such as disease, floods, etc. . While risk perception can be conceptualized in many ways, Rimal and Real view risk perception as a combination of a person’s perceived susceptibility and perceived severity of a risk and these are assessed based on threats that may occur in the future . While risk perception alone is typically not enough to predict behavioral intentions, it does help indicate the communicative needs of at-risk individuals and it can help guide message development , , – . Risk perception is an integral element for risk communication as it helps communicators align their message(s) with audience (i.e., at-risk individuals) concerns and needs , , . As may be evident, there is a gap in research related to investigating these phenomena, specifically, as related to BWNs and risk communication. As such, investigating risk perception in the current study is important as it has important implications for risk communication in the context of disasters with cascading risks, like BWNs resulting from Winter Storm Uri. Winter Storm Uri was a disaster with various cascading risks, with one such risk being possible water contamination. The lack of electricity due to the winter storm made communicating with at-risk individuals challenging. Previous research about BWNs notes that “the quick response of utility managers is important to protect more consumers, and using the news media as the only means to protect consumers may not provide high levels of public health protection,” (p. 2051) . Not only does the context of Uri exemplify an extremely complex disaster, but it also highlights the communication difficulties that utility managers, city and state governments, and at-risk individuals faced when ‘normal’ communication channels—such as news media and social media—were not fully available or able to reach all at-risk populations due to limitations related to travel, electrical and cellphone service, etc. – . As such, the goal of this study was to investigate BWN-affected individuals’ risk perceptions, water quality perceptions, and perceived efficacy as related to compliance with BWNs during a complex disaster with cascading risks: Winter Storm Uri. Additionally, the aforementioned phenomena can reveal communicative exigencies of affected individuals as well as help inform recommendations for future risk communication about BWNs that emerge from larger disasters. Current study There is limited risk communication research focused on drinking water issues, especially during severe weather events, which often involve cascading risk(s) , . Despite awareness of an active BWN, compliance with suggested protective actions varies among affected populations. BWN compliance rates are reported to range from 36 to 98% . Previous experience with a risk may increase risk perception and encourage positive protective actions , , – . Further, “Past experience with a hazard is generally thought to influence one’s recognition that a risk exists and increases motivation to protect one-self,” (p. 1841) . As such, the following hypothesis was employed to investigate the relationship between past experience with a BWN and risk perception during Winter Storm Uri: H1 Previous experience with a BWN(s) will alter individuals' perceptions of water quality and the risk of acquiring a waterborne disease during Uri. Due to the complex nature of Winter Storm Uri and its cascading risks, communicating with affected individuals was difficult. For instance, one study reported that more than half of Uri-affected respondents were unsure or confused about whether they were under a BWN during Uri , highlighting a severe risk communication issue. This finding implies serious communication gaps between officials/utility companies and the public. Perhaps even more concerning is that extant research denotes that racial minority communities (e.g., Black and Latinx communities) and lower-income household tend to have less access to clean, safe water . More specifically, during Uri, racial minorities and lower-income households experienced more severe issues, like burst pipes . On the other hand, communities with larger populations of non-Hispanic White residents, single family homes, and higher-income households experienced a smaller percentage of power outages after the storm , meaning they likely had access to electricity, communication technologies, and running water sooner than other communities with higher populations of racial minorities and lower-income households. This information further complicates risk communication surrounding Uri, especially related to BWNs, but also highlights the need for better risk communication that is specifically tailored to diverse populations’ communicative needs. Another factor to consider when crafting risk communication for BWNs is increasing compliance among various at-risk populations. Common reasons for non-compliance during a BWN include forgetfulness, perceived inconvenience, appearance of clean water, not believing the initial notification of a BWN, and failing to receive communication about an issued BWN , , , . Furthermore, low risk perception is consistently reported as a primary reason for non-compliance – . However, Americans have consistently expressed concerns about contamination of drinking water with 83% reporting a “great deal” or “fair” amount of worry in 2022 . Perceptions about the quality and safety of water may interact with risk perceptions and efficacy beliefs to impact decisions about protective actions , , . Both self- and response-efficacy are important in relation to compliance with recommended protective actions during disasters. Efficacy plays a role in individuals’ perceived risk , which is important to know when creating risk communication. However, there were vast communication issues surrounding Winter Storm Uri. The interactive role of risk perception and efficacy , , and compliance in the context of BWN’s following winter storms is not well understood. Understanding these phenomena could contribute to more effective risk communication practices during complex disasters, such as Uri. Therefore, EPPM and extant research suggesting that risk perception and efficacy beliefs play a role in people’s decisions to take protective actions (or not) when experiencing potential water contamination, inform the following hypothesis: H2 Higher levels of perceived efficacy will be related to individuals’ (risk) perceptions of water quality and the risk of acquiring a waterborne disease during Uri. H3 General risk perception(s), efficacy, and (risk) perception of water quality will influence compliance with BWNs during Uri. A more detailed discussion of literature on EPPM, risk and crisis communication, and methods are provided in Supplemental Information . To address gaps in understanding how risk perceptions influence response to BWNs surrounding a weather-related crisis, the 2021 Winter Storm Uri, and to evaluate our hypotheses, we conducted a survey of Texas and Oklahoma residents aimed at understanding how individuals perceive risk surrounding BWNs associated with Uri. BWNs, like those issued during Uri, are risk messages , . However, crisis communication generally warns individuals about a current threat, risk messages warn individuals about a potential threat that could occur in the future – . Since BWNs aim to inform affected populations about potential water contamination and advise them to take protective actions, they can be classified as risk messages. The Extended Parallel Process Model (EPPM) suggests efficacy plays a key role in determining whether individuals enact protective behaviors . EPPM asserts the effectiveness of crisis and risk communication depends, in part, on conveying efficacy (response + self) information and informing publics about risks in a way that incites action versus fear control , . Further, at-risk individuals tend to view ‘perceived effectiveness in protecting health’ during a water contamination emergency as “the most important correlate of protective action,” which implies that risk communicators must clearly explain how and why a protective action will protect an individual’s health (p. 887) . However, it is not clear if this communication strategy also applies in a novel disaster context that has many cascading risks, like Winter Storm Uri. Cascading risks complicate communication during disasters. Cascading risks are largely associated with “the anthropogenic domain and the vulnerability component of risk. This results in a disaster escalation process. In other words, it focuses mainly on the management of social and infrastructure nodes” (p. 2253) . Simply, cascading risk(s) occur when one hazardous event (e.g., Winter Storm Uri) triggers other events/risks, which produce even more severe consequences (e.g., BWNs, infrastructure failure). Thus, effective communication is key to protecting health and safety when a disaster has cascading risks. More particularly, clear communication of BWNs can enhance individuals’ efficacy levels by persuading them to properly boil their tap water, or use another suitable water source (e.g., bottled water) to mitigate contamination risks, especially during a disaster with cascading risks like Winter Storm Uri , – . To create effective risk messages, communicators must understand risk perception and how it influences actions of at-risk individuals. Simply, risk perception is “the belief that one is vulnerable” to some future risk such as disease, floods, etc. . While risk perception can be conceptualized in many ways, Rimal and Real view risk perception as a combination of a person’s perceived susceptibility and perceived severity of a risk and these are assessed based on threats that may occur in the future . While risk perception alone is typically not enough to predict behavioral intentions, it does help indicate the communicative needs of at-risk individuals and it can help guide message development , , – . Risk perception is an integral element for risk communication as it helps communicators align their message(s) with audience (i.e., at-risk individuals) concerns and needs , , . As may be evident, there is a gap in research related to investigating these phenomena, specifically, as related to BWNs and risk communication. As such, investigating risk perception in the current study is important as it has important implications for risk communication in the context of disasters with cascading risks, like BWNs resulting from Winter Storm Uri. Winter Storm Uri was a disaster with various cascading risks, with one such risk being possible water contamination. The lack of electricity due to the winter storm made communicating with at-risk individuals challenging. Previous research about BWNs notes that “the quick response of utility managers is important to protect more consumers, and using the news media as the only means to protect consumers may not provide high levels of public health protection,” (p. 2051) . Not only does the context of Uri exemplify an extremely complex disaster, but it also highlights the communication difficulties that utility managers, city and state governments, and at-risk individuals faced when ‘normal’ communication channels—such as news media and social media—were not fully available or able to reach all at-risk populations due to limitations related to travel, electrical and cellphone service, etc. – . As such, the goal of this study was to investigate BWN-affected individuals’ risk perceptions, water quality perceptions, and perceived efficacy as related to compliance with BWNs during a complex disaster with cascading risks: Winter Storm Uri. Additionally, the aforementioned phenomena can reveal communicative exigencies of affected individuals as well as help inform recommendations for future risk communication about BWNs that emerge from larger disasters. There is limited risk communication research focused on drinking water issues, especially during severe weather events, which often involve cascading risk(s) , . Despite awareness of an active BWN, compliance with suggested protective actions varies among affected populations. BWN compliance rates are reported to range from 36 to 98% . Previous experience with a risk may increase risk perception and encourage positive protective actions , , – . Further, “Past experience with a hazard is generally thought to influence one’s recognition that a risk exists and increases motivation to protect one-self,” (p. 1841) . As such, the following hypothesis was employed to investigate the relationship between past experience with a BWN and risk perception during Winter Storm Uri: H1 Previous experience with a BWN(s) will alter individuals' perceptions of water quality and the risk of acquiring a waterborne disease during Uri. Due to the complex nature of Winter Storm Uri and its cascading risks, communicating with affected individuals was difficult. For instance, one study reported that more than half of Uri-affected respondents were unsure or confused about whether they were under a BWN during Uri , highlighting a severe risk communication issue. This finding implies serious communication gaps between officials/utility companies and the public. Perhaps even more concerning is that extant research denotes that racial minority communities (e.g., Black and Latinx communities) and lower-income household tend to have less access to clean, safe water . More specifically, during Uri, racial minorities and lower-income households experienced more severe issues, like burst pipes . On the other hand, communities with larger populations of non-Hispanic White residents, single family homes, and higher-income households experienced a smaller percentage of power outages after the storm , meaning they likely had access to electricity, communication technologies, and running water sooner than other communities with higher populations of racial minorities and lower-income households. This information further complicates risk communication surrounding Uri, especially related to BWNs, but also highlights the need for better risk communication that is specifically tailored to diverse populations’ communicative needs. Another factor to consider when crafting risk communication for BWNs is increasing compliance among various at-risk populations. Common reasons for non-compliance during a BWN include forgetfulness, perceived inconvenience, appearance of clean water, not believing the initial notification of a BWN, and failing to receive communication about an issued BWN , , , . Furthermore, low risk perception is consistently reported as a primary reason for non-compliance – . However, Americans have consistently expressed concerns about contamination of drinking water with 83% reporting a “great deal” or “fair” amount of worry in 2022 . Perceptions about the quality and safety of water may interact with risk perceptions and efficacy beliefs to impact decisions about protective actions , , . Both self- and response-efficacy are important in relation to compliance with recommended protective actions during disasters. Efficacy plays a role in individuals’ perceived risk , which is important to know when creating risk communication. However, there were vast communication issues surrounding Winter Storm Uri. The interactive role of risk perception and efficacy , , and compliance in the context of BWN’s following winter storms is not well understood. Understanding these phenomena could contribute to more effective risk communication practices during complex disasters, such as Uri. Therefore, EPPM and extant research suggesting that risk perception and efficacy beliefs play a role in people’s decisions to take protective actions (or not) when experiencing potential water contamination, inform the following hypothesis: H2 Higher levels of perceived efficacy will be related to individuals’ (risk) perceptions of water quality and the risk of acquiring a waterborne disease during Uri. H3 General risk perception(s), efficacy, and (risk) perception of water quality will influence compliance with BWNs during Uri. A more detailed discussion of literature on EPPM, risk and crisis communication, and methods are provided in Supplemental Information . To address gaps in understanding how risk perceptions influence response to BWNs surrounding a weather-related crisis, the 2021 Winter Storm Uri, and to evaluate our hypotheses, we conducted a survey of Texas and Oklahoma residents aimed at understanding how individuals perceive risk surrounding BWNs associated with Uri. Previous experience with a BWN(s) will alter individuals' perceptions of water quality and the risk of acquiring a waterborne disease during Uri. Due to the complex nature of Winter Storm Uri and its cascading risks, communicating with affected individuals was difficult. For instance, one study reported that more than half of Uri-affected respondents were unsure or confused about whether they were under a BWN during Uri , highlighting a severe risk communication issue. This finding implies serious communication gaps between officials/utility companies and the public. Perhaps even more concerning is that extant research denotes that racial minority communities (e.g., Black and Latinx communities) and lower-income household tend to have less access to clean, safe water . More specifically, during Uri, racial minorities and lower-income households experienced more severe issues, like burst pipes . On the other hand, communities with larger populations of non-Hispanic White residents, single family homes, and higher-income households experienced a smaller percentage of power outages after the storm , meaning they likely had access to electricity, communication technologies, and running water sooner than other communities with higher populations of racial minorities and lower-income households. This information further complicates risk communication surrounding Uri, especially related to BWNs, but also highlights the need for better risk communication that is specifically tailored to diverse populations’ communicative needs. Another factor to consider when crafting risk communication for BWNs is increasing compliance among various at-risk populations. Common reasons for non-compliance during a BWN include forgetfulness, perceived inconvenience, appearance of clean water, not believing the initial notification of a BWN, and failing to receive communication about an issued BWN , , , . Furthermore, low risk perception is consistently reported as a primary reason for non-compliance – . However, Americans have consistently expressed concerns about contamination of drinking water with 83% reporting a “great deal” or “fair” amount of worry in 2022 . Perceptions about the quality and safety of water may interact with risk perceptions and efficacy beliefs to impact decisions about protective actions , , . Both self- and response-efficacy are important in relation to compliance with recommended protective actions during disasters. Efficacy plays a role in individuals’ perceived risk , which is important to know when creating risk communication. However, there were vast communication issues surrounding Winter Storm Uri. The interactive role of risk perception and efficacy , , and compliance in the context of BWN’s following winter storms is not well understood. Understanding these phenomena could contribute to more effective risk communication practices during complex disasters, such as Uri. Therefore, EPPM and extant research suggesting that risk perception and efficacy beliefs play a role in people’s decisions to take protective actions (or not) when experiencing potential water contamination, inform the following hypothesis: Higher levels of perceived efficacy will be related to individuals’ (risk) perceptions of water quality and the risk of acquiring a waterborne disease during Uri. General risk perception(s), efficacy, and (risk) perception of water quality will influence compliance with BWNs during Uri. A more detailed discussion of literature on EPPM, risk and crisis communication, and methods are provided in Supplemental Information . To address gaps in understanding how risk perceptions influence response to BWNs surrounding a weather-related crisis, the 2021 Winter Storm Uri, and to evaluate our hypotheses, we conducted a survey of Texas and Oklahoma residents aimed at understanding how individuals perceive risk surrounding BWNs associated with Uri. Preliminary analysis Some 99.9% of respondents reported they were affected by the storm from February 14 to February 26 . More than half (53.2%) of the respondents reported that they had no running water, 58% had no electricity, 75% had low water pressure, 28% had discolored water, 21% reported water with a bad smell, and 31% had a frozen water pipe. The majority (83.2%) of the respondents received BWNs or related advisories. Descriptive statistics Overall, results show that perceptions of risks surrounding BWNs during Uri were high, with means score was 3.5 on a five-point scale (all reported results are based on a five-point scale, unless noted otherwise). Perceived severity of the risk associated with BWNs saw a higher mean ( M = 4.29, SD = 0.72) than susceptibility ( M = 2.41, SD = 1.50). Respondents reported overall satisfaction with household water during normal conditions (median response = somewhat satisfied, M = 3.81, SD = 1.22), perceived it as safe (median response = somewhat safe, M = 4.09, SD = 1.11), and rated the quality as higher than average (median response = good, M = 2.86 out of 4, SD = 0.82). Testing hypothesis 1 (H1) A significant difference was observed in respondents’ perception of water quality based on whether they had previous experience with BWNs (U = 71,076.0, z = − 2.174, n = 791, p = 0.030). No significant difference was detected in perceived risk of acquiring a waterborne disease based on previous experience with BWNs (U = 66,877.5, z = − 0.052, n = 733, p = 0.959). However, a difference in risk perception was observed between respondents that were under a BWN and those that were not (U = 23,453.5, z = − 2.932, n = 707, p = 0.003), with those under a BWN scoring higher on the risk perception scale (median = 3.5 vs. 3.0). A similar difference was not observed in perceived water quality between these groups that were and were not under a BWN ( p = 0.204). Nonetheless, even when restricting the analysis to only respondents that report being under a BWN, the difference in perceived water quality held up (U = 47,841.0, z = − 2.664, n = 662, p = 0.008) and no difference was detected in perceived risk (U = 46,032.5, z = − 0.283, n = 613, p = 0.777). These results offer partial support for H1, that previous experience with a BWN(s) alters an individuals' perceptions of water quality and risk of acquiring a waterborne disease during Uri. Testing hypothesis 2 (H2) Small but significant correlations were observed between the levels of perceived efficacy and water quality and risk perceptions (Table ). Efficacy had a small but significant correlation with the perceived risk of acquiring a waterborne disease during Uri ( [12pt]{minimal} $${ }_{B}$$ τ B =0.055, n = 723, p = 0.043). Efficacy also correlated with water quality perception during Uri ( [12pt]{minimal} $${ }_{B}$$ τ B =0.200, n = 760, p < 0.0005). Both relationships provide evidence supporting H2. Family income was found to correlate, sometimes inversely, with all perceptions investigated (Table ). Testing hypothesis 3 (H3) Several binary logistic regression models were investigated to evaluate H3 (Table ). Respondents’ ability to boil water was always a significant variable ( p < 0.0005). When respondents reported they had some ability to boil water, they were more than 4.6 times more likely to comply with the BWN (OR = 4.63–4.70). Interestingly, when respondents reported not having any limitation to boil water (i.e., full ability to boil water), the likelihood was slightly less (OR = 2.73–2.89, p [12pt]{minimal} $$$$ ≤ 0.006). Family income and race were often significant as well, although the level of significance was not consistent ( p = 0.019–0.064). Nonetheless, respondents who reported a family income greater than $35,000 were twice as likely to boil water (OR = 1.96–3.49) compared to those making less than $35,000. Race was at least moderately significant ( p [12pt]{minimal} $$$$ ≤ 0.07) in all models. While there were 35 respondents who reported two or more races, the number of non-white respondents was small (n [12pt]{minimal} $$$$ ≤ 10), limiting the statistical power of our analysis. Nonetheless, across all models, American Indian or Alaskan Native respondents were about 80% less likely to report boiling water than respondents who reported to be White ( p [12pt]{minimal} $$$$ ≤ 0.010). Age, sex, and level of education of respondents were not found to be significant predictors of BWN adherence ( p > 0.1). Therefore, models with these variables were not included in Table . The first model (Table , Model 1) evaluated the likelihood that respondents boiled water based on perceptions of risk and efficacy as well as respondents’ ability to boil water, as defined by Day et al. and the American Water Works Association (AWWA) , and whether a minor was in the household. While the model explained a small percentage of the variability in response (r 2 = 0.196), it was significant (χ 2 = 61.496, p < 0.0005). Risk perception, as defined by Rimal and Real , had an insignificant ( p = 0.489) inverse relationship with the likelihood respondents reported boiling water. Perceived efficacy (OR = 1.75, p = 0.019) and ability to boil water ( p < 0.0005) were positively associated with adherence to BWNs. The presence of a minor in the household increased the likelihood that survey respondents reported boiling their water by 70%, although this relationship was moderately significant in this model ( p = 0.068). The next model (Table , Model 2) evaluated the same parameters as Model 1 but replaced the risk perception variable with the reported perception of water quality. This model behaved similarly overall (r 2 = 0.211, χ 2 = 70.119, p < 0.0005). Respondents’ perception of water quality was positively associated with the likelihood of boiling water (OR = 1.28), although this was only marginally significant ( p = 0.055). For each unit increase in water quality scale, there is approximately a 28% increase in the likelihood water will be boiled before consumption. Like the previous model, perceived efficacy (OR = 1.56, p = 0.033) and ability to boil water ( p < 0.0005) positively influenced BWN adherence. The presence of a minor in the household nearly doubled the likelihood that water was boiled (OR = 1.98, p = 0.017). In Model 3, we included both perceptions of risk and water quality. Similar to Model 1, risk perception was not a significant predictor ( p = 0.750). Similar to Model 2, the perception of water quality was again positively associated with whether respondents boiled water, although this was not significant ( p = 0.162). The influence of respondents perceived efficacy (OR = 1.59, p = 0.65) and whether a minor was in the household (OR = 1.73, p = 0.058) observed in this model was similar to results observed for Models 1 and 2. Because the risk perception scale was insignificant in all of the binary logistic regression models investigated, we also explored the influence of the susceptibility and severity scales that constituted the risk perception scale defined by Rimal and Real . In this set of models (Table ), predictor variables that were at least moderately significant in previous models were included. Water quality perception (OR = 1.16–1.19, p = 0.215–0.266), perceived efficacy (OR = 1.45–1.70, p = 0.025 to 0.145), whether a minor was in the household (OR = 1.75–1.80, p = 0.042 to 0.054), the ability to boil water ( p < 0.0005), family income ( p = 0.020–0.028) and race ( p = 0.63–0.074) behaved similarly across all models and was consistent with Models 1–3. The focus of Models 4–6, measures of susceptibility and severity, were not found to be significant predictors with p values greater than 0.122 in all cases. Some 99.9% of respondents reported they were affected by the storm from February 14 to February 26 . More than half (53.2%) of the respondents reported that they had no running water, 58% had no electricity, 75% had low water pressure, 28% had discolored water, 21% reported water with a bad smell, and 31% had a frozen water pipe. The majority (83.2%) of the respondents received BWNs or related advisories. Overall, results show that perceptions of risks surrounding BWNs during Uri were high, with means score was 3.5 on a five-point scale (all reported results are based on a five-point scale, unless noted otherwise). Perceived severity of the risk associated with BWNs saw a higher mean ( M = 4.29, SD = 0.72) than susceptibility ( M = 2.41, SD = 1.50). Respondents reported overall satisfaction with household water during normal conditions (median response = somewhat satisfied, M = 3.81, SD = 1.22), perceived it as safe (median response = somewhat safe, M = 4.09, SD = 1.11), and rated the quality as higher than average (median response = good, M = 2.86 out of 4, SD = 0.82). A significant difference was observed in respondents’ perception of water quality based on whether they had previous experience with BWNs (U = 71,076.0, z = − 2.174, n = 791, p = 0.030). No significant difference was detected in perceived risk of acquiring a waterborne disease based on previous experience with BWNs (U = 66,877.5, z = − 0.052, n = 733, p = 0.959). However, a difference in risk perception was observed between respondents that were under a BWN and those that were not (U = 23,453.5, z = − 2.932, n = 707, p = 0.003), with those under a BWN scoring higher on the risk perception scale (median = 3.5 vs. 3.0). A similar difference was not observed in perceived water quality between these groups that were and were not under a BWN ( p = 0.204). Nonetheless, even when restricting the analysis to only respondents that report being under a BWN, the difference in perceived water quality held up (U = 47,841.0, z = − 2.664, n = 662, p = 0.008) and no difference was detected in perceived risk (U = 46,032.5, z = − 0.283, n = 613, p = 0.777). These results offer partial support for H1, that previous experience with a BWN(s) alters an individuals' perceptions of water quality and risk of acquiring a waterborne disease during Uri. Small but significant correlations were observed between the levels of perceived efficacy and water quality and risk perceptions (Table ). Efficacy had a small but significant correlation with the perceived risk of acquiring a waterborne disease during Uri ( [12pt]{minimal} $${ }_{B}$$ τ B =0.055, n = 723, p = 0.043). Efficacy also correlated with water quality perception during Uri ( [12pt]{minimal} $${ }_{B}$$ τ B =0.200, n = 760, p < 0.0005). Both relationships provide evidence supporting H2. Family income was found to correlate, sometimes inversely, with all perceptions investigated (Table ). Several binary logistic regression models were investigated to evaluate H3 (Table ). Respondents’ ability to boil water was always a significant variable ( p < 0.0005). When respondents reported they had some ability to boil water, they were more than 4.6 times more likely to comply with the BWN (OR = 4.63–4.70). Interestingly, when respondents reported not having any limitation to boil water (i.e., full ability to boil water), the likelihood was slightly less (OR = 2.73–2.89, p [12pt]{minimal} $$$$ ≤ 0.006). Family income and race were often significant as well, although the level of significance was not consistent ( p = 0.019–0.064). Nonetheless, respondents who reported a family income greater than $35,000 were twice as likely to boil water (OR = 1.96–3.49) compared to those making less than $35,000. Race was at least moderately significant ( p [12pt]{minimal} $$$$ ≤ 0.07) in all models. While there were 35 respondents who reported two or more races, the number of non-white respondents was small (n [12pt]{minimal} $$$$ ≤ 10), limiting the statistical power of our analysis. Nonetheless, across all models, American Indian or Alaskan Native respondents were about 80% less likely to report boiling water than respondents who reported to be White ( p [12pt]{minimal} $$$$ ≤ 0.010). Age, sex, and level of education of respondents were not found to be significant predictors of BWN adherence ( p > 0.1). Therefore, models with these variables were not included in Table . The first model (Table , Model 1) evaluated the likelihood that respondents boiled water based on perceptions of risk and efficacy as well as respondents’ ability to boil water, as defined by Day et al. and the American Water Works Association (AWWA) , and whether a minor was in the household. While the model explained a small percentage of the variability in response (r 2 = 0.196), it was significant (χ 2 = 61.496, p < 0.0005). Risk perception, as defined by Rimal and Real , had an insignificant ( p = 0.489) inverse relationship with the likelihood respondents reported boiling water. Perceived efficacy (OR = 1.75, p = 0.019) and ability to boil water ( p < 0.0005) were positively associated with adherence to BWNs. The presence of a minor in the household increased the likelihood that survey respondents reported boiling their water by 70%, although this relationship was moderately significant in this model ( p = 0.068). The next model (Table , Model 2) evaluated the same parameters as Model 1 but replaced the risk perception variable with the reported perception of water quality. This model behaved similarly overall (r 2 = 0.211, χ 2 = 70.119, p < 0.0005). Respondents’ perception of water quality was positively associated with the likelihood of boiling water (OR = 1.28), although this was only marginally significant ( p = 0.055). For each unit increase in water quality scale, there is approximately a 28% increase in the likelihood water will be boiled before consumption. Like the previous model, perceived efficacy (OR = 1.56, p = 0.033) and ability to boil water ( p < 0.0005) positively influenced BWN adherence. The presence of a minor in the household nearly doubled the likelihood that water was boiled (OR = 1.98, p = 0.017). In Model 3, we included both perceptions of risk and water quality. Similar to Model 1, risk perception was not a significant predictor ( p = 0.750). Similar to Model 2, the perception of water quality was again positively associated with whether respondents boiled water, although this was not significant ( p = 0.162). The influence of respondents perceived efficacy (OR = 1.59, p = 0.65) and whether a minor was in the household (OR = 1.73, p = 0.058) observed in this model was similar to results observed for Models 1 and 2. Because the risk perception scale was insignificant in all of the binary logistic regression models investigated, we also explored the influence of the susceptibility and severity scales that constituted the risk perception scale defined by Rimal and Real . In this set of models (Table ), predictor variables that were at least moderately significant in previous models were included. Water quality perception (OR = 1.16–1.19, p = 0.215–0.266), perceived efficacy (OR = 1.45–1.70, p = 0.025 to 0.145), whether a minor was in the household (OR = 1.75–1.80, p = 0.042 to 0.054), the ability to boil water ( p < 0.0005), family income ( p = 0.020–0.028) and race ( p = 0.63–0.074) behaved similarly across all models and was consistent with Models 1–3. The focus of Models 4–6, measures of susceptibility and severity, were not found to be significant predictors with p values greater than 0.122 in all cases. The Winter Storm Uri produced various cascading risks for individuals living in Texas and Oklahoma. Findings from this research highlight, (a) that most Uri-affected respondents believed the water risks were severe, (b) that some demographic variables impacted BWN compliance, while previous BWN experiences decreased water quality perceptions but did not increase risk perceptions, implying a possible risk paradox effect , (c) that higher levels of perceived efficacy correlated to higher levels of BWN compliance, and (d) risk perception had an inverse relationship to respondents’ boiling their water. These results highlight the need for effective risk communication during these types of disasters, as it could be the difference in compliance with protective actions , . As noted previously, non-compliance during a BWN can be a result of forgetfulness, perceived inconvenience, appearance of clean water, not believing the initial notification of a BWN, and/or failing to receive communication about an issued BWN , , , . Additionally, low risk perception is consistently reported as a primary reason for non-compliance with BWNs – . As suggested by our results, a lack of clear communication, perceived inconvenience, inability to boil water, and, at times, influence from risk perception impacted respondents’ BWN compliance. However, communicating BWNs was complicated during Uri. More than half (58%) of respondents reported that they had lost electricity, complicating their access to BWN messages. Yet, the majority (83.2%) of the respondents did report receiving BWNs or related advisories at some point during Uri or soon after Uri. A small number of respondents (3.6%) reported that they were unsure whether they were under a BWN . These findings have important implications for theory, practice, and future inquiry. Moreover, results from this research can inform future risk communication praxis in the context of BWNs , , . Understanding risk perception, water quality perceptions, and perceived efficacy (in relation to BWN compliance) are important elements to understanding the communicative needs among at-risk individuals. These phenomena are important for developing effective and tailored risk communication , , . Results indicate that most respondents believed the water risks associated with Uri were severe and, thus, many had high risk perception about these events. However, previous experience with BWNs did not significantly influence risk perceptions, but perceived efficacy did show a (small) significant correlate with individuals’ perceived risk of acquiring a waterborne disease. These results can be partially explained by risk paradox , which has significant implications for risk communication. It is often assumed that if an individual has high risk perception about a threat, they will be more likely to prepare and/or enact risk mitigation behavior(s); however, the opposite can also occur . Sometimes, individuals with high-risk perception and/or previous experience with a particular risk still do not adequately prepare for future risks, for a variety of reasons. First, individuals may understand the risk posed by BWNs, but choose to accept the risk, perhaps because they are overburdened by other risks, such as those caused by Uri (e.g., lack of electricity, inability to travel to get supplies) , , . Second, individuals may understand the risk posed by BWNs, but they may see someone else as responsibility for enacting the protective action (e.g., boiling water), such as the head-of-the-house, a spouse, a parent, etc. . Third, individuals may understand the risks and would like to enact the protective action (e.g., boil water), but they may lack the resources to do so . For instance, during Uri, many individuals did not have electricity, which, in many cases, hindered their ability to boil water. Furthermore, travel conditions were not safe during Uri. Texas and Oklahoma generally lacked the infrastructure to clear roadways in a safe and timely manner – . Thus, developing risk communication in the context of Winter Storm Uri was complicated. There has been limited research examining BWNs that occur due to weather-related disasters , . Severe weather events, such as Uri, can create cascading risks, leading to additional challenges with communication , . Thus, these findings contribute new knowledge about a specific form and context for crisis and risk communication. Results are supportive of EPPM propositions as efficacy is positively correlated with increased risk mitigation behavior (i.e., boiling water). Additionally, perceived efficacy had a positive relationship with perceptions of risk and water quality. Some of our results, therefore, are consistent with extant research that has used EPPM in other contexts, such as in a hypothetical weather-related emergency and a radiological “dirty” bomb event , hearing-loss protection for agricultural workers , smokers’ risks and readiness to quit , and colorectal cancer screenings . However, the nature of Winter Storm Uri presents important, contextual factors that require further inquiry. According to EPPM, exposure to a fear-appeal message prompts an individual to appraise the threat and then appraise their ability to prevent the threat/comply with protective actions , . However, perceptions of BWNs during an extreme weather disaster like Uri may fundamentally differ from BWNs that occur outside of a weather disaster. BWNs during extreme weather disasters may be one of many risks and risk messages that individuals are receiving and appraising. Stated another way, risk communication within the larger context of a crisis may differ from risk communication in normal times. During ‘regular’ BWNs, individuals may feel confident in their ability to comply, but during an extreme event like Uri, individuals’ ‘regular’ perceived efficacy and sense of threat may be altered since Uri disrupted power, made roads impassable, and created other risks. In many cases individuals may have been unable to comply with the BWN. Thus, in this context, BWNs could have been considered a ‘dread risk’ since they were one of numerous cascading risks , . Dread risk accounts for whether a given threat is perceived as very severe, controllable, catastrophic, fatal, increasing, involuntary, and whether it evokes fear and worry (i.e., dread) . As noted, most respondents believed water risks during Uri to be severe (i.e., high threat). Yet, other research related to BWNs during extreme weather disasters—like Hurricane Katrina—found that individuals had low levels of perceived risk . Further, Vedachalam et al.’s (2016) meta-analysis on compliance with BWNs found that, “awareness of BWA was moderately high, except in situations involving extreme weather,” (p. 136) . Variations in risk perception around BWNs suggest a need to examine how dread risk and EPPM tenets function across typical BWN events and during extreme weather disasters. Such research should also continue to examine how different demographic variables impact efficacy beliefs and risk perceptions, inclusive to water quality perceptions. Age, gender, and level of education were not found to influence the likelihood of whether respondents complied with BWNs during Uri, although income level was influential in increasing BWN compliance. Lai et al. reported similar findings related to income, but contradictory findings related to age and level of education, noting that “respondents who were older and had higher levels of education and income were likely to have a wider range of disaster information repertoires,” (p. 747) like emergency supplies . These results highlight the need for future research as well as the need to better understand at-risk populations so that risk messages more effectively promote efficacy and acknowledge various risk perceptions . This study also suggests that current conceptualizations of risk perception may be too simplistic. Risk perception is currently conceptualized and measured primarily as a combination of severity + susceptibility for generic, nonspecific threats . However, “threat” can be perceived in more complex ways than just “severity” and “susceptibility.” In the context of Uri, for example, respondents answered questions about their perceived risks surrounding water borne diseases (which are the ‘threat’ that BWNs aim to mitigate). Though this potential risk is communicated in a BWN, these questions do not link this risk specifically to the context of the ongoing disaster (i.e., Uri). Therefore, an individual’s overall perception of risk may be captured in the current conceptualization, but risk in relation to a specific risk event may not. Different conceptualization(s) may impact results, which is especially important because many disasters have cascading effects and numerous risks that emerge due to the initial disaster event. During Uri, impacted populations experienced extreme winter weather, power outages, loss of heat, frozen pipes, damage to buildings and infrastructure, travel restrictions, water issues and over 200 individuals died . Little research specifically involving communication has examined these forms of cascading risks . When asked about risk perception surrounding water borne disease (i.e., the risk associated with BWNs), respondents may have been thinking about this risk in relation to other risks posed by Uri. Results from this study suggest that extreme disasters can impact efficacy levels. While affected individuals may have high efficacy (self + response) outside of disaster contexts, a lack of resources due to a disaster can impact efficacy during the event when individuals are trying to enact protective actions . As Witte purports in EPPM, individuals may have the intention to comply with a protective action—which is guided in part by perceived efficacy—but what stops them from executing the action may relate to their (lack of) skills and/or environmental constraints . This nuance was exemplified during Uri, as people may have intended to boil water, but could not do so due to loss of electricity. In addition, this disaster also complicated how risk perception and efficacy function as related to messaging. Stated another way, efficacy messages that cannot be followed due to environmental factors/constraints may impact risk perception and response behaviors. The complicated nature of sending crisis messages during Uri also influences how risk perception and efficacy typically function to influence at-risk individuals’ response behaviors. For instance, EPPM suggests that messages that individuals perceive as threatening can produce adaptive, desired responses (e.g., boiling water) when both perceived threat and efficacy are high , . Yet, when individuals cannot receive potentially “threatening messages” that signal risk (i.e., BWN), EPPM’s assumption may not hold. Efficacious messages are not only important to help disaster-affected individuals comply with protective actions, but they are also important because they can help publics’ practice preparedness during pre-crisis times , . Further, efficacious messages can help at-risk individuals reduce uncertainty and better understand risks . In these situations, efficacious messages may prompt information seeking, and increase knowledge sufficiency about the event . However, the very nature of Uri complicated this action for individuals since many did not have power due to the storm (i.e., cascading risks). Information seeking, as related to efficacy, is important because it can mediate the effects of perceived susceptibility (one aspect of risk perception) and anxiety, decrease message rejection, partially mediate effects linked to perceived susceptibility and fear, and thus, potentially lead to an increase in overall message acceptance . As such, future research should examine risk communication during events like Winter Storm Uri and query how affected publics’ efficacy levels are impacted by such communication. A non-representative, cross-sectional, survey was administered using Qualtrics XM (Qualtrics, Provo, UT). All methods were carried out in accordance with the methods approved by the Institutional Review Board (IRB) at the University of Texas at Tyler’s (IRB-FY2021-129) and Wayne State University (IRB-21-02-3278). Using the IRB-approved recruitment script, all respondents were presented an information sheet prior to their participation in this research and consented before starting the online survey. To participate in the study, respondents needed to live in Texas or Oklahoma during February 14–February 26, 2021 (Winter Storm Uri and related BWN parameters) and be at least 18 years of age. Respondents were asked to verify this information at the beginning of the survey and enter their city, state, and five-digit zip code to confirm their residence. Responses from adults (18 + years old) living in Texas and Oklahoma during the Winter Storm Uri were collected March 2 through April 21, 2021 (Fig. ). The survey took respondents approximately seven minutes (median response time) to complete. Some 99.9% of respondents reported they were affected by Uri from February 14–26, 2021. Overall, there were a total of 893 respondents; 775 from Texas, 101 from Oklahoma (including Native American reservations), and 17 other respondents (see Table ). Data collection began via snowball sampling and a targeted Facebook advertisement campaign and occurred from March to May 2021. For snowball sampling, researchers posted the survey link on their social media pages. Using the IRB-approved recruitment script, researchers also asked their social networks to take and/or share the survey. Additionally, a paid-for-advertisement campaign was placed on Facebook to promote the survey to individuals in Texas and Oklahoma based on user data. The advertisement promoted the IRB-approved recruitment script and the Qualtrics link. Researchers also shared the survey link with their non-social media social networks, such as university colleagues and academic communities. Risk perception was calculated as the product of susceptibility and severity, using four adapted items from Rimal and Real . Following Rimal and Real , response scores for four questions focused on perceived susceptibility to and severity of waterborne disease were averaged to provide an indexed measure of risk perception. Additionally, the average scores for susceptibility and severity were also investigated individually to determine how these perceptions influence risk-mitigating behavior. Perceptions of water quality were assessed with three questions adapted from the AWWA survey on public perceptions of tap water safety . Because the scales for responses to these three questions were different, the scores were normalized and averaged to constitute a water quality perception measure. Six items from Witte et al.’s Risk Behavior Diagnosis Scale were used to assess perceived efficacy related to risk. Average scores for the six questions were calculated and used as a continuous variable defining perceived efficacy . Additional details regarding survey questions are presented in the . Respondents were asked to indicate gender identity, age, race, number of people in their household, if children live in the household, family income, level of education, and employment status , . Statistical analysis All statistical analyses were performed in SPSS (Version 29, IBM). Hypothesis 1 (“Previous experience with a BWN(s) will alter individuals' perceptions of water quality and the risk of acquiring a waterborne disease during Uri”) was evaluated using Mann–Whitney U tests, because these data are not normally distributed. Hypothesis 2 (“Higher levels of perceived efficacy will be related to individuals’ perceptions of water quality and the risk of acquiring a waterborne disease during Uri”) was evaluated by assessing correlations between perceived efficacy and perceptions of risk and water quality. Because variables investigated included those that are not normally distribution (e.g., water quality perception) and ordinal (e.g., family income) the non-parametric Kendell’s tau b ( [12pt]{minimal} $${ }_{B}$$ τ B ) was used to assess correlations relevant to H2. Hypothesis 3 (“Perceptions of efficacy, water quality, and risk will influence compliance with BWNs during Uri”) was evaluated using a series of binary logistic regression models. Only respondents that received a BWN or similar notification were included in this analysis. The likelihood that respondents boiled water was based on the general equation: [12pt]{minimal} $$logit={L}_{i}=ln(_{1}}{1-{P}_{1}})={ }_{0}+{ }_{1}{x}_{1,i}+ +{ }_{k}{x}_{k,i}$$ l o g i t = L i = l n P 1 1 - P 1 = β 0 + β 1 x 1 , i + ⋯ + β k x k , i where [12pt]{minimal} $${L}_{i}$$ L i is the odds that a survey respondent boils water; P 1 is the probability of outcome 1 (i.e. the respondent boils water); [12pt]{minimal} $${x}_{k, i}$$ x k , i are predictor variables such as perceptions of risk, water quality, and efficacy, and respondents’ gender, education, age, income, ability to boil water, living with minor(s), and previous BWNs experience; [12pt]{minimal} $${ }_{k}$$ β k are fitted coefficients; and k is the number of predictors. For regression models, perceptions of risk, water quality and efficacy are treated as continuous variables. More information regarding collinearity of model predictors can be found in the . Limitations Several limitations should be considered when interpreting these results. First, respondents were recruited through Facebook advertisements. While this limited potential respondents to those who utilize and have access to Facebook, it allowed for sampling of the disaster-affected population (Texas and Oklahoma) through ads targeting users’ geo-location tool. Further, this method of recruiting provided an expedient way of reaching respondents soon after Uri while many were still under a BWN. Second, to protect respondent privacy, rather than asking respondents for their street level address, the survey only requested respondents identify their city, state and zip code. Therefore, our analysis was limited to matching responses to the smallest spatial unit possible, which typically was the city. This limitation prevented us from performing further spatial analyses or incorporating other census information that might have enhanced our understanding of local conditions that may have influenced respondent responses to BWNs. Third, while widely used and well-established, the risk perception survey items had low reliability scores in this study. Though the survey items asked respondents about future risk, the low reliability may be related to individuals responding based on current feelings of susceptibility, due to being under a BWN at the time. It is also possible that some respondents may have considered frozen pipes, a lack of resources and power, BWNs, and other cascading effects from Uri as a singular risks event rather than viewing them as individual risks emerging from a larger disaster. Thus, measuring risk perception during an ongoing disaster may require a more nuanced approached. Future research should further examine such an approach. All statistical analyses were performed in SPSS (Version 29, IBM). Hypothesis 1 (“Previous experience with a BWN(s) will alter individuals' perceptions of water quality and the risk of acquiring a waterborne disease during Uri”) was evaluated using Mann–Whitney U tests, because these data are not normally distributed. Hypothesis 2 (“Higher levels of perceived efficacy will be related to individuals’ perceptions of water quality and the risk of acquiring a waterborne disease during Uri”) was evaluated by assessing correlations between perceived efficacy and perceptions of risk and water quality. Because variables investigated included those that are not normally distribution (e.g., water quality perception) and ordinal (e.g., family income) the non-parametric Kendell’s tau b ( [12pt]{minimal} $${ }_{B}$$ τ B ) was used to assess correlations relevant to H2. Hypothesis 3 (“Perceptions of efficacy, water quality, and risk will influence compliance with BWNs during Uri”) was evaluated using a series of binary logistic regression models. Only respondents that received a BWN or similar notification were included in this analysis. The likelihood that respondents boiled water was based on the general equation: [12pt]{minimal} $$logit={L}_{i}=ln(_{1}}{1-{P}_{1}})={ }_{0}+{ }_{1}{x}_{1,i}+ +{ }_{k}{x}_{k,i}$$ l o g i t = L i = l n P 1 1 - P 1 = β 0 + β 1 x 1 , i + ⋯ + β k x k , i where [12pt]{minimal} $${L}_{i}$$ L i is the odds that a survey respondent boils water; P 1 is the probability of outcome 1 (i.e. the respondent boils water); [12pt]{minimal} $${x}_{k, i}$$ x k , i are predictor variables such as perceptions of risk, water quality, and efficacy, and respondents’ gender, education, age, income, ability to boil water, living with minor(s), and previous BWNs experience; [12pt]{minimal} $${ }_{k}$$ β k are fitted coefficients; and k is the number of predictors. For regression models, perceptions of risk, water quality and efficacy are treated as continuous variables. More information regarding collinearity of model predictors can be found in the . Several limitations should be considered when interpreting these results. First, respondents were recruited through Facebook advertisements. While this limited potential respondents to those who utilize and have access to Facebook, it allowed for sampling of the disaster-affected population (Texas and Oklahoma) through ads targeting users’ geo-location tool. Further, this method of recruiting provided an expedient way of reaching respondents soon after Uri while many were still under a BWN. Second, to protect respondent privacy, rather than asking respondents for their street level address, the survey only requested respondents identify their city, state and zip code. Therefore, our analysis was limited to matching responses to the smallest spatial unit possible, which typically was the city. This limitation prevented us from performing further spatial analyses or incorporating other census information that might have enhanced our understanding of local conditions that may have influenced respondent responses to BWNs. Third, while widely used and well-established, the risk perception survey items had low reliability scores in this study. Though the survey items asked respondents about future risk, the low reliability may be related to individuals responding based on current feelings of susceptibility, due to being under a BWN at the time. It is also possible that some respondents may have considered frozen pipes, a lack of resources and power, BWNs, and other cascading effects from Uri as a singular risks event rather than viewing them as individual risks emerging from a larger disaster. Thus, measuring risk perception during an ongoing disaster may require a more nuanced approached. Future research should further examine such an approach. Supplementary Information. |
Evaluation of a deformable image registration algorithm for image‐guided thermal ablation of liver tumors on clinically acquired MR‐temperature maps | 2b9c66eb-86de-450f-8231-353f316abe0e | 11788246 | Surgical Procedures, Operative[mh] | INTRODUCTION Percutaneous thermal ablation of oligo metastatic disease or early‐stage hepatocellular carcinoma (HCC) is an established therapy since guidelines on many disease modalities recommend thermoablation as a first‐line treatment according to the number and size of the lesion. , , , , , Interventional thermoablation procedures aim at destroying pathological tissues via a localized energy deposit. Ideally, the resulting ablated zone is aimed to cover the lesion borders with additional margins while avoiding unwanted damage to the surrounding tissues. The size of HCC varies a lot (from <1 to >10 cm) but standard treatment with a single probe under imaging guidance assumes a maximum size of 3 cm In the absence of precise temperature mapping during the procedure, empirical energy delivery (power and emission time) are selected by the clinician according to recommendations of the vendor (typically based on ex vivo experiments). As a result, personalized treatment remains difficult and the prediction of the resulting ablated zone remains imprecise since, in vivo, the size of the ablation zone is affected by perfusion, heat‐sink effects due to large vessels, and individual tissue composition. Thermal ablation procedures are associated with a 6%−12% local recurrence due to incomplete lesion coverage despite contrast‐enhanced control images acquired at the end of the procedures. , Real time quantitative monitoring of thermal energy deposition appears necessary to improve therapeutic procedures and to avoid heat‐induced complications in surrounding healthy tissues. MRI has the ability to non‐invasively monitor local temperature changes during the thermal therapies using quantitative real time temperature mapping techniques. , , The linear dependence of proton resonance frequency (PRF) shift on temperature is the basis of MR thermal monitoring. The standard method is based on phase mapping techniques using gradient echo imaging and measures the relative change in temperature throughout the acquisition in each voxel of the image. Knowing the local tissue temperature, it is then possible to predict cell death using the thermal dose calculation in each voxel. In particular, real time monitoring of the thermal dose distribution at the target region can be used to determine (and minimize) the required duration of the thermal ablation, whereas monitoring of the temperature distribution in adjacent risk structures is required to be able to interrupt the ablation if predefined temperature thresholds should be exceeded. Nevertheless, respiratory motion causes significant challenges in abdominal organ imaging. Apnea cannot be an option for a typical treatment with duration of several minutes and synchronization of the image acquisition with the physiological motion is currently used. , This approach has been applied recently with FLASH sequences during MWA on patients. , However, these approaches remain relatively slow with an update time equal to the typical duration of the physiological motion (about 6−7.5 s for one slice with FLASH sequence), leading to limited spatial coverage and temporal resolution of the thermometry. As a result, the characterization of the spatial extent of the ablation and precise computation of the accumulated thermal dose from temperature maps is limited. An alternative three‐step approach is to acquire the image continuously using very fast acquisition sequence (<100 ms per slice), such as single‐shot EPI, to eliminate intra‐scan motion. Inter‐scan motion, corresponding to the change in organ position (∼ 4–5 voxels or 9–12 mm apart) between two successive scans, must be corrected to follow the temporal evolution of the temperature in each voxel throughout the procedure. For this purpose, deformable image registration (DIR) methods such as optical flow (OF) algorithms exploit the conservation of intensity between successive images and estimate a displacement field per voxel that reflects the motion of each region of the image. The estimated motion can then be applied to phase images in a two‐step procedure: (i) Each phase image is registered to a fixed reference position based on the estimated motion field. (ii) Assuming a linear relationship between periodic organ displacement and phase variation, a phase correction also referred to as correction of magnetic susceptibility artifacts is performed to compute the temperature. In such context, local intensity changes during MWA (due to changes in tissue MR properties with heating) may be interpreted by the algorithm as a local motion and lead to erroneous estimation of local vector fields and final temperature estimates. Errors up to 10°C have been reported in static agar gel experiments using this strategy. A novel motion compensation algorithm was proposed to constrain the computation of motion fields during the heating period and to deliver non‐compromised temperature maps. This algorithm has been evaluated in silico and on preclinical data but not in the context of clinical ablation. The purpose of this study is to evaluate a novel DIR algorithm that enhances the generation of thermal maps aligned to a reference position, a critical step for calculating cumulative thermal dose and, consequently, for real‐time control and evaluation of interventional procedure progress as explained above. The method was retrospectively evaluated on datasets from patient that underwent MWA. The proposed approach was first evaluated on motion‐free datasets (gated acquisitions) to compare the robustness of the proposed OF method with the original method and to ensure the presence of a gold standard reference. Then, to demonstrate the feasibility and advantages of the proposed method under actual treatment conditions, the method was applied to motion datasets (fixed‐frequency acquisitions).
METHODS Study population Data were derived from patients who underwent microwave ablation (MWA) of primary or secondary liver tumors under MRI monitoring at the University Hospital of Munich within a prospective trial (Clinical Trial Register Number: DRKS00028515) between December 2020 and July 2023. The additional post hoc analysis was approved by the ethics committee of LMU Munich, and informed consent was waived due to the retrospective nature of the study. A total of 11 patients (36% female, 64% male, age 65 ± 9 years) with 13 lesions were included in the analysis. Reablation after needle repositioning was performed in two patients (G#4–G#5) and (FF#7–FF#8). Ablation procedure During the procedure, each patient was under general anesthesia with the permanent care of medical staff. An AveCure microwave system (MedWave, San Diego, USA) was used to perform the ablation using a 14‐gauge large antenna inserted percutaneously under MRI guidance. The device was connected to a generator located outside the Faraday cage using a shielded cable provided by the manufacturer. Ablation duration was set to 9 ± 2 min with a target temperature of 120°C (based on the lesion size, the recommendations of the vendor, and previous experience with the MWA system) and a delay of 30 s was observed before starting the energy deposition. The size of the ablation zone (on day one after ablation) is typically smaller than expected from vendor recommendation. To overcome too‐small ablation zones, the ablation time is usually increased in order to reach sufficiently large ablation volumes. With the chosen parameters, ablation volumes were clinically adequate in the presented cases. Thermometry acquisition A stack of slices was acquired dynamically on a 1.5T MRI scanner (Magnetom Aera, Siemens Healthineers, Erlangen, Germany) using a single‐shot gradient‐echo echo planar imaging (EPI) sequence: matrix size = 128 × 128, slice thickness = 3 mm, no slice gap, TE = 18 ms, TR = 2000 ms, FA = 90°, pixel bandwidth = 1445 Hz/pixel, phase encoded direction = left‐right, GRAPPA acceleration factor = 2, 6/8 partial Fourier, echo spacing = 0.78 ms, echo train length 47. Parameters varying between different acquisitions (total acquisition time, number of slices, number of repetitions, field of view, spatial resolution …) are listed in Table . To avoid aliasing of the patient's arms, two saturation bands were positioned on each side of the patient. Two strategies were chosen. MRI acquisitions of five microwave treatments were performed under respiratory gating during the exhalation phase using a cushion positioned in the abdomen of the patient. To freeze the motion, gated acquisitions limit the acquisition window to the most stable part of respiration (i.e., approximately 1 s in expiration), thereby reducing spatial coverage. The maximum image update rate is then equal to the respiratory period (approximately 0.2 Hz). Any small shift in the respiratory cycle between the reference image and the current image will introduce both a positional error in the voxels located on the image and a difference in magnetic susceptibility. This can lead to significant temperature bias. Eight acquisitions were also performed without respiratory gating. In this case, the acquisitions are triggered every 2 s (or at a fixed frequency of 0.5 Hz), which is about twice as fast as the respiratory rate of patients under general anesthesia. The stack of slices was acquired in paracoronal or parasagittal to minimize through‐plane motion and to locate the microwave antenna in the central slice of the stack. The MRI DICOM phase and magnitude data were transmitted in real time to a workstation to calculate temperature maps using the software “Certis Solution” version 1.2.0 (Certis Therapeutics, Pessac, France) and displayed in the console room with a delay and a frame rate of approximately 2 s. The frame rate was limited by the acquisition rate of the sequence listed in Table . A different image and temperature reconstruction framework is used in this study. Image reconstruction The spine coil integrated into the MRI bed and a loop coil positioned on the abdomen and surrounding the insertion point of the device were used for data acquisition (13 receiver channels for image reconstruction). The raw data files were converted into ISMRM Raw Data format and reconstructed using the Gadgetron framework. Initial steps included regridding, EPI ghost‐Nyquist correction and coil compression. The Gadgetron implementation of GRAPPA and partial Fourier (PF) reconstruction were then applied to correct for phase‐encoding undersampling in image reconstruction. Deformable image registration For each image in the time series, the displacement field (u,v) was estimated from the magnitude images and used to register both magnitude and phase images to a fixed reference position. The reference position is defined as the median position estimated from the different positions observed over the first 10 stacks of slices acquired. In practice, it will find a location near the peak of the breathing cycle for gated acquisition (dark green dots in Figure ) while it will find a location in the middle of the respirator cycle for fixed frequency acquisition (dark red dots in Figure ). This solution minimizes the displacement when estimating and applying the vector field. Since OF algorithms rely on the local conservation of the intensity, the drop in signal intensity induced by heating and changes in tissue properties may introduce errors in the estimated motion (u, v). Therefore, two different OF algorithms have been evaluated: The conventional OF using a Horn and Schunck (H&S) implementation computes vector fields representing displacement between each new magnitude image and the image at the reference position. The PCA‐based OF method introduced in (see Figure ) identifies spatial and temporal consistencies in the motion of the observed region through preparative learning covering several breathing cycles. This enables, during hyperthermia, the elimination of mis‐registration inherent to tissue heating. The method requires a preparative learning step covering several breathing cycles before starting the ablative procedure (∼15 frames). The 2D estimated motion fields (u, v) are collected together with the registered phase images using the conventional OF method since local variations of intensity are unexpected. A PCA is then performed on the motion fields of the learning step to extract eigenvector maps and eigenvalues. During the interventional procedure (from the sixteenth frame to the end), for each new incoming image, the 2D motion is estimated as a linear combination of the previously computed eigenvectors. Conventional OF and PCA‐based OF methods were both used to register magnitude/phase images in order to evaluate the impact of the algorithms on MR thermometry, dosimetry (computed using the equivalent minutes at 43°C), and lesion volume estimation. 2.1 Correction of respiratory‐induced susceptibility artifacts Correction of respiration‐induced susceptibility artifacts was then performed by parameterizing the phase using a PCA on a pixel‐by‐pixel basis. Details on this method can be found in Maclair et al or refs. . The approach is divided into two steps: a learning phase during which the influence of the displacement of the phase susceptibility is estimated using a model parameterization, and an intervention step during which the phase correction is estimated based on the actual motion state and subtracted from the current temperature image to remove the motion‐induced susceptibility artifacts. It should be noted that the learning phase (from the first frame to the fifteenth frame) and the intervention phase (from the sixteenth frame to the end) can be defined in the same way as in the previous step concerning image registration. In the learning phase, a PCA was applied to a collection of motion fields to obtain motion descriptors. A first‐order variation of local phase changes due to motion is then considered and can be written as a linear combination of motion descriptors and a set of parameterized magnetic field models. During the intervention, the largest PCA‐based motion descriptors were estimated from the current motion field and the background phase φ ref was computed for each incoming acquisition. For clarity, it should be pointed out that the methodology uses two methods based on Principal Component Analysis. The first, referred to throughout the article as “PCA‐based,” indicates that the motion fields during the intervention phase will be a linear combination of those observed during the preparatory phase. The second, called “Correction of respiratory‐induced susceptibility artifacts” indicates that the background phase φ ref is a model parametrization. The second was thus applied to the motion fields of the two OF algorithms: “Conventional” and “PCA‐based.” 2.2 Temperature calculation Temperature calculation was performed using the PRF method that computes temperature change Δ T from the difference between a given phase image φ t acquired during treatment and a reference phase image φ ref acquired prior to heating. Δ T = φ t − φ ref . γ . σ . TE . B 0 . − 1 where γ is the gyromagnetic ratio (≈42.58 MHz T−1), σ = −0.0094 ppm·°C−1 is the PRF temperature coefficient, B 0 . is the magnetic field strength (1.5 T here) and T E is the echo time. The temperature change estimation ΔT was carried out using the phase subtraction described in Equation (1) but with three different reference phases: Gold standard method: with fixed reference phase image φ ref Conventional OF: with the computed background phase φ ref computed from motion fields estimated with the conventional OF method. PCA‐based OF: with the computed background phase φ ref computed from motion fields estimated with the PCA‐based OF method. Spatial‐temporal drift correction and temporal filtering using a first‐order low‐pass Butterworth filter with a cutoff frequency of 0.14 Hz were finally applied based on this initial implementation. 2.3 Thermal dose and lesion volume estimation The cumulative thermal dose (TD) also described as the cumulative equivalent minutes (CEM) was computed from the temperature images using the Sapareto equation. The latter establishes an empirical relationship between the absolute temperature, the exposure time, and cell death. The lesion volume was estimated by taking an equivalent dose of 240 min at 43°C (CEM 43 ), as it is the theoretical threshold of cell death. The initial temperature was set to 37°C for each patient. 2.4 Data analysis The data analysis has two objectives: first, to compare the robustness of the proposed OF method with the original method. Second, to demonstrate the feasibility and advantages of the proposed method under actual treatment conditions. For the first objective, image data sets from five treatments performed under gated MRI acquisitions are used, assuming an absence of residual inter‐scan motion. Under these conditions, the gold standard temperature map ΔT is calculated by simple phase subtraction without DIR. Temperature maps were then calculated after DIR using conventional OF and PCA‐based OF methods. Differences in motion field, temperature, and lesion volume size estimation were compared for the three methods using the metrics described below. For the second objective, image data sets from eight treatments were performed without MRI gating. Under these conditions, the standard subtraction (formerly the gold standard) method was not relevant due to the presence of motion but was calculated and displayed for information purposes. Nevertheless, to a certain extent, the standard subtraction was considered, in some voxels, as a good indicator of temperature: for some cases, the limited motion (due to general anesthesia) and the large and long ablation (the spatial gradient temperature between voxels was relatively low) helped to estimate the “true” temperature. The two different OF algorithms and the resulting metrics were then compared qualitatively. In gated acquisitions, the acquisition frame rate was considered equal to the patient's respiratory rate imposed by the ventilator. The exact frame update was extracted from the timestamp of the kspace data. To characterize the regularity of the respiratory motion and image similarity, the Intercorrelation coefficient was computed between the current magnitude image and the reference magnitude image selected from the first 10 images and located in the middle of the respiratory cycle. The metric is computed on the whole image and averaged overall slices. Corr I , J = ∑ I − I ^ J − J ^ ∑ I − I ^ 2 ∑ J − J ^ 2 where ( I , J ) are the current magnitude and reference magnitude images. Motion correction performance and image similarity quality was then evaluated with the three methods by computing over all dynamic acquisitions the normalized root mean square error (NRMSE) between the current and the reference stack of images: NRMSE I , J = ∑ i = 1 N I i − J i 2 N where ( I , J ) are the current magnitude and reference magnitude stack of images and N the total number of voxels within the stack of images. The potential artifact on temperature images introduced by erroneous motion correction due to signal intensity variation on magnitude images was quantified by calculating the resulting bias in image registration. To do this, the averaged endpoint error (AEE) was calculated for each repetition over a ROI of 19 × 19 voxels centered on the ablated area. The metric is calculated on the extent of the ablation, which corresponds to a minimum of 4 slices and up to 20 slices depending on the patient and acquisition setup. As the SNR reduction is very localized in the vicinity of the needle, only voxels with temperature higher than 10°C during the course of the ablation were included in the spatial averaging. The statistical analysis in the whisker diagram was then computed over 100 dynamic acquisitions taken during the heating period. EE ( rep ) = ( u − u ref ) 2 + ( v − v ref ) 2 and AEE rep = 1 N ∑ i = 1 N E E i rep where ( u , v ) and ( u ref , v ref ) are the estimated and the reference motion, respectively. ( u ref , v ref ) is the motion estimated using the gold standard method while ( u , v ) is the motion computed individually with the conventional OF and the PCA‐based OF. And i are the voxel location within the volume where the temperature is higher than 10°C during the course of the ablation. Thermometry performance was evaluated for each motion compensation algorithm for computing: The NRMSE and the absolute temperature error. The evaluation was carried out on a 19 × 19 ROI centered on the ablated area. Such metrics were used to measure and compare the temperature bias between the gold standard and the proposed DIR methods. The statistical analysis in the whisker diagram was computed over 100 dynamic acquisitions taken during the heating period. Dosimetry and lesion volume estimation performance were evaluated for each motion compensation algorithm by computing: The accumulated thermal dose (CEM 43 ), the time to reach the accumulated thermal dose threshold (in dynamic acquisition), the difference in time to reach the accumulated thermal dose threshold. These metrics were used to measure and compare the potential bias between the gold standard and the proposed motion compensation algorithms. The lesion volume estimated in cm 3 over 10 consecutive dynamic acquisitions. The bias in volume estimation between the gold standard and the proposed DIR methods is then assessed using the Bland–Altman plot.
Data were derived from patients who underwent microwave ablation (MWA) of primary or secondary liver tumors under MRI monitoring at the University Hospital of Munich within a prospective trial (Clinical Trial Register Number: DRKS00028515) between December 2020 and July 2023. The additional post hoc analysis was approved by the ethics committee of LMU Munich, and informed consent was waived due to the retrospective nature of the study. A total of 11 patients (36% female, 64% male, age 65 ± 9 years) with 13 lesions were included in the analysis. Reablation after needle repositioning was performed in two patients (G#4–G#5) and (FF#7–FF#8).
During the procedure, each patient was under general anesthesia with the permanent care of medical staff. An AveCure microwave system (MedWave, San Diego, USA) was used to perform the ablation using a 14‐gauge large antenna inserted percutaneously under MRI guidance. The device was connected to a generator located outside the Faraday cage using a shielded cable provided by the manufacturer. Ablation duration was set to 9 ± 2 min with a target temperature of 120°C (based on the lesion size, the recommendations of the vendor, and previous experience with the MWA system) and a delay of 30 s was observed before starting the energy deposition. The size of the ablation zone (on day one after ablation) is typically smaller than expected from vendor recommendation. To overcome too‐small ablation zones, the ablation time is usually increased in order to reach sufficiently large ablation volumes. With the chosen parameters, ablation volumes were clinically adequate in the presented cases.
A stack of slices was acquired dynamically on a 1.5T MRI scanner (Magnetom Aera, Siemens Healthineers, Erlangen, Germany) using a single‐shot gradient‐echo echo planar imaging (EPI) sequence: matrix size = 128 × 128, slice thickness = 3 mm, no slice gap, TE = 18 ms, TR = 2000 ms, FA = 90°, pixel bandwidth = 1445 Hz/pixel, phase encoded direction = left‐right, GRAPPA acceleration factor = 2, 6/8 partial Fourier, echo spacing = 0.78 ms, echo train length 47. Parameters varying between different acquisitions (total acquisition time, number of slices, number of repetitions, field of view, spatial resolution …) are listed in Table . To avoid aliasing of the patient's arms, two saturation bands were positioned on each side of the patient. Two strategies were chosen. MRI acquisitions of five microwave treatments were performed under respiratory gating during the exhalation phase using a cushion positioned in the abdomen of the patient. To freeze the motion, gated acquisitions limit the acquisition window to the most stable part of respiration (i.e., approximately 1 s in expiration), thereby reducing spatial coverage. The maximum image update rate is then equal to the respiratory period (approximately 0.2 Hz). Any small shift in the respiratory cycle between the reference image and the current image will introduce both a positional error in the voxels located on the image and a difference in magnetic susceptibility. This can lead to significant temperature bias. Eight acquisitions were also performed without respiratory gating. In this case, the acquisitions are triggered every 2 s (or at a fixed frequency of 0.5 Hz), which is about twice as fast as the respiratory rate of patients under general anesthesia. The stack of slices was acquired in paracoronal or parasagittal to minimize through‐plane motion and to locate the microwave antenna in the central slice of the stack. The MRI DICOM phase and magnitude data were transmitted in real time to a workstation to calculate temperature maps using the software “Certis Solution” version 1.2.0 (Certis Therapeutics, Pessac, France) and displayed in the console room with a delay and a frame rate of approximately 2 s. The frame rate was limited by the acquisition rate of the sequence listed in Table . A different image and temperature reconstruction framework is used in this study.
The spine coil integrated into the MRI bed and a loop coil positioned on the abdomen and surrounding the insertion point of the device were used for data acquisition (13 receiver channels for image reconstruction). The raw data files were converted into ISMRM Raw Data format and reconstructed using the Gadgetron framework. Initial steps included regridding, EPI ghost‐Nyquist correction and coil compression. The Gadgetron implementation of GRAPPA and partial Fourier (PF) reconstruction were then applied to correct for phase‐encoding undersampling in image reconstruction.
For each image in the time series, the displacement field (u,v) was estimated from the magnitude images and used to register both magnitude and phase images to a fixed reference position. The reference position is defined as the median position estimated from the different positions observed over the first 10 stacks of slices acquired. In practice, it will find a location near the peak of the breathing cycle for gated acquisition (dark green dots in Figure ) while it will find a location in the middle of the respirator cycle for fixed frequency acquisition (dark red dots in Figure ). This solution minimizes the displacement when estimating and applying the vector field. Since OF algorithms rely on the local conservation of the intensity, the drop in signal intensity induced by heating and changes in tissue properties may introduce errors in the estimated motion (u, v). Therefore, two different OF algorithms have been evaluated: The conventional OF using a Horn and Schunck (H&S) implementation computes vector fields representing displacement between each new magnitude image and the image at the reference position. The PCA‐based OF method introduced in (see Figure ) identifies spatial and temporal consistencies in the motion of the observed region through preparative learning covering several breathing cycles. This enables, during hyperthermia, the elimination of mis‐registration inherent to tissue heating. The method requires a preparative learning step covering several breathing cycles before starting the ablative procedure (∼15 frames). The 2D estimated motion fields (u, v) are collected together with the registered phase images using the conventional OF method since local variations of intensity are unexpected. A PCA is then performed on the motion fields of the learning step to extract eigenvector maps and eigenvalues. During the interventional procedure (from the sixteenth frame to the end), for each new incoming image, the 2D motion is estimated as a linear combination of the previously computed eigenvectors. Conventional OF and PCA‐based OF methods were both used to register magnitude/phase images in order to evaluate the impact of the algorithms on MR thermometry, dosimetry (computed using the equivalent minutes at 43°C), and lesion volume estimation.
Correction of respiratory‐induced susceptibility artifacts Correction of respiration‐induced susceptibility artifacts was then performed by parameterizing the phase using a PCA on a pixel‐by‐pixel basis. Details on this method can be found in Maclair et al or refs. . The approach is divided into two steps: a learning phase during which the influence of the displacement of the phase susceptibility is estimated using a model parameterization, and an intervention step during which the phase correction is estimated based on the actual motion state and subtracted from the current temperature image to remove the motion‐induced susceptibility artifacts. It should be noted that the learning phase (from the first frame to the fifteenth frame) and the intervention phase (from the sixteenth frame to the end) can be defined in the same way as in the previous step concerning image registration. In the learning phase, a PCA was applied to a collection of motion fields to obtain motion descriptors. A first‐order variation of local phase changes due to motion is then considered and can be written as a linear combination of motion descriptors and a set of parameterized magnetic field models. During the intervention, the largest PCA‐based motion descriptors were estimated from the current motion field and the background phase φ ref was computed for each incoming acquisition. For clarity, it should be pointed out that the methodology uses two methods based on Principal Component Analysis. The first, referred to throughout the article as “PCA‐based,” indicates that the motion fields during the intervention phase will be a linear combination of those observed during the preparatory phase. The second, called “Correction of respiratory‐induced susceptibility artifacts” indicates that the background phase φ ref is a model parametrization. The second was thus applied to the motion fields of the two OF algorithms: “Conventional” and “PCA‐based.”
Temperature calculation Temperature calculation was performed using the PRF method that computes temperature change Δ T from the difference between a given phase image φ t acquired during treatment and a reference phase image φ ref acquired prior to heating. Δ T = φ t − φ ref . γ . σ . TE . B 0 . − 1 where γ is the gyromagnetic ratio (≈42.58 MHz T−1), σ = −0.0094 ppm·°C−1 is the PRF temperature coefficient, B 0 . is the magnetic field strength (1.5 T here) and T E is the echo time. The temperature change estimation ΔT was carried out using the phase subtraction described in Equation (1) but with three different reference phases: Gold standard method: with fixed reference phase image φ ref Conventional OF: with the computed background phase φ ref computed from motion fields estimated with the conventional OF method. PCA‐based OF: with the computed background phase φ ref computed from motion fields estimated with the PCA‐based OF method. Spatial‐temporal drift correction and temporal filtering using a first‐order low‐pass Butterworth filter with a cutoff frequency of 0.14 Hz were finally applied based on this initial implementation.
Thermal dose and lesion volume estimation The cumulative thermal dose (TD) also described as the cumulative equivalent minutes (CEM) was computed from the temperature images using the Sapareto equation. The latter establishes an empirical relationship between the absolute temperature, the exposure time, and cell death. The lesion volume was estimated by taking an equivalent dose of 240 min at 43°C (CEM 43 ), as it is the theoretical threshold of cell death. The initial temperature was set to 37°C for each patient.
Data analysis The data analysis has two objectives: first, to compare the robustness of the proposed OF method with the original method. Second, to demonstrate the feasibility and advantages of the proposed method under actual treatment conditions. For the first objective, image data sets from five treatments performed under gated MRI acquisitions are used, assuming an absence of residual inter‐scan motion. Under these conditions, the gold standard temperature map ΔT is calculated by simple phase subtraction without DIR. Temperature maps were then calculated after DIR using conventional OF and PCA‐based OF methods. Differences in motion field, temperature, and lesion volume size estimation were compared for the three methods using the metrics described below. For the second objective, image data sets from eight treatments were performed without MRI gating. Under these conditions, the standard subtraction (formerly the gold standard) method was not relevant due to the presence of motion but was calculated and displayed for information purposes. Nevertheless, to a certain extent, the standard subtraction was considered, in some voxels, as a good indicator of temperature: for some cases, the limited motion (due to general anesthesia) and the large and long ablation (the spatial gradient temperature between voxels was relatively low) helped to estimate the “true” temperature. The two different OF algorithms and the resulting metrics were then compared qualitatively. In gated acquisitions, the acquisition frame rate was considered equal to the patient's respiratory rate imposed by the ventilator. The exact frame update was extracted from the timestamp of the kspace data. To characterize the regularity of the respiratory motion and image similarity, the Intercorrelation coefficient was computed between the current magnitude image and the reference magnitude image selected from the first 10 images and located in the middle of the respiratory cycle. The metric is computed on the whole image and averaged overall slices. Corr I , J = ∑ I − I ^ J − J ^ ∑ I − I ^ 2 ∑ J − J ^ 2 where ( I , J ) are the current magnitude and reference magnitude images. Motion correction performance and image similarity quality was then evaluated with the three methods by computing over all dynamic acquisitions the normalized root mean square error (NRMSE) between the current and the reference stack of images: NRMSE I , J = ∑ i = 1 N I i − J i 2 N where ( I , J ) are the current magnitude and reference magnitude stack of images and N the total number of voxels within the stack of images. The potential artifact on temperature images introduced by erroneous motion correction due to signal intensity variation on magnitude images was quantified by calculating the resulting bias in image registration. To do this, the averaged endpoint error (AEE) was calculated for each repetition over a ROI of 19 × 19 voxels centered on the ablated area. The metric is calculated on the extent of the ablation, which corresponds to a minimum of 4 slices and up to 20 slices depending on the patient and acquisition setup. As the SNR reduction is very localized in the vicinity of the needle, only voxels with temperature higher than 10°C during the course of the ablation were included in the spatial averaging. The statistical analysis in the whisker diagram was then computed over 100 dynamic acquisitions taken during the heating period. EE ( rep ) = ( u − u ref ) 2 + ( v − v ref ) 2 and AEE rep = 1 N ∑ i = 1 N E E i rep where ( u , v ) and ( u ref , v ref ) are the estimated and the reference motion, respectively. ( u ref , v ref ) is the motion estimated using the gold standard method while ( u , v ) is the motion computed individually with the conventional OF and the PCA‐based OF. And i are the voxel location within the volume where the temperature is higher than 10°C during the course of the ablation. Thermometry performance was evaluated for each motion compensation algorithm for computing: The NRMSE and the absolute temperature error. The evaluation was carried out on a 19 × 19 ROI centered on the ablated area. Such metrics were used to measure and compare the temperature bias between the gold standard and the proposed DIR methods. The statistical analysis in the whisker diagram was computed over 100 dynamic acquisitions taken during the heating period. Dosimetry and lesion volume estimation performance were evaluated for each motion compensation algorithm by computing: The accumulated thermal dose (CEM 43 ), the time to reach the accumulated thermal dose threshold (in dynamic acquisition), the difference in time to reach the accumulated thermal dose threshold. These metrics were used to measure and compare the potential bias between the gold standard and the proposed motion compensation algorithms. The lesion volume estimated in cm 3 over 10 consecutive dynamic acquisitions. The bias in volume estimation between the gold standard and the proposed DIR methods is then assessed using the Bland–Altman plot.
RESULTS Figure shows the influence of respiration and/or liver motion on image similarity metrics in three different cases using (i) a temporal plot of intensity profile through the liver and (ii) the intercorrelation coefficient of magnitude images through the acquisition. The first case (G#1) is a respiratory‐gated acquisition, no significant difference is observed in the intensity profile, and the intercorrelation coefficient is higher than 0.95 indicating a high image similarity. A slight decrease is observed at the end of the procedure. In the second case (FF#2) in Figure , the acquisition was carried out without respiratory gating at a fixed frequency of 0.5 Hz. Regular small oscillations, linked to breathing, were visible both on the intensity profile and on the intercorrelation coefficient, which remained above 0.9. The last case (FF#8) in Figure is also carried out without respiratory gating at a fixed frequency of 0.5 Hz. Small oscillations were observed at the beginning of the procedure on both panels followed by a sudden change around dynamic acquisition #40 ( t = 80 s) linked to a sudden contraction of the organ during ablation. The intercorrelation coefficient then falls from 0.9 to 0.8 and gradually rises back to 0.85 at the end of the procedure. This last case was therefore excluded. Figure compares image similarity quality through the acquisition using the intercorrelation coefficient of magnitude images for both gated and non‐gated acquisitions. All gated acquisitions had a score higher than 0.95, while the fixed frequency acquisitions have a score lower but always higher than 0.9. 3.1 Gated acquisition results Figure shows a representative case (FF #6) of MW ablation with magnitude images at t = 0 s (before energy delivery) and t = 400 s (during the heating period) and t = 700 s (during the cooling period). A black hypo intense signal appears around the needle during ablation. It has disappeared toward the end of the ablation and has been replaced by a smaller white hyper intense signal. Figure show temporal measurements of temperature and magnitude as a function of time in a 3 × 3 kernel of voxels near the ablation spot. An approximate maximum temperature change of +50°C was observed. At the end of the procedure, a decrease of temperature is observed although the acquisition was stopped before temperature returned to the baseline. In this experiment, a maximal signal decrease of 47% of original magnitude signal intensity was observed. Figure compares the NRMSE of magnitude images through the acquisition with and without DIR for both gated and non‐gated acquisitions. For gated acquisition, the use of DIR has a moderate impact in NRMSE even if a significant decrease is noticeable for the cases G#1 to G#4. For non‐gated acquisition, the NRMSE calculated without motion correction (in gold) has both a wider distribution and a higher value than for the gated acquisition. The addition of OF approaches drastically reduces both the width of the distribution and the median for all acquisitions. Lastly, no major differences were observed between the two algorithms Conventional OF (in blue) and PCA‐based OF (in red) with the exception of the case FF#3 which is slightly outperformed using the Conventional OF approaches. Figure compares the thermometry performance with the proposed OF algorithms on a respiratory‐gated acquisition (G#3). The gold standard method is a simple phase subtraction with a fixed reference phase image. This assumption remains valid only if there is no motion during the acquisition, which has been verified above. The first and second rows (Figure ) show the magnitude images and a zoomed view in which the liver and MW needle are clearly visible. Although no motion is expected to be present, the conventional OF estimated, at dynamic acquisition #100, a higher vector field (Figure ) than the PCA‐based OF. The maximum EE (Figure ) was 2.67 mm with OF and 0.98 mm with PCA‐based OF. At first glance, the temperature map (Figure ) displays a similar pattern, but some variations can be noticed. Using the conventional OF method, the temperature NRMSE (Figure ) was found higher than 10°C in a large part of the voxel at the vicinity of the needle. As a result, the maximum temperature reached is wrong and the estimation of coagulation necrosis via the thermal dose will be strongly biased. Such a temperature difference is therefore important from a clinical point of view for the safety and accuracy of the procedure. In contrast, using the PCA‐based method, the temperature NRMSE was found below 2°C in most voxels. A few voxels located on the needle are impacted by the needle artifact itself. Temperature evolution in time in a ROI of 5 × 5 pixels was plotted in the right panel (Figure ) and illustrates the bias introduced by both tested methods versus the gold standard (in gold). Temperature errors up to 15°C−20°C are visible in 2 voxels. This bias can either lead to over or underestimation of the measurement in the same voxel (top left) through the acquisition or to underestimation of the measurement (at the center). To illustrate the reproducibility and reliability of the method, one additional case (G#5) is available in Figure . The selected slice is the one next to the one centered on the needle. Temperature error up to 10°C is clearly visible, Figure compares the dosimetry performance with the proposed OF algorithms on the same respiratory‐gated acquisition (G#3). Again, the first and second rows (Figure ) show the magnitude images and a zoomed view. The accumulated thermal dose map (Figure ) displays a similar pattern but some variations can be noticed. Using the conventional OF method, the mask of estimated lesions is smaller by 1–3 voxels in diameter which corresponds to 2.5–7.5 mm. The time to thermal dose threshold map (Figure ) indicates that close to the tip, the cell necrosis threshold is reached in less than 12 dynamic acquisitions while in the border zone, up to 75 dynamic acquisitions were required. The difference in time to reach the accumulated thermal dose threshold map (Figure ) indicates that the conventional OF creates a delay (in orange) or an advance (in blue) of 20 dynamic acquisitions (∼40 s) in lesion size estimation. Accumulated thermal dose evolution in time in a ROI of 5 × 5 pixels was plotted in the right panel (Figure ) and illustrates the bias introduced by both tested methods versus the gold standard (in gold). To illustrate the reproducibility and reliability of the method, one additional case (G#5) are available in Figure corresponds to the dosimetry computation of Figure . A quantitative report of thermometry performance is presented in Figure for all gated acquisitions. Again, the temperature NRMSE was found up to 20°C for the conventional OF method. A statistically significant decrease in averaged EE is observed for the PCA‐based OF with values close to 2.5°C, except for case G#4. Figure compares the lesion size estimation performance through the procedure. Each color indicates a comparison between the estimated volume using the gold standard versus the conventional OF (in blue) or PCA‐based OF (in red). An underestimation of the lesion size was found by both OF algorithms. The PCA‐based OF reported a bias of 0.5 cm 3 with the 95% confidence interval below 2 cm 3 (0.5 cm 3 /−1.5 cm 3 ) while the conventional OF reported a bias of 0.93 cm 3 and a greater dispersion (0.8 cm 3 /−2.6 cm 3 ). Fixed frequency acquisition results . Figure compares the thermometry performance with the proposed OF algorithms on a non‐gated acquisition (FF#6). The first, second and third columns (Figure ) show the magnitude images and a zoomed view at t = 0 s and t = 400 s. In this example, the image plane is oriented perpendicularly to the MW needle. The conventional OF estimated, at dynamic acquisition #200 ( t = 400 s), a higher vector field (Figure ) than the PCA‐based OF. Magnitude signal and temperature evolution in time in a ROI of 3 × 3 pixels was plotted in the right panel (Figure ). A net decrease in magnitude signal (up to 71%) is visible (top left voxel). The temperature estimation (Figure ) without motion correction (in gold) led to strong fluctuations or loss of temperature measurement (around dynamic acquisition #180, top left voxel). The conventional OF algorithm recovers a stable measurement at the cost of a bias ranging from 10°C to 25°C. The proposed PCA‐based OF recovers both a stable and precise temperature measurement without bias. To illustrate the reproducibility and reliability of the method, three additional cases (two for FF#6 and FF#2) are available in Figures . The temperature estimation without motion correction and using standard subtraction (in gold) led to strong fluctuations or loss of temperature measurement which could be recovered by using the PCA‐based algorithm. Figure compares the lesion size estimation performance through the procedure on a non‐gated acquisition. An underestimation of the Conventional OF versus the PCA‐based OF algorithms was found with a bias of 0.47 cm 3 .
Gated acquisition results Figure shows a representative case (FF #6) of MW ablation with magnitude images at t = 0 s (before energy delivery) and t = 400 s (during the heating period) and t = 700 s (during the cooling period). A black hypo intense signal appears around the needle during ablation. It has disappeared toward the end of the ablation and has been replaced by a smaller white hyper intense signal. Figure show temporal measurements of temperature and magnitude as a function of time in a 3 × 3 kernel of voxels near the ablation spot. An approximate maximum temperature change of +50°C was observed. At the end of the procedure, a decrease of temperature is observed although the acquisition was stopped before temperature returned to the baseline. In this experiment, a maximal signal decrease of 47% of original magnitude signal intensity was observed. Figure compares the NRMSE of magnitude images through the acquisition with and without DIR for both gated and non‐gated acquisitions. For gated acquisition, the use of DIR has a moderate impact in NRMSE even if a significant decrease is noticeable for the cases G#1 to G#4. For non‐gated acquisition, the NRMSE calculated without motion correction (in gold) has both a wider distribution and a higher value than for the gated acquisition. The addition of OF approaches drastically reduces both the width of the distribution and the median for all acquisitions. Lastly, no major differences were observed between the two algorithms Conventional OF (in blue) and PCA‐based OF (in red) with the exception of the case FF#3 which is slightly outperformed using the Conventional OF approaches. Figure compares the thermometry performance with the proposed OF algorithms on a respiratory‐gated acquisition (G#3). The gold standard method is a simple phase subtraction with a fixed reference phase image. This assumption remains valid only if there is no motion during the acquisition, which has been verified above. The first and second rows (Figure ) show the magnitude images and a zoomed view in which the liver and MW needle are clearly visible. Although no motion is expected to be present, the conventional OF estimated, at dynamic acquisition #100, a higher vector field (Figure ) than the PCA‐based OF. The maximum EE (Figure ) was 2.67 mm with OF and 0.98 mm with PCA‐based OF. At first glance, the temperature map (Figure ) displays a similar pattern, but some variations can be noticed. Using the conventional OF method, the temperature NRMSE (Figure ) was found higher than 10°C in a large part of the voxel at the vicinity of the needle. As a result, the maximum temperature reached is wrong and the estimation of coagulation necrosis via the thermal dose will be strongly biased. Such a temperature difference is therefore important from a clinical point of view for the safety and accuracy of the procedure. In contrast, using the PCA‐based method, the temperature NRMSE was found below 2°C in most voxels. A few voxels located on the needle are impacted by the needle artifact itself. Temperature evolution in time in a ROI of 5 × 5 pixels was plotted in the right panel (Figure ) and illustrates the bias introduced by both tested methods versus the gold standard (in gold). Temperature errors up to 15°C−20°C are visible in 2 voxels. This bias can either lead to over or underestimation of the measurement in the same voxel (top left) through the acquisition or to underestimation of the measurement (at the center). To illustrate the reproducibility and reliability of the method, one additional case (G#5) is available in Figure . The selected slice is the one next to the one centered on the needle. Temperature error up to 10°C is clearly visible, Figure compares the dosimetry performance with the proposed OF algorithms on the same respiratory‐gated acquisition (G#3). Again, the first and second rows (Figure ) show the magnitude images and a zoomed view. The accumulated thermal dose map (Figure ) displays a similar pattern but some variations can be noticed. Using the conventional OF method, the mask of estimated lesions is smaller by 1–3 voxels in diameter which corresponds to 2.5–7.5 mm. The time to thermal dose threshold map (Figure ) indicates that close to the tip, the cell necrosis threshold is reached in less than 12 dynamic acquisitions while in the border zone, up to 75 dynamic acquisitions were required. The difference in time to reach the accumulated thermal dose threshold map (Figure ) indicates that the conventional OF creates a delay (in orange) or an advance (in blue) of 20 dynamic acquisitions (∼40 s) in lesion size estimation. Accumulated thermal dose evolution in time in a ROI of 5 × 5 pixels was plotted in the right panel (Figure ) and illustrates the bias introduced by both tested methods versus the gold standard (in gold). To illustrate the reproducibility and reliability of the method, one additional case (G#5) are available in Figure corresponds to the dosimetry computation of Figure . A quantitative report of thermometry performance is presented in Figure for all gated acquisitions. Again, the temperature NRMSE was found up to 20°C for the conventional OF method. A statistically significant decrease in averaged EE is observed for the PCA‐based OF with values close to 2.5°C, except for case G#4. Figure compares the lesion size estimation performance through the procedure. Each color indicates a comparison between the estimated volume using the gold standard versus the conventional OF (in blue) or PCA‐based OF (in red). An underestimation of the lesion size was found by both OF algorithms. The PCA‐based OF reported a bias of 0.5 cm 3 with the 95% confidence interval below 2 cm 3 (0.5 cm 3 /−1.5 cm 3 ) while the conventional OF reported a bias of 0.93 cm 3 and a greater dispersion (0.8 cm 3 /−2.6 cm 3 ). Fixed frequency acquisition results . Figure compares the thermometry performance with the proposed OF algorithms on a non‐gated acquisition (FF#6). The first, second and third columns (Figure ) show the magnitude images and a zoomed view at t = 0 s and t = 400 s. In this example, the image plane is oriented perpendicularly to the MW needle. The conventional OF estimated, at dynamic acquisition #200 ( t = 400 s), a higher vector field (Figure ) than the PCA‐based OF. Magnitude signal and temperature evolution in time in a ROI of 3 × 3 pixels was plotted in the right panel (Figure ). A net decrease in magnitude signal (up to 71%) is visible (top left voxel). The temperature estimation (Figure ) without motion correction (in gold) led to strong fluctuations or loss of temperature measurement (around dynamic acquisition #180, top left voxel). The conventional OF algorithm recovers a stable measurement at the cost of a bias ranging from 10°C to 25°C. The proposed PCA‐based OF recovers both a stable and precise temperature measurement without bias. To illustrate the reproducibility and reliability of the method, three additional cases (two for FF#6 and FF#2) are available in Figures . The temperature estimation without motion correction and using standard subtraction (in gold) led to strong fluctuations or loss of temperature measurement which could be recovered by using the PCA‐based algorithm. Figure compares the lesion size estimation performance through the procedure on a non‐gated acquisition. An underestimation of the Conventional OF versus the PCA‐based OF algorithms was found with a bias of 0.47 cm 3 .
DISCUSSION The present study performed at 1.5T using a commercially available multi‐slice EPI sequence investigated the impact of a new 2D DIR workflow to reduce potential bias in MR‐temperature estimation related to changes in T1/T2 tissue properties during the heating period. The proposed acquisition includes 13–20 slices acquired in an interleaved pattern in coronal or sagittal orientation. Phase image quality in all cases was found excellent as shown in Figure , this criterion is a prerequisite for obtaining high‐quality temperature maps. The study first checked the impact of the triggering method on the image quality. The acquisitions were done either using respiratory gating or at a constant update rate of 2 s. The intercorrelation coefficient is a simple and fast metric to calculate for quantifying the degree of similarity between two images. Here, it is calculated over the whole image and can therefore be influenced by any disturbance of the magnitude signal on the liver or outside. Nevertheless, it proved to be very robust in characterizing the regularity of movements and the stability of image positions throughout each experiment (Figure and ). The computation of NRMSE of magnitude images (Figure ) through the acquisition also confirmed the absence of residual movement during gated acquisitions. The validity of the “no motion hypothesis” during the gated acquisition allowed us to investigate and compare the proposed OF algorithms against a gold‐standard approach. While alternative gold‐standard devices (such as probes) are sometimes accessible in phantom or preclinical experiments, such an option cannot be envisioned in clinical practice. We then progressively present the potential bias that is generated on motion field estimation by the conventional OF method. This bias propagated for all calculated metrics, whether it be the temperature (Figure ) or the accumulated thermal dose, the time to reach the thermal dose threshold (Figure ), or the lesion size (Figure ). As shown in Figure and Figure , the bias is much more significant at the core of the ablation because temperatures and associated decrease in magnitude signal are higher than at the periphery of the heated area where it becomes insignificant or null. This bias therefore introduces a significant error ranging from a few degrees to 20°C−30°C on the maximum estimated temperature increase. The error can be positive or negative, even though most of the time an underestimate of the actual temperature is observed. An error of assessment on this criterion introduces a safety risk for the procedure. During liver tumor ablation, high‐power MW settings are used to overcome the heat sink effect of perfusion, resulting in high‐temperature increase (sometimes higher than 100°C). A careful monitoring of the temperature is therefore needed. An error in the maximum temperature measured is also critical for all applications using a feedback control loop for automatic temperature regulation. Although the initial work on this subject is well known, these approaches still remain marginal in clinical practice, and are mainly coupled to protocols on static organs using HIFU energy. , Improving the safety of the liver tumor ablation procedure requires both precise monitoring of the procedure and optimization of the ablation parameters depending on tissue response. Implementing such an approach would address both of these issues. The PCA‐based OF succeeded in recovering the expected temperature measurements where conventional OF failed. A net decrease (up to 15°C) in NRMSE is observed in Figure . One major advantage of the proposed approach is that there is no penalty since it corrects what does not work and offers the same level of accuracy in locations where the decrease in magnitude signal is small or null (Figures ). In most cases, the calculated temperature measurements were sufficiently accurate to provide a complete characterization of the thermal field during ablation (Figure or Figure ). One part of the work sets out to show the reader how an error in the estimation of the vector field can ultimately impact the thermal dose, and therefore the estimation of the final ablated volume from thermal dose images. This error can also be reflected in the time required to reach the necrosis threshold. In this case, the bias on lesion size is not present at the core of the ablation, as the heating times and the energy sent are sufficient to reach the thermal damage threshold for both algorithms. Surprisingly, the bias still acts at the periphery of the ablation, with an underestimate of the temperature leading to an underestimate of the ablation volume. This intuition is confirmed in the five cases presented in the Bland–Altman graph comparing the two methods, where a greater underestimate (∼ 1 cm 3 ) is found with the conventional algorithm. It is important to note here that the statistics and measurements are carried out on the ROI of size 19 × 19 x ∼7 slices centered on the tip of the needle for visualization purposes. A second reason is to keep the percentage of voxels affected by the drop in magnitude signal not too low. The aim of the quantification is therefore not to cover the entire volume of ablation performed nor to characterize the clinical result. The second part of the study applied all approaches in another datasets acquired at a fixed frequency of 0.5 Hz without respiratory gating. The motion was relatively minimal as the patients were under general anesthesia. Typically, the images in the case FF#1 are mostly static but other cases present visible motion of the organs between consecutive images. Such impacts have been quantified by the NRMSE of magnitude intensity (Figure ) and the intercorrelation coefficient (Figure ). The computation of the PRF method therefore needed two corrections, one for freezing the motion between consecutive images, a second for correcting for magnetic susceptibility artifacts due to breathing. The two OF algorithms were applied and as a result, we observed a significant decrease in NRMSE of magnitude intensity (Figure ), indicating that voxel misregistration through the acquisition was equally corrected by both algorithms. A first case study with up to 80% signal decrease of magnitude signal is presented. Temperature estimation was lost or fluctuated in the absence of motion correction ( standard subtraction ) but was recovered by both OF algorithms. A clear bias was visible between both algorithms with an underestimation of 10°C to 25°C. In this specific case, the standard subtraction approach is readable and it was easy to conclude that the proposed PCA‐based OF did recover the correct temperature measurement. On the contrary, the presence of movement makes it difficult, if not impossible, to read standard subtraction temperature data in Figure were recovered by the proposed OF algorithms. The new algorithm created no penalty where the decrease in magnitude signal is small or null (Figure ). Again, in most cases, the calculated temperature measurements were sufficiently accurate to provide a complete characterization of the thermal field during ablation (Figure or Figure ). Lastly, as previously reported in Figure for gated acquisition, a bias in lesion size estimation of approximately 0.5 cm 3 was also reported between the two OF methods for fixed‐frequency acquisitions.
LIMITATIONS Temperature rise induces changes in the tissues MRI properties, resulting in a drop in signal magnitude. However, this is not the only phenomenon that can produce such an effect. During high‐power MW ablation, the tissue temperature might reach locally the boiling point. The presence of air bubbles due to evaporation previously reported by , will create a bulk magnetic susceptibility artifact visible in Figure in magnitude and temperature images. While not demonstrated here, the PCA‐based OF will help in minimizing potential bias induced by the presence of bubbles. Nevertheless, the correction of such artifacts , , was not investigated in this work. The presence of spontaneous motions like a strong contraction of the liver happened in case FF#7 during the heating and were identified by the intercorrelation coefficient metrics. When analyzing the images of Fig S1C, movements are present both in the image plane and out of the plane. In this specific case, the use of 2D slice‐to‐slice OF algorithms is obsolete and requires a 3D approach. Such a scenario will strongly impact both the registration and the motion‐induced susceptibility correction and associated temperature estimation. The correction of such artifacts was not investigated in this work. Clinical perspectives require in‐line integration of the algorithm in the Gadgetron framework. The second step is the validation of the computational time and associated latency. A CPU computation time, reported in the previous study, of 25 ms per image was found. At a fixed frequency of 0.5 Hz even with 20 slices (25 ms x 20 = 500 ms), the existing version would be therefore suitable for real‐time processing in the context of liver tumor ablation, without significant lag between acquisition and temperature display.
CONCLUSION Motion field estimation using OF algorithm can be significantly affected by local variation of signal intensity on magnitude images associated with local tissue heating. A dedicated deformable image algorithm “PCA‐based” was designed and evaluated in a clinical setting in motion‐free datasets and apply then with ones with motion. An accurate assessment of the vector field and temperature was achieved, enabling the size of the lesions to be better quantified. The approach was evaluated using single‐shot EPI, a widely available sequence used in diffusion, perfusion, and fMRI that provides high‐quality phase images of the liver in clinical practice and offers wide spatial coverage (6 cm depth) while maintaining a fast acquisition rate (20 slices / 2 s) to monitor large lesions formation.
Valéry Ozenne and Bruno Quesson are co‐funders and shareholders of Certis Therapeutics. Pierre Bour, Thibaut Faller, and Manon Desclides are employees of Certis Therapeutics.
Supporting Information
|
The Gastric Microbiota Invade the Lamina Propria in | 8e48cae1-6fc9-48db-ab96-74220007734d | 11865006 | Anatomy[mh] | Introduction Gastric adenocarcinoma (GAC) is the fourth leading cause of cancer‐related deaths, accounting for 7.7% of cancer mortalities worldwide . Helicobacter pylori infection is associated with approximately 70% of GAC cases and the global prevalence of H. pylori infection is 43.9% . Correa's cascade describes the histopathological changes to the gastric mucosa during progression to GAC . H. pylori first triggers chronic gastritis (CG), which can lead to gastric intestinal metaplasia (GIM), dysplasia and finally GAC. Although genotypes of H. pylori that encode the virulence factors CagA, VacA and HtrA are more associated with severe disease outcomes, the mechanism by which a minority of H. pylori infections (1%) develop GAC is not fully understood. In recent years, multiple 16S ribosomal RNA (16S rRNA) and metagenomic sequencing studies have profiled the human gastric microbiota during health and H. pylori ‐ associated disease . The proposed model is that the healthy human stomach harbors a distinct microbial community structure and that upon H. pylori infection, this community structure shifts towards a dominance of H. pylori in CG. As carcinogenesis progresses, multiple other bacterial species displace H. pylori and dominate the GAC microbiota. However, sequencing studies are often prone to issues with contamination and do not offer spatial resolution. A longstanding question is whether the gastric microbiota play a causative or correlative role in GAC. Here, we provide high‐resolution spatial images of H. pylori and non‐ H. pylori bacteria in patients with H. pylori ‐positive or negative CG and GIM. Importantly, we provide direct evidence of non‐ H. pylori bacteria invading the lamina propria in the early to middle stages of Correa's cascade, suggesting that non‐ H. pylori bacteria might play a synergistic role in the cause of these disease states.
Materials and Methods 2.1 Sample Preparation and Pretreatment Protocol for Automated Tissue Staining Punch biopsy gastric corpus tissue samples were collected from consenting patients at the Queen Elizabeth Hospital, Birmingham, fixed in formalin and prepared at the Human Biomaterial Resource Centre (HBRC), University of Birmingham by embedding in paraffin (Ethics #17–285) and retrieved from archived stocks at HBRC. Exclusion criteria were previous H. pylori eradication therapy, antibiotic treatment 4 weeks prior to endoscopy, tissue from the cardia, fundus, or antrum and patients under 30 years old. Routine histologic evaluation of hematoxylin and eosin (H&E)‐stained gastric mucosal sections was used for H. pylori diagnosis. Tissue sections (4 μM thick) were prepared for RNAscope and immunohistochemistry (IHC) staining, using a Leica BOND RX Fully Automated Research Stainer. Prior to staining, sections were prepared in the BOND RX, following three short protocols: (1) Deparaffinization, rehydration, hydrogen peroxide and distilled water wash, (2) Heating to 100°C in target retrieval buffer ER2 (pH 9; AR9640) for 45 min and manual washing with distilled water followed by 100% ethanol and drying at 60°C and (3) Incubation in protease III solution for 30 min, followed by a final wash in distilled water. 2.2 Optimization of RNAscope Treatments and Immunohistochemistry Antibody Concentrations for 5‐Plex Automated Staining Automated RNAscope C1 and C2 probes against H. pylori (ACD‐Bio, #542938) and Eubacteria (ACD‐Bio, #464468), respectively, were first tested on healthy colon tissue in a fluorescent multiplex assay, using an RNAscope LS Multiplex Fluorescent Reagent Kit (ACD Bio, #322800), following the standard protocol recommended for this platform. Pretreatment of tissue slides for automated tissue staining with lysozyme was not used as this was found to affect tissue integrity. However, automated tissue staining with lysozyme (micro bacteria detection protocol provided by ACD‐Bio) and a 3‐plex panel (E‐Cadherin and RNAscope probes against H. pylori and Eubacteria) was used on all samples to ensure the bacterial signal was not underestimated in comparison to 5‐plex stained images (data not shown). A positive and negative control probe section was used on every RNAscope run to validate and assess its quality and the sensitivity of the assay. The bacterial gene dapB (ACD Bio, #312038) was used as a negative control to confirm the absence of background noise, and a cocktail of housekeeping genes polr2A C1, ppiB C2 and ubc C3 (ACD Bio, #320868) was used as a positive control to validate the detection of the signal and the tissue integrity. H. pylori gene sequences can bind to both 16S rRNA probes against H. pylori and Eubacteria, whilst non‐ H. pylori bacteria stain with only the Eubacteria probe. All antibodies used in IHC steps were optimized prior, using chromogenic DAB staining of healthy colon tissue in addition to a Leica Bond Polymer Refine Detection kit (Leica, #DS9800). Antigen retrieval was tested using pH 6 (Leica Bond TM Epitope Retrieval 1, #AR9961) and pH 9 (Leica Bond TM Epitope Retrieval 2, #AR9640) buffers by heating to 100°C for 20 min. Three different dilutions were tested for each antibody, as recommended by the manufacturer. Ideal staining pattern and intensity was assessed and approved by a pathologist, whereby slides were then used as reference throughout the validation process. All antibodies were then tested for compliance with the RNAscope pretreatment, to ensure stability after protease III digestion. Once each antibody was assigned an Opal fluorophore, single fluorescence assays were directly compared against DAB‐stained colon tissue to optimize Opal concentration. To assess epitope stability during the following heat steps and to define the order of the addition of antibodies in the multiplex sequence, each antibody was tested individually in the different positions of the panel. Further probe and antibody information can be found in Table . 2.3 Automated 5‐Plex Co‐ RNAscope In Situ Hybridization/Immunohistochemistry Protocol The automated RNAscope Multiplex Fluorescent LS assay (ACD Bio, #322800) was conducted using a Leica BOND RX according to manufacturer's instructions, incubating the sections with C1 and C2 probes against H. pylori and Eubacteria, respectively, for 2 h. An automated IHC staining protocol was immediately followed to fluorescently label E‐cadherin, MUC5AC and MUC2. Between each staining cycle, a heat‐induced stripping step with pH 6 solution was added. Images were acquired using a Vectra Polaris TM multispectral whole slide scanner. Exposure times on the Vectra Polaris Slide scanner for the DAPI, 480, 520, 570, 620, and 690 channels were 1.13 ms, 2.47 ms, 36.06 ms, 5.61 ms, 29.58 ms, and 7.46 ms, respectively. 2.4 Counterstaining and Section Visualization Sections were counterstained with spectral DAPI (Akoya Biosciences) and mounted with ProLong Diamond Antifade Mountant, according to the manufacturer's instructions. Mounted sections were stored in the dark at 4°C until viewing. During optimization steps, sections were visualized using a Zeiss LSM 900 confocal microscope or Mantra 2 Quantitative Pathology Digital Workstation (Akoya Biosciences) and final images were obtained on a Vectra Polaris TM multispectral whole slide scanner (Akoya Biosciences) and saved as a .qptiff file. Images were viewed at 40× magnification using Phenochart Whole Slide Viewer with PhenoImager HT (Akoya Biosciences) or open source QuPath software (Version 0.4.3) . All .qptiff image files were spectrally unmixed by importing and stamping in Phenochart Whole Slide Viewer (Akoya Biosciences), unmixed and exported in InForm software (Akoya Biosciences) and finally restitched as a BIGTIFF file using Visiopharm software (Visiopharm, Hørsholm, Denmark). 2.5 Qualitative and Quantitative Image Analysis For quantitation of target markers in whole slide tissue sections, manual image analysis was conducted using QuPath. For each whole slide scan, tissue regions were first annotated and defined as a region of interest (ROI). Tissue detection was then performed based on the average values of all channels using pixel thresholders. A pixel thresholder was used to calculate mean tissue area (μM 2 ), which was exported. Separate pixel thresholders were then created for each individual Opal channel, corresponding to H. pylori , Eubacteria, MUC5AC or MUC2. The thresholders were saved, and average area annotation measurements (μM 2 ) were obtained for each channel. To account for H. pylori double staining and detection in both the H. pylori and Eubacteria channels, Eubacteria area was quantified by calculating total Eubacteria channel area minus combined H. pylori and Eubacteria channel areas. For each whole slide scan, average percentage area coverage of each marker of interest was calculated by dividing individual channel area over total tissue area x 100. Data were exported from Excel to GraphPad Prism 9 (Version 9.5.1). For qualitative scoring of bacterial invasion, images were viewed using the Phenochart Whole Slide Viewer (Akoya Biosciences) or QuPath. Invasion was scored, whereby 0 = no invasion, 1 = sparse invasion, 2 = moderate invasion (patches of bacteria across sample) and 3 = high invasion (multiple clear regions of bacterial invasion across sample). A Mann–Whitney test was used for statistical analysis in which * indicates p < 0.05, ** indicates p < 0.01, *** indicates p < 0.001 and ns indicates nonsignificant. Data are presented as median ± SEM for each group. 2.6 H&E and Modified Gram Staining Additional sections were stained with H&E for further inflammation analysis and scoring by an independent pathologist. A modified Gram stain was also followed for staining of bacteria in additional tissue sections .
Sample Preparation and Pretreatment Protocol for Automated Tissue Staining Punch biopsy gastric corpus tissue samples were collected from consenting patients at the Queen Elizabeth Hospital, Birmingham, fixed in formalin and prepared at the Human Biomaterial Resource Centre (HBRC), University of Birmingham by embedding in paraffin (Ethics #17–285) and retrieved from archived stocks at HBRC. Exclusion criteria were previous H. pylori eradication therapy, antibiotic treatment 4 weeks prior to endoscopy, tissue from the cardia, fundus, or antrum and patients under 30 years old. Routine histologic evaluation of hematoxylin and eosin (H&E)‐stained gastric mucosal sections was used for H. pylori diagnosis. Tissue sections (4 μM thick) were prepared for RNAscope and immunohistochemistry (IHC) staining, using a Leica BOND RX Fully Automated Research Stainer. Prior to staining, sections were prepared in the BOND RX, following three short protocols: (1) Deparaffinization, rehydration, hydrogen peroxide and distilled water wash, (2) Heating to 100°C in target retrieval buffer ER2 (pH 9; AR9640) for 45 min and manual washing with distilled water followed by 100% ethanol and drying at 60°C and (3) Incubation in protease III solution for 30 min, followed by a final wash in distilled water.
Optimization of RNAscope Treatments and Immunohistochemistry Antibody Concentrations for 5‐Plex Automated Staining Automated RNAscope C1 and C2 probes against H. pylori (ACD‐Bio, #542938) and Eubacteria (ACD‐Bio, #464468), respectively, were first tested on healthy colon tissue in a fluorescent multiplex assay, using an RNAscope LS Multiplex Fluorescent Reagent Kit (ACD Bio, #322800), following the standard protocol recommended for this platform. Pretreatment of tissue slides for automated tissue staining with lysozyme was not used as this was found to affect tissue integrity. However, automated tissue staining with lysozyme (micro bacteria detection protocol provided by ACD‐Bio) and a 3‐plex panel (E‐Cadherin and RNAscope probes against H. pylori and Eubacteria) was used on all samples to ensure the bacterial signal was not underestimated in comparison to 5‐plex stained images (data not shown). A positive and negative control probe section was used on every RNAscope run to validate and assess its quality and the sensitivity of the assay. The bacterial gene dapB (ACD Bio, #312038) was used as a negative control to confirm the absence of background noise, and a cocktail of housekeeping genes polr2A C1, ppiB C2 and ubc C3 (ACD Bio, #320868) was used as a positive control to validate the detection of the signal and the tissue integrity. H. pylori gene sequences can bind to both 16S rRNA probes against H. pylori and Eubacteria, whilst non‐ H. pylori bacteria stain with only the Eubacteria probe. All antibodies used in IHC steps were optimized prior, using chromogenic DAB staining of healthy colon tissue in addition to a Leica Bond Polymer Refine Detection kit (Leica, #DS9800). Antigen retrieval was tested using pH 6 (Leica Bond TM Epitope Retrieval 1, #AR9961) and pH 9 (Leica Bond TM Epitope Retrieval 2, #AR9640) buffers by heating to 100°C for 20 min. Three different dilutions were tested for each antibody, as recommended by the manufacturer. Ideal staining pattern and intensity was assessed and approved by a pathologist, whereby slides were then used as reference throughout the validation process. All antibodies were then tested for compliance with the RNAscope pretreatment, to ensure stability after protease III digestion. Once each antibody was assigned an Opal fluorophore, single fluorescence assays were directly compared against DAB‐stained colon tissue to optimize Opal concentration. To assess epitope stability during the following heat steps and to define the order of the addition of antibodies in the multiplex sequence, each antibody was tested individually in the different positions of the panel. Further probe and antibody information can be found in Table .
Automated 5‐Plex Co‐ RNAscope In Situ Hybridization/Immunohistochemistry Protocol The automated RNAscope Multiplex Fluorescent LS assay (ACD Bio, #322800) was conducted using a Leica BOND RX according to manufacturer's instructions, incubating the sections with C1 and C2 probes against H. pylori and Eubacteria, respectively, for 2 h. An automated IHC staining protocol was immediately followed to fluorescently label E‐cadherin, MUC5AC and MUC2. Between each staining cycle, a heat‐induced stripping step with pH 6 solution was added. Images were acquired using a Vectra Polaris TM multispectral whole slide scanner. Exposure times on the Vectra Polaris Slide scanner for the DAPI, 480, 520, 570, 620, and 690 channels were 1.13 ms, 2.47 ms, 36.06 ms, 5.61 ms, 29.58 ms, and 7.46 ms, respectively.
Counterstaining and Section Visualization Sections were counterstained with spectral DAPI (Akoya Biosciences) and mounted with ProLong Diamond Antifade Mountant, according to the manufacturer's instructions. Mounted sections were stored in the dark at 4°C until viewing. During optimization steps, sections were visualized using a Zeiss LSM 900 confocal microscope or Mantra 2 Quantitative Pathology Digital Workstation (Akoya Biosciences) and final images were obtained on a Vectra Polaris TM multispectral whole slide scanner (Akoya Biosciences) and saved as a .qptiff file. Images were viewed at 40× magnification using Phenochart Whole Slide Viewer with PhenoImager HT (Akoya Biosciences) or open source QuPath software (Version 0.4.3) . All .qptiff image files were spectrally unmixed by importing and stamping in Phenochart Whole Slide Viewer (Akoya Biosciences), unmixed and exported in InForm software (Akoya Biosciences) and finally restitched as a BIGTIFF file using Visiopharm software (Visiopharm, Hørsholm, Denmark).
Qualitative and Quantitative Image Analysis For quantitation of target markers in whole slide tissue sections, manual image analysis was conducted using QuPath. For each whole slide scan, tissue regions were first annotated and defined as a region of interest (ROI). Tissue detection was then performed based on the average values of all channels using pixel thresholders. A pixel thresholder was used to calculate mean tissue area (μM 2 ), which was exported. Separate pixel thresholders were then created for each individual Opal channel, corresponding to H. pylori , Eubacteria, MUC5AC or MUC2. The thresholders were saved, and average area annotation measurements (μM 2 ) were obtained for each channel. To account for H. pylori double staining and detection in both the H. pylori and Eubacteria channels, Eubacteria area was quantified by calculating total Eubacteria channel area minus combined H. pylori and Eubacteria channel areas. For each whole slide scan, average percentage area coverage of each marker of interest was calculated by dividing individual channel area over total tissue area x 100. Data were exported from Excel to GraphPad Prism 9 (Version 9.5.1). For qualitative scoring of bacterial invasion, images were viewed using the Phenochart Whole Slide Viewer (Akoya Biosciences) or QuPath. Invasion was scored, whereby 0 = no invasion, 1 = sparse invasion, 2 = moderate invasion (patches of bacteria across sample) and 3 = high invasion (multiple clear regions of bacterial invasion across sample). A Mann–Whitney test was used for statistical analysis in which * indicates p < 0.05, ** indicates p < 0.01, *** indicates p < 0.001 and ns indicates nonsignificant. Data are presented as median ± SEM for each group.
H&E and Modified Gram Staining Additional sections were stained with H&E for further inflammation analysis and scoring by an independent pathologist. A modified Gram stain was also followed for staining of bacteria in additional tissue sections .
Results 3.1 Localization of H. pylori and the Gastric Microbiota in Chronic Gastritis and Gastric Intestinal Metaplasia Due to anatomical differences and unique microenvironments within the gastric mucosa driving microbial variation , archived formalin‐fixed paraffin‐embedded (FFPE) tissue blocks from only gastric corpus tissue were retrieved from the HBRC ( n = 32) for the purpose of this study. The associated clinical pathology reports confirmed samples were H. pylori ‐ positive CG ( n = 9) or GIM ( n = 10) and H. pylori ‐ negative CG ( n = 6) or GIM ( n = 7). H&E‐stained tissue sections were prepared from the archived FFPE gastric corpus tissue blocks. Again, using routine histologic evaluation, an independent pathologist confirmed H. pylori status and graded all samples for inflammation (Table ). There was no significant difference in inflammation between H. pylori ‐ positive and H. pylori ‐ negative samples (data not shown). To detect H. pylori or non‐ H. pylori bacteria, sections were stained using RNAscope in situ hybridization (ISH) probes against H. pylori ‐ specific or conserved sequences of the 16S rRNA gene, respectively. The probe “Eubacteria” was used to detect non‐ H. pylori bacteria. IHC enabled detection of the host cell markers E‐cadherin, Mucins 5AC (MUC5AC) and 2 (MUC2) using automated tissue staining for CG and GIM samples. Goblet cells, present only in GIM, secrete MUC2, whereas MUC5AC is found in both healthy and diseased gastric mucus. Representative images of GIM and CG samples are presented in Figure and Figure , respectively. Quantification of indicated markers are shown in Figure . Comparable levels of MUC5AC were observed across the patient groups and, as expected, there was a significant increase in MUC2 in GIM patients (Figure C&D). Consistent with previous studies , H. pylori exclusively colonized the gastric glands (Figure ). The presence of Eubacteria correlated with H. pylori infection in CG and GIM but were significantly reduced in the absence of H. pylori infection ( n = 13) (Figure ). This contrasts with the current dogma that there is a unique microbiota in H. pylori ‐ negative individuals . Interestingly, there was an increase in levels of Eubacteria in H. pylori ‐ positive GIM samples compared with H. pylori ‐ positive CG. Although this was not statistically significant, this raised the possibility that the presence of GIM‐specific MUC2 could provide intestinal‐like microniches within the stomach that promote bacterial colonization. As such, we observed colocalization of non‐ H. pylori bacteria with MUC2 in 2/10 H. pylori ‐ positive and 1/7 H. pylori ‐ negative GIM patients (Figure ). 3.2 Association of H. pylori With Invasion of Non‐ H. pylori Bacteria to the Lamina Propria Bacterial invasion is associated with multiple diseases of the gastrointestinal tract, causing damage to the epithelial architecture by triggering inflammatory responses . Whilst H. pylori colonized the gastric glands, non‐ H. pylori bacteria invaded the lamina propria in 6/9 patients with H. pylori ‐ associated CG and 8/10 H. pylori ‐ positive patients with GIM (Figure ). Qualitative scoring of Eubacterial invasion showed significantly more invasion in H. pylori ‐ positive CG and GIM than H. pylori ‐ negative CG and GIM (Figure ). Representative images of H. pylori ‐ positive CG and GIM patients with an invasion score of 3 are shown in Figure H. pylori ‐ negative patients with an invasion score of 0 can be seen in Figure . Very few Eubacteria are seen in samples with an invasion score of 0. It is most likely that these rare occurrences of Eubacteria in the stomach represent transient bacteria, rather than stable colonization . Invasion of H. pylori into the lamina propria is very rare across all patient samples, yet an instance of this can be seen in Figure . Representative images of samples with invasion scores of 1–2 can be seen in Figure . 3.3 Visualization of Non‐ H. pylori Bacteria Using a Modified Gram Stain Next, we aimed to confirm bacterial invasion at the cellular level in patients with H. pylori ‐ positive CG ( n = 4) or GIM ( n = 4). As can be seen in Figure , microniches of Gram‐positive bacteria could be identified in all 4 H. pylori ‐ positive GIM patients. Visualization of Gram‐negative bacteria with this technique was not clear and H&E‐stained sections were superior for identifying H. pylori (data not shown). Bacteria could not readily be identified in H. pylori ‐ positive CG patients using this technique (data not shown). This is consistent with the lower levels of non‐ H. pylori bacteria that were detected in CG patients in comparison to GIM patients (Figure ). Although further work is required to understand the biological consequences of invasive bacteria in carcinogenesis, this modified Gram stain technique could be used alongside the current histological diagnostic tools to stratify patients at higher risk for developing GAC.
Localization of H. pylori and the Gastric Microbiota in Chronic Gastritis and Gastric Intestinal Metaplasia Due to anatomical differences and unique microenvironments within the gastric mucosa driving microbial variation , archived formalin‐fixed paraffin‐embedded (FFPE) tissue blocks from only gastric corpus tissue were retrieved from the HBRC ( n = 32) for the purpose of this study. The associated clinical pathology reports confirmed samples were H. pylori ‐ positive CG ( n = 9) or GIM ( n = 10) and H. pylori ‐ negative CG ( n = 6) or GIM ( n = 7). H&E‐stained tissue sections were prepared from the archived FFPE gastric corpus tissue blocks. Again, using routine histologic evaluation, an independent pathologist confirmed H. pylori status and graded all samples for inflammation (Table ). There was no significant difference in inflammation between H. pylori ‐ positive and H. pylori ‐ negative samples (data not shown). To detect H. pylori or non‐ H. pylori bacteria, sections were stained using RNAscope in situ hybridization (ISH) probes against H. pylori ‐ specific or conserved sequences of the 16S rRNA gene, respectively. The probe “Eubacteria” was used to detect non‐ H. pylori bacteria. IHC enabled detection of the host cell markers E‐cadherin, Mucins 5AC (MUC5AC) and 2 (MUC2) using automated tissue staining for CG and GIM samples. Goblet cells, present only in GIM, secrete MUC2, whereas MUC5AC is found in both healthy and diseased gastric mucus. Representative images of GIM and CG samples are presented in Figure and Figure , respectively. Quantification of indicated markers are shown in Figure . Comparable levels of MUC5AC were observed across the patient groups and, as expected, there was a significant increase in MUC2 in GIM patients (Figure C&D). Consistent with previous studies , H. pylori exclusively colonized the gastric glands (Figure ). The presence of Eubacteria correlated with H. pylori infection in CG and GIM but were significantly reduced in the absence of H. pylori infection ( n = 13) (Figure ). This contrasts with the current dogma that there is a unique microbiota in H. pylori ‐ negative individuals . Interestingly, there was an increase in levels of Eubacteria in H. pylori ‐ positive GIM samples compared with H. pylori ‐ positive CG. Although this was not statistically significant, this raised the possibility that the presence of GIM‐specific MUC2 could provide intestinal‐like microniches within the stomach that promote bacterial colonization. As such, we observed colocalization of non‐ H. pylori bacteria with MUC2 in 2/10 H. pylori ‐ positive and 1/7 H. pylori ‐ negative GIM patients (Figure ).
Association of H. pylori With Invasion of Non‐ H. pylori Bacteria to the Lamina Propria Bacterial invasion is associated with multiple diseases of the gastrointestinal tract, causing damage to the epithelial architecture by triggering inflammatory responses . Whilst H. pylori colonized the gastric glands, non‐ H. pylori bacteria invaded the lamina propria in 6/9 patients with H. pylori ‐ associated CG and 8/10 H. pylori ‐ positive patients with GIM (Figure ). Qualitative scoring of Eubacterial invasion showed significantly more invasion in H. pylori ‐ positive CG and GIM than H. pylori ‐ negative CG and GIM (Figure ). Representative images of H. pylori ‐ positive CG and GIM patients with an invasion score of 3 are shown in Figure H. pylori ‐ negative patients with an invasion score of 0 can be seen in Figure . Very few Eubacteria are seen in samples with an invasion score of 0. It is most likely that these rare occurrences of Eubacteria in the stomach represent transient bacteria, rather than stable colonization . Invasion of H. pylori into the lamina propria is very rare across all patient samples, yet an instance of this can be seen in Figure . Representative images of samples with invasion scores of 1–2 can be seen in Figure .
Visualization of Non‐ H. pylori Bacteria Using a Modified Gram Stain Next, we aimed to confirm bacterial invasion at the cellular level in patients with H. pylori ‐ positive CG ( n = 4) or GIM ( n = 4). As can be seen in Figure , microniches of Gram‐positive bacteria could be identified in all 4 H. pylori ‐ positive GIM patients. Visualization of Gram‐negative bacteria with this technique was not clear and H&E‐stained sections were superior for identifying H. pylori (data not shown). Bacteria could not readily be identified in H. pylori ‐ positive CG patients using this technique (data not shown). This is consistent with the lower levels of non‐ H. pylori bacteria that were detected in CG patients in comparison to GIM patients (Figure ). Although further work is required to understand the biological consequences of invasive bacteria in carcinogenesis, this modified Gram stain technique could be used alongside the current histological diagnostic tools to stratify patients at higher risk for developing GAC.
Discussion Sequencing technologies have enabled characterization of the human microbiota in health and disease in multiple organs. However, applying 16S rRNA PCR‐based sequencing to low biomass samples, such as the skin and the stomach, can introduce sampling and technical errors in comparison to higher biomass samples and do not offer spatial resolution. Here, we have circumvented these issues by applying advanced imaging technologies to directly visualize the gastric microbiota in CG and GIM. By combining detection of targeted sequences within the bacterial 16S rRNA gene with immunostaining against host cell markers E‐cadherin, MUC5AC, and MUC2 we show that H. pylori exclusively occupies the gastric glandular niche, which is in agreement with previous studies . We have also shown the presence of significantly more non‐ H. pylori bacteria in H. pylori ‐ infected CG and GIM patients, whereas there are barely detectable levels in H. pylori ‐negative CG patients and only slightly more present in H. pylori ‐ negative GIM patients. This is in contrast to sequencing studies, which suggest a distinct microbiome exists in H. pylori ‐ negative disease states . These discrepancies may be due to many factors contributing to an over‐representation of the gastric microbiota using sequencing techniques, such as amplification‐based methods capturing bacterial DNA remnants, contamination issues and detection of transient rather than persistent bacteria . Indeed, this study highlights the crucial need for researchers to use a combination of sequencing , spatial profiling, and culture techniques to fully resolve the ecology of the gastric microbiota during health and disease. The most important observation made in this study was that non‐ H. pylori bacteria invaded the lamina propria in 67% of H. pylori ‐ positive CG and 80% of H. pylori ‐ positive GIM. The high prevalence of invasive bacteria in H. pylori ‐ positive patients amongst this cohort of patients was surprising. However, this could be explained by these patients being symptomatic and therefore attending for further clinical investigations. Our preliminary studies using intestinal‐type GAC tissue from gastrectomy samples highlighted that comparisons between CG and GIM gastric punch biopsy samples and gastrectomy samples are incredibly difficult to make with spatial biology approaches. The size, stage and molecular subtypes of GAC samples demands that three‐dimensional reconstructions of consecutive tissue sections are required to fully resolve intratumoral microbial communities. As such, these samples were excluded from this study until this comprehensive analysis of GAC samples can be made. The gastrointestinal tract provides an important barrier to pathogen invasion. Translocation of bacteria and their antigens/metabolites across the intestinal epithelial barrier is associated with gastrointestinal infections and a range of diseases, such as inflammatory bowel diseases and metabolic diseases . Activation of immune cells in the lamina propria drives pathology in these disorders via the production of pro‐inflammatory cytokines and reactive oxygen species. Thus, it is conceivable that invasive bacteria within the gastric lamina propria synergistically activate the immune response in H. pylori ‐ associated CG and GIM. Further work is underway to triangulate the immune response to invasive bacteria within the gastric mucosa. Given that these archived gastric tissue samples were retrieved via the local tissue bank (HBRC), we did not have access to the associated H. pylori clinical isolates from these patients. Recently, Sharafutdinov and colleagues reported that H. pylori strains encoding the trimeric form of the HtrA serine protease, which proteolytically cleaves the cell junction proteins occludin, claudin‐8 and E‐cadherin, are associated with a higher GAC risk than strains encoding the monomeric form . It is therefore conceivable that individuals infected with strains of H. pylori that secrete the trimeric HtrA cause the gastric epithelial cell barrier to become “leakier”, thereby facilitating invasion of bacteria to the lamina propria. However, it is also possible that H. pylori ‐ associated inflammation can cause disruption to the epithelial barrier by direct and indirect interactions between gastric epithelial cells and mucosal immune cells . Thus, we propose that H. pylori facilitates opportunistic invasion of the lamina propria by transiting microbes. Further work is required to compare bacterial invasion in patients that are infected with strains of H. pylori encoding the trimeric or monomeric HtrA. Unfortunately, we did not have access to healthy patients for endoscopy and so the microbial landscape in the healthy, H. pylori ‐ negative gastric mucosa is yet to be determined using imaging‐based approaches. Additionally, the prevalence and precise identity of invasive bacteria during the carcinogenic cascade must be determined in a larger cohort of patients. Interestingly, we have also shown that the levels of non‐ H. pylori bacteria increased from CG to GIM (Figure ) and that microniches of MUC2 and non‐ H. pylori bacteria colocalization are apparent in GIM tissue samples (Figure ). The region of non‐ H. pylori bacteria shown in Figure represents the largest single region of non‐ H. pylori bacteria detected amongst all patient samples. Although the levels of Eubacteria was relatively low in this patient (0.14%), it is conceivable that this microniche of bacterial colonization could be due to their use of the proton pump inhibitor, Rabeprazole (Table ). However, our relatively low number of patients with known PPI use (Table ) limits our ability to correlate the use of PPI with changes in the gastric microbiota. Nonetheless, our data provides the first observation of non‐ H. pylori bacteria with MUC2 within the gastric mucosa. Further work is required to understand whether these unique precancer microenvironments contribute to the estimated 1%–10% of GIM patients who progress to developing GAC . Limitations of this study were that 94% of patients were male, cases of autoimmune gastritis within the H. pylori ‐negative CG samples is unknown and PPI use is only known for a small number of patients within this cohort (Table ). Further studies should address these important considerations, particularly the use of proton pump inhibitors, when assessing the role of H. pylori and the microbiota in gastric cancer. Clinically, GIM has been termed the “point of no return”, given that antibiotic eradication of H. pylori in GIM provides only minimal benefit to a patient's risk of developing GAC in comparison to treatment of H. pylori ‐ associated CG . Additionally, patients with GIM are monitored for progression to GAC with 3‐yearly endoscopic surveillance due to a lack of convincing evidence on the use of biomarkers . Here, we show that a rapid and cost‐effective modified Gram stain could identify non‐ H. pylori bacteria in GIM, whilst it was less superior at visualizing H. pylori than H&E staining (data not shown). Therefore, further work could focus on optimizing the modified Gram stain to better visualize Gram‐negative bacteria, or, modified Gram staining in conjunction with H&E staining of consecutive tissue sections could be utilized for the diagnosis of invasive bacteria and H. pylori , respectively. Although future studies must determine whether invasive bacteria drive progression to GAC, it is possible that this represents a novel biomarker for GAC and that antibiotic treatment of GIM patients to eradicate both H. pylori and invasive bacteria could be implemented as a preventative treatment for GAC. In conclusion, we have observed that invasive bacteria are associated with H. pylori infection in the early and middle stages of Correa's carcinogenic cascade and therefore represent an attractive target for microbiome‐based interventions in the prevention, diagnosis, and management of GAC.
Conceptualization: C.D.S.‐L. and A.E.R.‐P. Methodology: H.J.G, A.T., J.J., J.L.M, K.H., R.A., I.G., W.B. and A.E.R.‐P. Data analysis: H.J.G., A.T., J.J., J.L.M., K.H., Z.A., I.R.H., J.A.C. and A.E.R.‐P. Writing: H.J.G., J.A.C. and A.E.R.‐P. Funding acquisition: A.E.R.‐P.
The authors declare no conflicts of interest.
Figure S1. Figure S2. Figure S3. Tables S1–S2.
|
Remote testing in Abbiategrasso (RTA): results from a counterbalanced cross-over study on direct-to-home neuropsychology with older adults | dbf16c94-5e5b-40d2-946d-20268cce8990 | 9884598 | Physiology[mh] | Remote neuropsychological assessments, via voice phone calls or videoconference (Teleneuropsychology—TNP), have become more and more common because of the SARS-CoV-2 pandemic, since geriatric populations, who need to undergo cognitive screenings and follow-up assessments, are at higher risk of developing SARS-CoV-2 severe complications . However, the interest toward this practice dates long before the pandemic considering that older populations often have mobility limitations and/or live in underserved rural areas . A recent meta-analysis including most of the TNP studies carried out before the pandemic demonstrated good reliability of brief global cognitive assessments [i.e., Mini Mental State Examination and Montreal Cognitive Assessment (MoCA)], letter fluency and digit span tests when administered in older populations with and without neurocognitive disorders . However, pre-pandemic TNP was generally carried out with some assistance (e.g., patient was provided with necessary devices either at the clinic or at home, or directly assisted in setting up personal devices). The pandemic forced instead to what has been called “direct-to-home-neuropsychology” (DTH-NP) , in which the neuropsychologist video-calls patients at their home, hence without being able to help in setting up the testing environment or solve any technical issues that might arise. Such a scenario raises concerns regarding the feasibility of DTH-NP in older adults, such as: limited access to technology, poor familiarity with it (e.g., knowledge of videoconference platforms, difficulties in setting up the equipment), the non-controlled environment during the in-home evaluation (e.g., presence of other people, sounds, and animals), the management of technical problems (e.g., connection speed, audio–video interruptions), a feeling of distance that may hinder the construction of a therapeutic alliance . These concerns have been addressed by few studies carried out during the pandemic, which showed good feasibility and acceptability of DTH-NP, when considering the point of view of both users in a wide age range and neuropsychologists . Even telephone-based cognitive assessment (TBCA), that can be easily implemented in a DTH-NP context, seems well accepted in adult populations: Caze et al. found that TBCA increased attendance rates, in particular because of the diminished cost of the service. According to Lacritz et al., TBCA also resulted in good patients’ satisfaction. As for statistical comparability between remote and face-to-face assessments, Parks et al. confirmed that auditory attention, verbal fluency, and verbal episodic memory tests delivered direct-to-home maintain their validity (i.e., allow to characterize specific cognitive profiles outlined by a previous face-to-face assessment). However, this study did not use a cross-over counterbalanced design, which was instead implemented to demonstrate the reliability of an unsupervised computerized battery of tests, but in younger adults . To our knowledge, no other studies used such a design to evaluate a battery of neuropsychological tests delivered via DTH-NP in older adults. As for the Italian context, so far only the Telephone Interview for Cognitive Status (TICS) and the Telephone‑based Frontal Assessment Battery (t-FAB) (respectively, a brief global cognitive assessment and an executive functions screening) showed satisfying psychometric properties (i.e., reliability, validity), and usability in clinical populations . To date, if also a more extensive face-to-face neuropsychological assessment proposed via videoconference is feasible, acceptable, and reliable for direct-to-home administration, remain to our knowledge under-investigated. The overall aim of the current study is, hence, to evaluate the feasibility, acceptability, and comparability of DTH-NP in an Italian population of over 65 without previously diagnosed major neurocognitive disorder by means of a randomized cross-over study in which each participant performed both a remote and face-to-face neuropsychological assessment in a counterbalanced order. Two remote modalities were included (phone call and videoconference) to obtain a reliable overview on the barriers that could be encountered in a sample of older adults living in Italy, where, in 2019, only 34% of families composed of over-65 individuals reported internet access . The first specific aim of this study is to describe the procedures for the implementation of DTH-NP, assess the recruitment rate of an older Italian population living in a small town in Northern Italy (Abbiategrasso) in a study on DTH-NP and report the issues encountered. The second specific aim is to evaluate the acceptability of DTH-NP. The third specific aim is to assess the comparability between remote and face-to-face neuropsychological assessment, as defined similarly to Backx et al. as the fulfillment of two requirements: (1) intraclass correlation coefficients (ICC) between scores ≥ 0.6 (considered as a “good” agreement), and (2) agreement between them as assessed through the Bland–Altman plots analysis. We expect that the recruitment rate will be higher for the phone call modality than for the videoconference one and that DTH-NP will be well accepted by our population. Also, we hypothesize that, in line with previous results , verbally mediated cognitive tests will show better comparability between remote and face-to-face administration. Study design This study has a within-subject counterbalanced cross-over design. All participants took part in both a remote (phone call or videoconference) and face-to-face neuropsychological assessment of the main cognitive domains, followed by an acceptability questionnaire. To reduce learning effect bias and intra-rater variability, administration modality has been counterbalanced as following: Group 1 (i.e., remote assessment first) and Group 2 (i.e., face-to-face assessment first). Each participant took part in the second session (remote or face-to-face) 8 weeks later. For the full description of the study protocol, see Vaccaro et al. . Participants Recruitment of participants is contextualized in the Abbiategrasso Brain Bank project (ABB) of the Golgi Cenci Foundation, in which brain donors undergo a periodic multidimensional assessment . During the pandemic, before vaccines’ availability, in-person assessments were interrupted. Thus, between February and March 2021, we performed a telephone-based survey aimed at collecting information on how ABB participants experienced last year. In this context, their willingness and availability to take part in a study on remote neuropsychological assessment (via phone call or videoconference) was explored. Apart from expressing positive feedback about participation in this study in either one of the two proposed modalities, inclusion criteria were age between 65 and 85 years and absence of a neurocognitive disorder (mild cognitive impairment or dementia) based on the last follow-up available (2018) and/or a recent memory decline with impact on daily life reported during the survey. Finally, participants scheduled for the 2021 donation program follow-up were excluded, since they had to perform a neuropsychological battery similar to the one included in the present study. The study was conducted in accordance with the Declaration of Helsinki, and it was approved by the Ethics Committee Milano Area 3 (ASST Grande Ospedale Metropolitano Niguarda) on June 9th, 2021 (approval number: 387–09,062,021). Written informed consent, explaining study objectives, information on the methods employed, the possible risks, and the storage and processing of collected data in accordance with the European General Data Protection Regulation (GDPR 2016/679), was obtained from all participants. Sample size calculation Based on Backx et al. , the criteria used for sample size calculation were (1) to obtain at least fair reliability ( ρ = 0.40) for the ICC and (2) an effect size of 0.40 at 0.80% of power in a one-sample t test, employed for testing if the bias (mean of the differences between scores obtained in the two modalities) was significantly different from zero (in this case, power calculation was performed under different distributions: normal, Laplace, and logistic). The sample size calculation estimated that a sample of 45 would be sufficient to provide adequate power for the intraclass correlation coefficient (ICC) and a sample size of 52 for the one-sample t test with a normal distribution, of 35 with a Laplace distribution and 47 with a logistic one. Since biases were normally distributed, the required sample size was of 52 (plus 6 participants to take into account the 10% expected dropout rate due to possible acute clinical conditions or withdrawal interfering with study completion). Procedures Study procedures are detailed in Vaccaro et al. . Briefly, a healthcare assistant called the pre-selected subjects ( N = 58) to receive confirmation of their participation in the study and to check the preferred and/or available remote modality (phone call or videoconference). To attend the videoconference assessment, participants needed a desktop computer or laptop equipped with a video camera and a stable internet connection (cable, Wi-Fi, or hotspot). After acceptance, Group 1 subjects (remote assessment first) were sent a home letter containing the disclosure note on the study, informed consent, and detailed instructions for the remote assessment (e.g., choose a quiet room, be alone in the testing room until they could clearly hear the experimenter, close any open programs on their PC, use earphones if possible, turn up the volume on their phone, install Zoom), and some numbered sheets on which to draw the stimuli (if in videoconference mode). Group 2 (face-to-face assessment first) received instead the information note and the informed consent on the day of the appointment at the Golgi Cenci Foundation; at the end of the session, they also received the instructions for the remote session and the numbered sheets on which to draw the stimuli, if in videoconference mode. On the day of the remote session, participants who chose the phone call assessment were called on their phone (either fixed or mobile). For subjects in the videoconference modality, a link to access the Zoom videoconference was sent by e-mail one day prior to the appointment. Following international guidelines , the professional version of the software Zoom was used to ensure privacy and security through the end-to-end data encryption. To ensure videoconference stability, internet connection speed was assessed prior to the testing session and considered appropriate for a bandwidth of 50–150 kbps, as advised in Zoom guidelines . For tests involving visual stimuli, the experimenter shared the screen to make stimuli visible to the subjects and took a screenshot of the paper on which the subject drew. The assessment, both remotely and face-to-face, was carried out in a quiet room of the Golgi Cenci Foundation by the same trained neuropsychologist (VA). Neuropsychological battery Table shows the tests employed in the study and the remote administration modality. Some tests were administered only to participants in the videoconference modality, since they required the use of visual stimuli. Global cognition was assessed by the MoCA 5 min protocol for phone call administration and the MoCA Audiovisual for videoconference administration . Verbal learning and long-term memory were evaluated through the Italian version of the Rey Auditory Verbal Learning Test (RAVLT), resulting in an Immediate Recall and Delayed Recall Score . In the second session, a parallel version was used to avoid a learning effect . Short-term memory and working memory were assessed by the Digit Span Forward and Backward, two subtests of the Wechsler Adult Intelligence Scale (WAIS) . Longest Digit Span was used as raw score to indicate the maximum length of the sequence correctly repeated. This test, with minimal verbal requirements, was delivered in the 15 min interval between the RAVLT Immediate and Recall, together with the Mental Alternation test . This latter task requires oral alternation of numbers and letters, and it was merely used to achieve the interference time required by the RAVLT; hence, performances were not evaluated. Spontaneous verbal production, strongly related to executive functions, processing speed and semantic memory was evaluated by the Alternate Phonemic–Semantic Fluency test , which consists of three subtests administered consecutively: letter-cued (phonemic) fluency, category-cued (semantic) fluency, and alternate phonemic/semantic fluency. The individual’s ability to reason (i.e., solve problems, draw inferences, and classify, based on previous own knowledge) was investigated by the Verbal Judgments test . Since a parallel form does not exist, the same version of the test was used in the second session. The Constructional Praxis test assesses visuospatial abilities. In this test, subjects are requested to copy bi-dimensional and three-dimensional figures on a paper sheet. To investigate social cognition, we adopted the Italian version of the Reading the Mind in the Eyes test (RME) . Acceptability questionnaire After the completion of the first assessment section, the psychologist administered a questionnaire evaluating levels of anxiety, enjoyment, engagement, willingness to repeat the battery and its overall duration, on a five-point Likert scale: 1, not at all; 2, a little bit; 3, neutral; 4, slightly; 5, very much . Furthermore, the general level of difficulty to perform the assessment was rated on the same five-point Likert scale at both sessions. At the end of the second session, participants were also asked to state their preferred modality and the reasons for their choice (“If we were to call you back again for a neuropsychological assessment, would you prefer to come to the Foundation, participate remotely, or no preference? Why?”). Statistical analysis All Statistical analyses were performed using SPSS Statistics 20.0 (IBM SPSS Statistics for Windows, version 20.0,BM Corp. Released 2011, Armonk, N.Y., USA), the open-source software R, version 4.1.2 (R Foundation, Released 2021, Vienna, Austria) and G*power 3 . The two-tailed significance level was set at 0.05 for all analyses. Mean values with standard deviation (SD) or median with interquartile range (IQR) were used to summarize the quantitative variables, while frequencies and percentages were used to describe the categorical variables. Group differences in socio-demographic variables, remote administration modalities, and the time elapsed between the two sessions were investigated with the Fisher’s exact test for categorical variables and the t test for continuous variables. Normality of data distributions was assessed by the Kolmogorov–Smirnov test to guide choice of statistical test. The recruitment rate was calculated as the percentage of randomly extracted subjects who agreed to participate in the study. An independent sample t test or the non-parametric corresponding test (Mann–Whitney U test) was employed to compare the responses to the acceptability questionnaire at the first session. Furthermore, the number of unexecuted or uncompleted tests and the perceived difficulty ratings were compared between the two modalities (remote vs face-to-face) with the paired t test or Wilcoxon signed rank test. Preference toward the modality of neuropsychological administration (remote, face-to-face or no preference) was compared with a χ 2 test. To assess the reliability of each neuropsychological test, the scores obtained in the remote and face-to-face sessions were correlated by means of the Intra-Class Correlation (ICC) (single-rating, absolute-agreement, two-way random-effects model). Test F was performed to test if the ICC was significantly different from zero. Following popular clinical criteria for interpreting reliability of a test by means of the ICC , we used “poor” to indicate an ICC below 0.40, “fair” between 0.40 and 0.59, “good” between 0.60 and 0.74, and “excellent” between 0.75 and 1.00. Based on Backx et al. , our reliability criterion was set at ICC ≥ 0.6. In addition, Bland–Altman plots were produced to further analyze the agreement between the two administration modalities and a t test was performed to assess if the bias (mean of the differences between scores obtained in the two modalities) was significantly different from zero. Due to the low number of participants who chose the videoconference modality and to disentangle the effects related to different remote modalities (phone call and videoconference), the ICC and the agreement were also obtained for phone call modality only. This study has a within-subject counterbalanced cross-over design. All participants took part in both a remote (phone call or videoconference) and face-to-face neuropsychological assessment of the main cognitive domains, followed by an acceptability questionnaire. To reduce learning effect bias and intra-rater variability, administration modality has been counterbalanced as following: Group 1 (i.e., remote assessment first) and Group 2 (i.e., face-to-face assessment first). Each participant took part in the second session (remote or face-to-face) 8 weeks later. For the full description of the study protocol, see Vaccaro et al. . Recruitment of participants is contextualized in the Abbiategrasso Brain Bank project (ABB) of the Golgi Cenci Foundation, in which brain donors undergo a periodic multidimensional assessment . During the pandemic, before vaccines’ availability, in-person assessments were interrupted. Thus, between February and March 2021, we performed a telephone-based survey aimed at collecting information on how ABB participants experienced last year. In this context, their willingness and availability to take part in a study on remote neuropsychological assessment (via phone call or videoconference) was explored. Apart from expressing positive feedback about participation in this study in either one of the two proposed modalities, inclusion criteria were age between 65 and 85 years and absence of a neurocognitive disorder (mild cognitive impairment or dementia) based on the last follow-up available (2018) and/or a recent memory decline with impact on daily life reported during the survey. Finally, participants scheduled for the 2021 donation program follow-up were excluded, since they had to perform a neuropsychological battery similar to the one included in the present study. The study was conducted in accordance with the Declaration of Helsinki, and it was approved by the Ethics Committee Milano Area 3 (ASST Grande Ospedale Metropolitano Niguarda) on June 9th, 2021 (approval number: 387–09,062,021). Written informed consent, explaining study objectives, information on the methods employed, the possible risks, and the storage and processing of collected data in accordance with the European General Data Protection Regulation (GDPR 2016/679), was obtained from all participants. Based on Backx et al. , the criteria used for sample size calculation were (1) to obtain at least fair reliability ( ρ = 0.40) for the ICC and (2) an effect size of 0.40 at 0.80% of power in a one-sample t test, employed for testing if the bias (mean of the differences between scores obtained in the two modalities) was significantly different from zero (in this case, power calculation was performed under different distributions: normal, Laplace, and logistic). The sample size calculation estimated that a sample of 45 would be sufficient to provide adequate power for the intraclass correlation coefficient (ICC) and a sample size of 52 for the one-sample t test with a normal distribution, of 35 with a Laplace distribution and 47 with a logistic one. Since biases were normally distributed, the required sample size was of 52 (plus 6 participants to take into account the 10% expected dropout rate due to possible acute clinical conditions or withdrawal interfering with study completion). Study procedures are detailed in Vaccaro et al. . Briefly, a healthcare assistant called the pre-selected subjects ( N = 58) to receive confirmation of their participation in the study and to check the preferred and/or available remote modality (phone call or videoconference). To attend the videoconference assessment, participants needed a desktop computer or laptop equipped with a video camera and a stable internet connection (cable, Wi-Fi, or hotspot). After acceptance, Group 1 subjects (remote assessment first) were sent a home letter containing the disclosure note on the study, informed consent, and detailed instructions for the remote assessment (e.g., choose a quiet room, be alone in the testing room until they could clearly hear the experimenter, close any open programs on their PC, use earphones if possible, turn up the volume on their phone, install Zoom), and some numbered sheets on which to draw the stimuli (if in videoconference mode). Group 2 (face-to-face assessment first) received instead the information note and the informed consent on the day of the appointment at the Golgi Cenci Foundation; at the end of the session, they also received the instructions for the remote session and the numbered sheets on which to draw the stimuli, if in videoconference mode. On the day of the remote session, participants who chose the phone call assessment were called on their phone (either fixed or mobile). For subjects in the videoconference modality, a link to access the Zoom videoconference was sent by e-mail one day prior to the appointment. Following international guidelines , the professional version of the software Zoom was used to ensure privacy and security through the end-to-end data encryption. To ensure videoconference stability, internet connection speed was assessed prior to the testing session and considered appropriate for a bandwidth of 50–150 kbps, as advised in Zoom guidelines . For tests involving visual stimuli, the experimenter shared the screen to make stimuli visible to the subjects and took a screenshot of the paper on which the subject drew. The assessment, both remotely and face-to-face, was carried out in a quiet room of the Golgi Cenci Foundation by the same trained neuropsychologist (VA). Table shows the tests employed in the study and the remote administration modality. Some tests were administered only to participants in the videoconference modality, since they required the use of visual stimuli. Global cognition was assessed by the MoCA 5 min protocol for phone call administration and the MoCA Audiovisual for videoconference administration . Verbal learning and long-term memory were evaluated through the Italian version of the Rey Auditory Verbal Learning Test (RAVLT), resulting in an Immediate Recall and Delayed Recall Score . In the second session, a parallel version was used to avoid a learning effect . Short-term memory and working memory were assessed by the Digit Span Forward and Backward, two subtests of the Wechsler Adult Intelligence Scale (WAIS) . Longest Digit Span was used as raw score to indicate the maximum length of the sequence correctly repeated. This test, with minimal verbal requirements, was delivered in the 15 min interval between the RAVLT Immediate and Recall, together with the Mental Alternation test . This latter task requires oral alternation of numbers and letters, and it was merely used to achieve the interference time required by the RAVLT; hence, performances were not evaluated. Spontaneous verbal production, strongly related to executive functions, processing speed and semantic memory was evaluated by the Alternate Phonemic–Semantic Fluency test , which consists of three subtests administered consecutively: letter-cued (phonemic) fluency, category-cued (semantic) fluency, and alternate phonemic/semantic fluency. The individual’s ability to reason (i.e., solve problems, draw inferences, and classify, based on previous own knowledge) was investigated by the Verbal Judgments test . Since a parallel form does not exist, the same version of the test was used in the second session. The Constructional Praxis test assesses visuospatial abilities. In this test, subjects are requested to copy bi-dimensional and three-dimensional figures on a paper sheet. To investigate social cognition, we adopted the Italian version of the Reading the Mind in the Eyes test (RME) . After the completion of the first assessment section, the psychologist administered a questionnaire evaluating levels of anxiety, enjoyment, engagement, willingness to repeat the battery and its overall duration, on a five-point Likert scale: 1, not at all; 2, a little bit; 3, neutral; 4, slightly; 5, very much . Furthermore, the general level of difficulty to perform the assessment was rated on the same five-point Likert scale at both sessions. At the end of the second session, participants were also asked to state their preferred modality and the reasons for their choice (“If we were to call you back again for a neuropsychological assessment, would you prefer to come to the Foundation, participate remotely, or no preference? Why?”). All Statistical analyses were performed using SPSS Statistics 20.0 (IBM SPSS Statistics for Windows, version 20.0,BM Corp. Released 2011, Armonk, N.Y., USA), the open-source software R, version 4.1.2 (R Foundation, Released 2021, Vienna, Austria) and G*power 3 . The two-tailed significance level was set at 0.05 for all analyses. Mean values with standard deviation (SD) or median with interquartile range (IQR) were used to summarize the quantitative variables, while frequencies and percentages were used to describe the categorical variables. Group differences in socio-demographic variables, remote administration modalities, and the time elapsed between the two sessions were investigated with the Fisher’s exact test for categorical variables and the t test for continuous variables. Normality of data distributions was assessed by the Kolmogorov–Smirnov test to guide choice of statistical test. The recruitment rate was calculated as the percentage of randomly extracted subjects who agreed to participate in the study. An independent sample t test or the non-parametric corresponding test (Mann–Whitney U test) was employed to compare the responses to the acceptability questionnaire at the first session. Furthermore, the number of unexecuted or uncompleted tests and the perceived difficulty ratings were compared between the two modalities (remote vs face-to-face) with the paired t test or Wilcoxon signed rank test. Preference toward the modality of neuropsychological administration (remote, face-to-face or no preference) was compared with a χ 2 test. To assess the reliability of each neuropsychological test, the scores obtained in the remote and face-to-face sessions were correlated by means of the Intra-Class Correlation (ICC) (single-rating, absolute-agreement, two-way random-effects model). Test F was performed to test if the ICC was significantly different from zero. Following popular clinical criteria for interpreting reliability of a test by means of the ICC , we used “poor” to indicate an ICC below 0.40, “fair” between 0.40 and 0.59, “good” between 0.60 and 0.74, and “excellent” between 0.75 and 1.00. Based on Backx et al. , our reliability criterion was set at ICC ≥ 0.6. In addition, Bland–Altman plots were produced to further analyze the agreement between the two administration modalities and a t test was performed to assess if the bias (mean of the differences between scores obtained in the two modalities) was significantly different from zero. Due to the low number of participants who chose the videoconference modality and to disentangle the effects related to different remote modalities (phone call and videoconference), the ICC and the agreement were also obtained for phone call modality only. Participants selection Among the surveyed participants ( N = 237), 93 ABB donors were eligible for the present study. In accordance with sample size calculation results, we randomly extracted 58 participants to be contacted, randomized in 2 groups of equal size ( N = 29): Group 1(DTH assessment first) and Group 2 (face-to-face assessment first). Figure shows the flow chart of the study, including the detailed reasons for exclusion. Feasibility Participants recruited in the present study were 47 out of 58, resulting in a recruitment rate of 81% (See Fig. ). The reasons for refusal were: not interested in participating in the study ( N = 3), health problems ( N = 5), transfer ( N = 1). Among recruited participants, only ten (21%) selected the videoconference modality. The main reason was they did not have the requested digital devices to perform the videoconference evaluation ( N = 31, 66%), while the rest of participants chose the phone call due to low self-confidence in using the requested devices on their own and/or unavailability of assistance by acquaintances. Among those in the videoconference modality, only one participant required at-home assistance by a family member to set up the videoconference before the assessment. As reported in Table , Group 1 and Group 2 were comparable for the main socio-demographic characteristics (sex, age, education) and for the remote administration modalities. All recruited participants completed the study, but two of them changed from videoconference to phone call modality between sessions (both in Group 2), resulting in eight subjects performing the tests planned for the videoconference modality in both sessions. Acceptability The Mann–Whitney test revealed no differences between the two groups on the variables of the battery acceptability questionnaire (anxiety, enjoyment, interest, happiness to repeat the test, perceived duration) at the first session (all p > 0.05; see Supplementary Table 1). There were no significant differences between the number of unexecuted or uncompleted neuropsychological tests ( Z = 1.26, p = 0.21) nor in the perceived difficulty between remote and face-to-face modality ( Z = 0.58, p = 0.56). More precisely, in the remote modality, there were a total of 14 non-executed or uncompleted tests (11 Alternate Phonemic–Semantic Fluency Tests, 2 RAVLT, and 1 Digit Span Backward), while in the face-to-face modality there were 17 non-executed or uncompleted tests (15 Alternate Phonemic–Semantic Fluency Tests and 2 RAVLT). Specifically, the reasons for non-execution or non-completion were: 1 rejection (RAVLT), 7 exhaustion/anxiety (6 Alternate Phonemic–Semantic Fluency Tests, 1 RAVLT), 22 objective difficulties in completing the task (20 Alternate Phonemic–Semantic Fluency Tests, 2 RAVLT) and 1 invalidated execution since the subject admitted having written down the stimuli (Digit Span Forward). Finally, there was a significant difference between the preferences for a future evaluation ( χ 2 (2, 47) = 20.47, p < 0.001) with 6 preferences for remote administration (12.8%), 30 for face-to-face (63.8%), and 11 with no preference (23.4%). Comparability According to Cicchetti’s interpretation , Phonemic Fluency, Semantic Fluency, and Verbal Judgments Tests showed excellent reliability; RAVLT Recall, Digit Span Forward, and Alternate Phonemic–Semantic Fluency Tests showed good reliability; the MoCA-5 min and RAVLT Immediate tests showed fair reliability. Only the Digit Span Backward test showed poor reliability (Table ). The reliability results for the phone call modality were similar: only the Semantic and Phonemic Fluency Tests showed good instead of excellent reliability (see Supplementary Table 2). Bland–Altman plots assessing agreement between the scores obtained in the remote and face-to-face modalities are showed in Fig. (See Supplementary Table 3 for bias analysis). Biases were not significantly different from zero (hence indicating agreement between scores obtained in the remote and face-to-face modalities) for all tests except for the RAVLT Immediate and Recall. In particular, performances on RAVLT Immediate and Recall were higher in the remote modality. When analyzing scores obtained in the phone call modality only, Bland–Altman results were comparable: biases were significantly different from zero only for the RAVLT (Immediate and Recall) (See Supplementary Table 4. Note that Bland–Altman plots were not reported for this analysis). ICC and Bland–Altman results for the MoCA Audiovisual (good reliability and agreement criterion satisfied), Constructional Praxis and RME tests (excellent reliability and agreement criterion satisfied) should be interpreted cautiously since they were performed by only eight participants. Among the surveyed participants ( N = 237), 93 ABB donors were eligible for the present study. In accordance with sample size calculation results, we randomly extracted 58 participants to be contacted, randomized in 2 groups of equal size ( N = 29): Group 1(DTH assessment first) and Group 2 (face-to-face assessment first). Figure shows the flow chart of the study, including the detailed reasons for exclusion. Participants recruited in the present study were 47 out of 58, resulting in a recruitment rate of 81% (See Fig. ). The reasons for refusal were: not interested in participating in the study ( N = 3), health problems ( N = 5), transfer ( N = 1). Among recruited participants, only ten (21%) selected the videoconference modality. The main reason was they did not have the requested digital devices to perform the videoconference evaluation ( N = 31, 66%), while the rest of participants chose the phone call due to low self-confidence in using the requested devices on their own and/or unavailability of assistance by acquaintances. Among those in the videoconference modality, only one participant required at-home assistance by a family member to set up the videoconference before the assessment. As reported in Table , Group 1 and Group 2 were comparable for the main socio-demographic characteristics (sex, age, education) and for the remote administration modalities. All recruited participants completed the study, but two of them changed from videoconference to phone call modality between sessions (both in Group 2), resulting in eight subjects performing the tests planned for the videoconference modality in both sessions. The Mann–Whitney test revealed no differences between the two groups on the variables of the battery acceptability questionnaire (anxiety, enjoyment, interest, happiness to repeat the test, perceived duration) at the first session (all p > 0.05; see Supplementary Table 1). There were no significant differences between the number of unexecuted or uncompleted neuropsychological tests ( Z = 1.26, p = 0.21) nor in the perceived difficulty between remote and face-to-face modality ( Z = 0.58, p = 0.56). More precisely, in the remote modality, there were a total of 14 non-executed or uncompleted tests (11 Alternate Phonemic–Semantic Fluency Tests, 2 RAVLT, and 1 Digit Span Backward), while in the face-to-face modality there were 17 non-executed or uncompleted tests (15 Alternate Phonemic–Semantic Fluency Tests and 2 RAVLT). Specifically, the reasons for non-execution or non-completion were: 1 rejection (RAVLT), 7 exhaustion/anxiety (6 Alternate Phonemic–Semantic Fluency Tests, 1 RAVLT), 22 objective difficulties in completing the task (20 Alternate Phonemic–Semantic Fluency Tests, 2 RAVLT) and 1 invalidated execution since the subject admitted having written down the stimuli (Digit Span Forward). Finally, there was a significant difference between the preferences for a future evaluation ( χ 2 (2, 47) = 20.47, p < 0.001) with 6 preferences for remote administration (12.8%), 30 for face-to-face (63.8%), and 11 with no preference (23.4%). According to Cicchetti’s interpretation , Phonemic Fluency, Semantic Fluency, and Verbal Judgments Tests showed excellent reliability; RAVLT Recall, Digit Span Forward, and Alternate Phonemic–Semantic Fluency Tests showed good reliability; the MoCA-5 min and RAVLT Immediate tests showed fair reliability. Only the Digit Span Backward test showed poor reliability (Table ). The reliability results for the phone call modality were similar: only the Semantic and Phonemic Fluency Tests showed good instead of excellent reliability (see Supplementary Table 2). Bland–Altman plots assessing agreement between the scores obtained in the remote and face-to-face modalities are showed in Fig. (See Supplementary Table 3 for bias analysis). Biases were not significantly different from zero (hence indicating agreement between scores obtained in the remote and face-to-face modalities) for all tests except for the RAVLT Immediate and Recall. In particular, performances on RAVLT Immediate and Recall were higher in the remote modality. When analyzing scores obtained in the phone call modality only, Bland–Altman results were comparable: biases were significantly different from zero only for the RAVLT (Immediate and Recall) (See Supplementary Table 4. Note that Bland–Altman plots were not reported for this analysis). ICC and Bland–Altman results for the MoCA Audiovisual (good reliability and agreement criterion satisfied), Constructional Praxis and RME tests (excellent reliability and agreement criterion satisfied) should be interpreted cautiously since they were performed by only eight participants. The Remote Testing in Abbiategrasso (RTA) is among the first counterbalanced cross-over study aimed at evaluating feasibility, acceptability, and comparability of DTH-NP in a sample of over 65 without previously diagnosed major neurocognitive disorder. The results here presented show (1) an optimal recruitment rate (81%) in a DTH-NP study involving community-dwelling Italian older adults (mean age ~ 80), especially for the phone call modality (79%), (2) good acceptability of DTH-NP delivered through phone call or videoconference, and (3) remote–face-to-face administration comparability of verbally mediated tests (Digit Span Forward, Semantic, Phonemic and Alternate Phonemic–Semantic Fluency test, Verbal Judgments test). During the Sars-Cov-2 pandemic, clinicians and researchers working in the neuropsychology field had to “make a virtue out of necessity”, by trying to offer good practices on DTH-NP . Procedures employed in the present study were conceived to overcome limitations of DTH-NP in line with the published guidelines . In particular, we gave preliminary instructions on how to reduce distractions and communication issues, we tested internet connection before starting the videoconference session and instructed participants to close all the unnecessary programs on their devices to prevent technical issues. The high recruitment and completion rate of older adults with a mean age of ~ 80 years in a DTH-NP study is encouraging and could be partly ascribed to the fact that study participants were volunteers of the ABB project, hence with a good understanding of our research activities . It should also be noted that the refusals received were ascribed to the impossibility of attending the face-to-face session and not to the remote modality per se. Furthermore, our study design, including two remote modalities (phone call and videoconference), allowed us to better understand if, to date, the spreading of DTH-NP with older adults could be realistic. The infrequent choice of the videoconference modality in the investigated population, explained mostly by the high percentage of subjects without access to the technologies requested for the videoconference modality (computer or laptop with a camera), is consistent with previous data showing that Italy is one of the countries with the lowest level of Information and Communication Technologies use among older adults in Europe . Since in Italy many funds will be allocated for reinforcing telemedicine interventions, providers willing to implement DTH-NP in the clinical context should take this result into consideration. As previously reported, in concomitance of our study, Aiello et al. demonstrated the validity and usability in clinical populations of two popular cognitive screening tests (TICS and FAB) administered by telephone, highlighting the interest toward TBCA in older adults . Our study, by showing the feasibility of a more complete (hence longer) battery of tests evaluating different cognitive functions delivered via phone call, further confirms the potentialities of TBCA in older populations, notwithstanding its intrinsic limitations (e.g., non-controlled environment and cheating tendencies) further discussed below. DTH-NP acceptability was supported by the finding of no significant differences in the number of uncompleted/invalid tests or in the difficulty perceived in the two modalities, suggesting that exhaustion, anxiety, or technical issues attributable to the DTH-NP modality (e.g., distractions, drop in line, fatigue) should not affect feasibility of remote cognitive assessment. Impressions on enjoyment, interest, and anxiety related to the battery of tests and on its duration were comparable between the two administration modalities, suggesting that the proposed tests were well-suited for remote administration, even if the Alternate Fluency test was particularly challenging for this age group (among the two modalities, there were, in fact, 26 occurrences in which it was not completed). However, most attendees reported preference for a future evaluation in the face-to-face modality. The percentage of subjects preferring this modality is higher than those reported in previous studies in older adults and could be due to the fact that all participants were residents in Abbiategrasso and voluntary donors of the ABB; therefore, the place of the face-to-face evaluation was both close and familiar to them. When asked about the reasons for the preference, participants frequently answered that after many months spent isolated, they were happy to go out and have a social exchange with the foundation personnel. We recognize that this result could be different in case of greater geographical distances or in clinical populations, e.g., with diagnosed neurocognitive disorders. Our results on comparability (here defined as good reliability and agreement between measures), between remote and face-to-face neuropsychological assessments are altogether in line with previous studies . The tests that met both our comparability criteria are the Digit Span Forward, Phonemic, Semantic and Alternate Phonemic–Semantic Fluency, and the Verbal Judgments test. This finding further confirms that even when the assessment is carried out direct-to-home, results obtained at verbally mediated tasks are comparable to those obtained face-to-face. In particular, the verbal tasks matching our comparability requirements could be included in a preliminary telephonic screening of memory, language, and executive functions, eventually followed by a more extensive face-to-face assessment in case of results worthy of clinical case attention. Also the three tests delivered in the videoconference modality matched the requirements for comparability, but this result suffers from poor generalization because of the small sample size. Further studies including larger samples could help clarifying the reliability and validity of the Constructional Praxis test, a test developed in the Italian context and widely used in clinical settings, and the RME test, which, to our knowledge, has not been included in other studies on DTH-NP. As for the MOCA Audiovisual test, there are previous evidences of its reliability and validity for remote administration . Other tests did not match both the requirements for comparability (reliability and agreement) between administration modalities. In particular, the MOCA 5-min test did not reach the threshold for good reliability, even though it was just below it, while the Digit Span Backward test showed poor reliability; however, both tests showed agreement between the two modalities. Poor reliability of the Digit Span Backward test could be explained by a “cheating” tendency at this test, since we did not explicitly indicate not to transcribe the stimuli (this decision was driven by the need to not make them feel distrusted, and to not point them toward this solution). There was in fact one subject during phone call administration that wrote down the stimuli, invalidating the score, and other two subjects suspected to have done so (for this test, a cheating attempt could be more evident to the clinician than for other tests). Since it is relatively easy to write down numbers on a paper, we consider that the remote administration of the Digit Span Backward, particularly challenging, could be more reliable in the videoconference modality. The RAVLT showed instead fair reliability for the immediate recall and good reliability for the delayed one, but the criterion for the agreement has not been met for either sub-score. Indeed, there was a systematic difference between scores, being those obtained in the remote modality significantly higher than in the face-to-face condition. There could be multiple explanations for this unexpected finding. In line with previous literature reporting a percentage of older subjects cheating at cognitive tests , we cannot again exclude that some participants wrote down some of the words while the experimenter listed them. Another possible explanation is that anxiety due to face-to-face administration could have had an impact on the performance, even if the anxiety levels measured through the acceptability questionnaire were not significantly different in the two modalities. However, it cannot be excluded that, since the question on anxiety was referred to the overall experience and not to the specific tests, such question was not able to reliably detect the anxiety associated to a test assessing episodic memory, particularly affected by personal beliefs on self-efficacy . We acknowledge that the present study has some limitations. First, the choice of healthy participants does not allow generalizing our results to other clinical populations, which most likely would need the help of a caregiver in handling the DTH assessment by phone call or videoconference. However, as previously reported, over 65 adults without diagnosed neurocognitive impairment could represent the population that mostly benefits from a DTH screening for preventive or diagnostic purposes . Future studies could investigate if these results apply to clinical populations as well. We also acknowledge that the possibility to choose between two remote modalities lead some participants with a videoconference device to choose the phone call modality because of higher familiarity with this means of communication, preventing us to obtain sufficient data for a reliable analysis of the tests administered via videoconference and for a comparison between the two remote modalities. At the same time, this aspect could be seen as a strength of our study, since, as previously pointed out, it allows to provide an unbiased picture of the readiness for telemedicine in Italian older adults. As for the acceptability assessment, the fact that the questionnaire was orally administered by the same psychologist that assessed participants could be pointed out as a limitation, since the answers could be driven by social desirability. Future studies could include, e.g., self-rating questionnaires for each proposed test (directly online or sent by mail before the assessment and collected in a second moment). Finally, future studies could confirm if these results could be replicated in larger sample sizes. The present counterbalanced cross-over study evaluated the feasibility and acceptability of DTH-NP in an Italian population of older adults without previously diagnosed neurocognitive disorder. Description of procedures for delivering a neuropsychological battery via phone call and videoconference while participants are in the natural home environment addressed the principal issues accounted for DTH-NP in the scientific literature. Moreover, remote–face-to-face comparability of selected domain-specific neuropsychological tests was demonstrated. Further studies might investigate, within the Italian context, the validity of DTH-NP in clinical settings. Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 29 KB) |
Does Ophthalmology Need Philosophy? | d1c30c3e-95ce-4382-9dd7-5661131eabd1 | 8558692 | Ophthalmology[mh] | Ophthalmology has shown significant progress and achievements, particularly in the last 20 to 30 years. Surgical incisions are made on a micron scale, and drugs that act against pathological vascularization are providing more “successful” results in incurable diseases. Societal awareness of eye health has increased, and technological products for diagnosis and treatment have become widespread and more accessible. Thus, in many parts of the world, ophthalmology utilizes these advanced capabilities to reduce vision loss and improve people’s quality of life. Although science in general and ophthalmology in particular have made major progress, when examined objectively, one can recognize aspects of both that can be improved. , , , , One means of improving these aspects is a philosophical approach that subjects some established thoughts and behaviors to more rigorous examination and inquiry. , , , , , , Philosophy is not an “ivory tower” activity that quotes important philosophers, deals only with the theoretical realm, and is carried out with complex words. Although unnoticed, philosophy is an activity that positively or negatively affects life in many ways and determines the basic mental processes that guide life. Ophthalmology practices, like all activities of life, are shaped by some fundamental philosophical approaches. This article provides an introduction to philosophy and the areas in which philosophy and ophthalmology interact. Philosophy in Turkish means “love of wisdom,” and we will discuss the connection and relationship between ophthalmology and philosophy based on the premise that the main purpose of philosophy is to “acquire wisdom.” We use the concept of “wisdom” within the scope and historical meaning of the Turkish language, not implying a “mystical” wisdom of Far East or similar origin. We use a plain language, avoiding expressions and terms that are not used by people outside the field of academic philosophy, such as “geist,” “phenomenology,” and “transcendental.” This article will attempt to raise the call to carefully reexamine ophthalmology, which has historical ties and interaction with philosophy, within the conditions of our language (Turkish) and country.
As is widely known, “philosophy” is formed from the words philo and sophia and means “love of wisdom.” In the Turkish language, the concept of wisdom ( bilgelik ) is based on the root for “know” ( bil -) and has a common origin with words such as knowledge ( bilgi ), science ( bilim ), scientific ( bilimsel ), and consciousness ( bilinç ). “Know” ( bil -) in our language is also used in the sense of “having the power, skill, and ability to do” in the Turkish words for “able to do” ( yapa-bilmek ), “able to see” ( göre-bilmek ), and “able to know” ( bile-bilmek ). Considering the dictionary definitions, “wise” ( bilge ) can be defined as a person who has comprehensive knowledge and can use their knowledge correctly and beneficially, and “wisdom” can be defined as the state of evaluating what transpires with a superiority born of virtue and knowledge. Essentially, wisdom can be described as “to comprehensively know and be aware; to implement in a correct, beneficial, and virtuous way.”
Logic is among the essentials of philosophy and works on the principles of correct reasoning. Logic defines the methods and rules by which experienced realities and mental processes can be conveyed through words (or symbols) in a way that other minds can understand and process. Although there are different definitions, in our opinion, logic structurally examines the processes involved in using words (and symbols) connected to thoughts to enable realities to be conveyed in accordance with actuality. The most important logic topic that concerns ophthalmologists is logical fallacies. Logical fallacies are erroneous thought processes that occur unconsciously. Although ophthalmology is largely carried out through rigorous cognitive processes, logical fallacies can mislead patients and ophthalmologists just as everyone else. Publications on logical fallacies and how to reduce them are also found in medical literature. , , In this article we address several important and common fallacies concerning the field of ophthalmology. A common fallacy in life and ophthalmology is called post hoc ergo propter hoc in Latin, or the “post hoc fallacy” for short. This fallacy can be described as the assumption that unrelated events are connected because they occur in temporal proximity (one after the other). It manifests as an erroneous causal relationship drawn between events with very low or no probability of connection. It can be regarded as essentially an extension of the problem of induction in philosophy. As an example related to ophthalmology, if a person with atopic tendency and complaints of intense eye itching, redness, and off-white discharge uses antibiotic drops at home, leaves the city, and then attributes the cessation of their symptoms to the antibiotic, this is an example of post hoc fallacy. Their symptoms likely resolved because they moved away from an allergen, but they think the antibiotic cured their condition. Another example is to assume that intravitreal injections alone improved or worsened a patient’s condition, without adequately considering some other important factors, such as blood glucose regulation. To believe that the intravitreal drug is definitively effective or ineffective carries the possibility of post hoc fallacy, because the presumption of (in)efficacy is being made in a multifactorial clinical condition based on only one variable, without a comprehensive evaluation of the causal relationship. Statements such as “drug A is effective and safe in eye disease B” that bear truth/falsity and provision/judgment values are called propositions. When a phenomenon inconsistent with this proposition is observed, from a philosophical standpoint this proposition is no longer as strong as it was; in a sense, it is “refuted.” In this case, new scientific observations and studies are conducted in an attempt to gain a more comprehensive understanding of the subject of the proposition. A new study shows that drug A is more effective in patients with intraocular pressure of 20-25 mmHg. Therefore, the proposition becomes “drug A is effective and safe in eye disease B when intraocular pressure is between 20 and 25 mmHg.” Now more is known about disease B and drug A and there is more comprehensive knowledge of which patient group drug A will be effective in. The hypothetical phenomenon given here is an example of the “thought experiment” concept in philosophy. Thought experiments aim to scrutinize reality within the framework of existing information, according to reason and logic but with imaginary/hypothetical situations. This thought experiment is an example of the process of better understanding the incompletely understood disease B, creating a more correct approach, and avoiding post hoc fallacy. Post hoc fallacy can be seen in some patients in examples such as “my head hurts, my intraocular pressure is high,” or in exfoliative zonular weakness, “surgery was performed incorrectly, my lens shifted.” Headache may be associated with intraocular pressure in a group of patients; however, if intraocular pressure is implicated when the headache was actually caused by a factor such as tension or stress, then a post hoc fallacy was committed. Similarly, a patient with advanced exfoliation who undergoes normal cataract surgery and later attributes intraocular lens dislocation due to severe zonular weakness as “incorrect surgery” is another example of post hoc fallacy. The capsule is no longer adequately supported due to structural alterations in the patient’s eye, yet the patient believes the surgery was performed incorrectly. The problem with the post hoc fallacy is that it is difficult to determine whether successive events are truly connected, i.e., to determine causality. Even randomized controlled studies cannot fully solve this problem; the complexity of the human organism makes it difficult to reach the truth. , “Confounding” and “bias,” which exist in medicine and the nature of life, also create challenges in identifying causal relationships. Although this logical fallacy can be overcome to some extent through more careful observations and interpretations based on a better scientific method, it is an important problem of science and philosophy that has yet to be solved. Ophthalmologists can reduce the frequency of post hoc fallacy by being more aware of the fact that successive events can also occur by chance. Apart from the post hoc fallacy, some habits in the medical field may also lead to erroneous thinking and decision-making by physicians. The habits of appealing to authority and appealing to convention are also common fallacies. These are examples of logical fallacies in that well-known people may not always show the right path, or the majority may be misguided. , , , In the case of ophthalmology, it should also be borne in mind that despite being published in reputable journals, study results may be biased due to factors such as academic or financial concerns. Using publications with a high citation index and a practical orientation, as well as checking the accuracy of information related to the physicians’ workplace, hospital conditions, and patient group may help prevent these logical fallacies. , , , , Varner reported that there are problems in the ophthalmology literature regarding issues such as study validity and bias, patient selection and eligibility, compliance with standards of comparison, insufficient patient numbers, lack of comparison to the gold standard or placebo, confounders, and a lack of clear research objectives. An important part of logic studies is the branch of propositional logic. When many statements used in ophthalmology are examined within the framework of propositional logic, one can gain a more in-depth perspective in the diagnosis, treatment, and follow-up stages. For example, the expression “this person has glaucoma” is perceived as a true, clear, and understandable proposition by ophthalmologists. However, ophthalmologists being acquainted with subjects like whether a judgment is accepted as “true” because it is “concordant with the facts” or because it is “consistent with all other propositions of the system to which that proposition belongs” may contribute to more solid foundations of ophthalmological knowledge.
One means by which physicians can achieve wisdom is to scrutinize basic definitions and concepts that influence their thinking and practices, such as “disease,” “health,” “therapy,” “healing,” “innovation,” or “the latest treatment.” Although this area is considered to be related to the branch of ontology, which is translated into Turkish as “the philosophy of existence,” it also falls into the domain of epistemology and aims to provide a better understanding of the nature of reality. The reason ontology concerns ophthalmologists is that basic definitions and concepts influence their ways of thinking that lead practices in that field of knowledge. For example, the phrase “complete well-being” in the World Health Organization’s definition of health indicates a very high level of well-being and creates a goal that is difficult to achieve in real life. Another example of the importance of definitions and concepts for physicians and patients is statements such as “the latest treatment” or “innovation.” While such words can be presented as a hope and cure for the patient, they also carry meanings such as “treatment whose effects and side effects are not yet fully known.” Ontology examines words’ mental correlates in real life, thereby enriching perceptions and understanding of the subject and contributing to wisdom. Although philosophy scholars and philosophers examine such basic concepts in theoretical terms, physicians can make more realistic contributions to these examinations and explanations from real life.
Along with ontology, another important branch of philosophy is epistemology, or the philosophy of knowledge. Epistemology is defined as “a general reckoning with knowledge,” and it leads philosophical discussions such as “the nature of knowledge and justification” and “the position/attitude of skepticism.” Knowledge is defined as “justified true belief,” and valid and adequate indications that a proposition is true are accepted as evidence. Epistemology is a branch of philosophy that adopts a measured skepticism and seeks answers to questions such as “What is true knowledge?”, “What factors make knowledge true?”, and “Is the information given by people known as authorities always reliable?” Epistemology is one of the most fundamental areas of philosophy and deals with the “having true and comprehensive knowledge” aspect of wisdom. Seeking an answer to the question “Does industry funding influence research results?” is in fact an epistemological pursuit. A more detailed form of epistemology is the philosophy of science, which subjects scientific thought and practices to philosophical scrutiny and inquiry.
Ophthalmology, which is actually a branch of science, is most connected to philosophy through the philosophy of science. The philosophy of science is concerned with more closely examining, questioning, and understanding the procedures and processes called “scientific activity.” It has been stated that in traditional education, there is a “missing link” between science and philosophy and that philosophy’s contribution to science is of no interest whatsoever to scientists. Philosophy of science aims to contribute to many questions such as “What is science and its purpose?”, “What properties distinguish scientific knowledge from other types of knowledge?”, “What is scientific explanation?”, and “Under what conditions is science useful?” Ophthalmologists can examine and review their activities as science practitioners within the framework of the philosophy of science. This examination and review process may allow ophthalmology practices to further mature and be more open to development. Although many scientists and philosophers have contributed to the philosophy of science, Karl Popper and Thomas Kuhn in particular made significant changes in perspectives of science, the impact of which persist even today. Therefore, we will briefly discuss some ideas of these two science philosophers. Popper made the concept of “falsifiability” central to science. According to this idea, the distinction between scientific and non-scientific information is whether it can be falsified. Information that cannot be tested experimentally and falsified by the scientific method is not considered scientific, but is relegated to the realm of pseudoscience. Karl Popper’s concept of falsifiability, which promotes scientific skepticism, also offers ophthalmologists an important approach and useful research style. According to this research style, if ophthalmology knowledge and practices can be falsified by an experiment or observation, that knowledge and practice is scientific. For example, the proposition that “elevated intraocular pressure damages the optic nerve” is considered scientific because it can be confirmed or shown to be false by experiment or observation. It is observed that people with glaucomatous damage have high intraocular pressure, and it is understood that intraocular pressure damages the optic nerve. However, as time progresses and observations increase, the observation of a person with glaucomatous damage who does not have high intraocular pressure indicates a fault in the proposition “high intraocular pressure damages the optic nerve” and it becomes clear that another explanation for glaucomatous damage is needed. Thus, the explanation of low tension glaucoma emerges and glaucoma is better understood. An important point learned from Popper is that findings contrary to established knowledge and general belief should not be disregarded, because they will contribute to a better understanding of medical truths. Applying the falsifiability principle in daily life exposes the errors and fallacies of general beliefs and thoughts and allows them to be corrected and strengthened. Without a skeptical approach based on the falsifiability principle, ophthalmologists would probably still be diagnosing and monitoring glaucoma with Schiotz tonometry. Demonstrating the shortcomings of this device enabled follow-up and treatment to be performed using better methods. When current methods are also shown to be flawed, it will immediately open the way for more useful diagnostic and therapeutic methods for nearly all eye diseases. By means of the falsifiability principle, findings that falsify established practices are given more attention, theories and explanations are closer to the truth, and practices are improved. The scientific philosopher Kuhn introduced the concept of “paradigm shift,” which explains how scientific revolutions, or major changes in scientific understanding, have occurred throughout history. This explanation rejects the view that science is an activity that evolves and is perfected by the gradual accumulation of knowledge over time. According to Kuhn, people in a profession group, with the influence of their professional perspectives, develop scientific propositions (i.e., “paradigms”) that explain events within a certain framework. Although these scientific propositions do not always reflect the most accurate and truthful information, those within the group perceive them as true knowledge. Over time, however, new findings reveal important flaws in the existing paradigm, and a new paradigm is developed to explain the situation. An example of paradigm shift in ophthalmology is the transition from explaining glaucomatous damage by the mechanical effect of intraocular pressure to the explanation of vascular autoregulation, and even the transition to considering it an eye disease related to systemic neurodegeneration. The concept of paradigm shift draws attention to the fact that established ideas are understood and explained with the existing level of knowledge and that these truths can change with new information. Instead of assuming medical findings that contradict the general view are errors or inadequate observations, seeing them as an opportunity to improve the general view can initiate large-scale changes. In terms of ophthalmology, Kuhn’s major contribution is that existing knowledge is considered “valid according to the present understanding” and that more comprehensive understanding and perceptions of reality can be achieved through new findings and new perceptions. From Kuhn’s perspective, the attitude that will further advance ophthalmology is not research that replicates and confirms established knowledge, but adopting an approach that encourages development and change by demonstrating deficiencies in the current understanding. Popper and Kuhn have made some important contributions to the perceptions and application of science and the scientific method. In our opinion, the most important contributions of these two science philosophers are that they draw attention to the need for existing knowledge to nearly always be open to inquiry and even challenge. Ophthalmologists may be inclined to consider findings that are inconsistent with general knowledge and understanding as incomplete or inaccurate observations. Ophthalmology journals, like all journals, can fall prey to publication (or non-publication) bias, particularly toward articles stating that drugs and treatments are not effective. , , Publication bias in the field ophthalmology can be observed as a higher publication rate of studies with positive results, i.e., showing that there are benefits of treatment, especially in journals with a high impact factor. This suggests that studies showing that drugs and treatments are ineffective are less likely to be published, especially in high-impact journals. In addition, misconduct by those regarded as authority is met by silence due to the culture of respect for elders in the profession, which has persisted from the Hippocratic Oath to the present day. Although it is important to preserve ophthalmological traditions, which are an extension of our country’s culture, measured and logical objections to established inadequate practices can help ophthalmology advance in the right direction. An important feature of science and ophthalmology is the different approaches to science in countries or institutions that “produce knowledge” and those that “use knowledge.” Although the scientific method has the same standards, there may be differences among individuals and institutions that produce scientific knowledge and those who transfer and use it. The people, institutions, and countries that produce knowledge “promote” the scientific product with the inherent aim of ultimately profiting from it. For this promotion, inadequacies and flaws of a method may be overlooked while so-called “scientific” methods are used to convince others of its superiority. So-called “scientific” studies can also be seen in research and knowledge-generation processes for reasons such as academic promotion, recognition, and industry affiliations. , By seeking answers to questions such as “What are its inadequacies and advantages?” and “Does it contribute significantly to clinical practice?”, the users of scientific information and technological products can more accurately evaluate scientific products. This way of thinking contributes to a more comprehensive understanding of reality and wisdom through questions like “What is true knowledge?” and “Who benefits from this information?”, which are actually among the fundamental questions of epistemology.
The area of greatest interaction between philosophy and medicine is the field of deontology/ethics/moral philosophy. Changes in the last few decades have resulted in a silent shift from the concept of “deontology” to the concept of “ethics” in medical education and practice. As ethics, derived from “ethos,” is perceived as an area that more encompasses professional rules, we prefer the more comprehensive term “moral philosophy” in this article. Moral philosophy is the field of philosophy that discusses the morality of thoughts and behaviors through questions such as “What is the right behavior?”, “What is virtuous behavior?”, and “What makes a behavior moral?” In addition to big problems in the field of academic philosophy such as “Can there be moral standards other than religious edicts?”, moral philosophy can also be used for other everyday life problems. Frank discussions of questions such as “What boundaries make industrial relationships with physicians moral?”, “Is it morally appropriate to present a medical practice as a new treatment while in the research stage?”, and “Can a revenue/performance-based pricing system negatively affect the principle of doing no harm?” are also included in the field of moral philosophy. In addition, the philosophical and moral examination of the concepts of health law cited in malpractice claims such as “failure to inform,” “strict liability,” and “professional inexperience” also warrants philosophical inquiry in terms of expressing the physician’s viewpoint. It has been stated that for the art of medicine to be performed with decency, it must be determined not only by technical rules but also by medical ethics, and many criticisms of medicine arise from the patient feeling that they have been subjected to excessive and unnecessary interventions. Excessive medicalization has been called a real danger in many countries due to situations that can be described as the abuse of drugs and medicine. Biomedical ethics is also expected to answer the questions of what to do, what not to do, and how to solve problems encountered while conducting research or practicing the profession. Moreover, it becomes a moral imperative to test practices that are currently being presented as scientific, such as leeching, ozone therapy, homeopathy, and acupuncture, according to real scientific standards, and only allow practices that are not of “pseudoscience” status. Such trials by philosophical inquiry contribute to a deeper understanding of medical practices and the enhancement of their morality. The field of medical ethics has been shaped by the concept of “bioethics” since the 1970s, largely due to the contributions of philosophy and medical history scholars. Bioethics is a theoretical field of study that has contributed significantly to the strengthening of human rights in the field of medicine and health and to making medical practices more humane. Today, bioethical principles are used as a moral norm/standard in a wide range of areas, from health law to medical research. The Turkish Ophthalmological Association adopts the Professional Ethics Guide for Ophthalmologists: Ethical Principles and Professional Principles and determines the ethical principles of ophthalmological activities in our country. In addition, the Turkish Medical Association Professional Ethics Code is shaped by the four principles of bioethics. While some additions may be made, the core bioethical principles of beneficence, non-maleficence, justice, and autonomy have remained strong over the years. In fact, although ophthalmologists may not realize it, a significant proportion of everyday practices are shaped by these four principles of bioethics. For example, practices such as informed consent are carried out within the scope of the bioethical principles of patients being autonomous/self-governing and the procedure being beneficial for the patient. Although such bioethical practices create some difficulties for physicians and health institutions, they ensure the implementation of many practices that are for the patient’s benefit. While bioethics has made significant contributions to more humane medical practices, this field must also be subjected to philosophical inquiry by physicians, i.e., by those who apply these principles in practice. When the literature is examined, there are many articles that regard bioethical principles positively, as well as criticisms that these principles reflect the traditions of Western moral philosophy, politics, and social theory and are even a tool of moral imperialism. , Taking into account these and similar criticisms, the field of bioethics is due for a philosophical examination using the questioning tools of moral philosophy. Ophthalmologists, as people who are living and observing in real life, can make important contributions to the theoretical field of bioethics. Some moral standards developed at an office desk may not be compatible with the realities of daily life. Some well-intentioned theoretical practices can turn into impositions that strain the human dignity of physicians. Under the guise of actualizing bioethical norms, physicians may be exposed to practices outside the norms of human rights. Ophthalmologists should also try to utilize philosophical inquiry for support in the scientific and moral criticism of the legal norms associated with bioethical principles and to avoid being subject to undignified allegations of malpractice.
Science and philosophy had nearly the same meaning historically but have been divergent for several centuries, and today the connections between science and philosophy are rather obscure. For thousands of years, philosophy has involved thinking and producing written works on various subjects, whereas science has used mental abilities to innovate practices to make life better. Over time, the ties between science and philosophy have weakened; philosophy remained purely a field of intellectual production, while science continued on to become a field of intensive technology production and use, but limited in terms of inquiry. Even if ophthalmologists are not interested in philosophy, just asking the question “Is it possible that what is said may be untrue?” will constitute the first stage of wisdom. Philosophy can contribute to an ophthalmology practice that is firmly grounded and consistent with how physicians want themselves and their families treated, not swayed by the researchers, authors, and opinion leaders (although rare) who abuse the drug industry’s support. For such medical practices, there may be important benefits to revisiting and reconstructing both the “logos” and “ethos” areas of ophthalmology (i.e., ophthalmo-logos). A philosophical attitude that pursues wisdom makes an important contribution to more accurately observing, thinking about, and interpreting one’s experiences. This may enable a more comprehensive and sound evaluation of professional practices. The essence of the professional activity performed can be better recognized and understood. Philosophy provides individuals and the profession with valuable intellectual abilities and tools applicable in a broad range of contexts, from advancing ophthalmology research to defending against malpractice claims. For these reasons, ophthalmologists need philosophical activity and the wisdom they can gain from it.
|
Early Prediction of Fetal Macrosomia Through Maternal Lipid Profiles | f841938d-88ad-45a0-aa24-dcf3b8fdf074 | 11818448 | Biochemistry[mh] | The prevalence of metabolic disorders among pregnant women worldwide is steadily increasing, significantly varying based on factors such as dietary behaviors, cultural habits, average living standards in developed and developing countries, and maternal age. Annually, around 39 million pregnancies occur amidst maternal obesity, and, in some countries, its prevalence exceeds 60% . Maternal overweight and obesity are independent risk factors for the development of gestational diabetes mellitus (GDM). The overall global standardized prevalence of GDM is 14.0% (95% CI: 13.97–14.04%). The prevalence varies across regions: North America and the Caribbean—7.1%, Europe—7.8%, South and Central America—10.4%, Africa—14.2%, the Western Pacific—14.7%, Southeast Asia—20.8%, and the Middle East and North Africa—27.6% . As of 1 January 2019, in the Russian Federation, 4.58 million people (3.1% of the population) were registered with diabetes. Of these, 92% (4.2 million) had type 2 diabetes, 6% (256,000) had type 1 diabetes, and 2% (90,000) had other types of diabetes, including 8006 cases of gestational diabetes . Fetal macrosomia is diagnosed three times more often in children born to women with GDM than in children born to mothers with normal blood glucose levels. However, a trend toward excessive fetal growth can also occur in women without carbohydrate metabolism disorders. Known non-modifiable risk factors include ethnicity, advanced maternal age (>35 years), multiparity, family history of diabetes or macrosomia, and male infants ( p < 0.001) . The risk of perinatal loss in fetal macrosomia is higher than in children with normal birth weight . In pregnancies complicated by diabetes, fetuses with macrosomia develop a unique pattern of excessive growth characterized by central subcutaneous fat deposition around the abdomen and between the shoulder blades. This results in an increase in the shoulder circumference relative to the head, which, in 5–9% of cases, significantly increases the risk of shoulder dystocia, Erb’s palsy, brachial plexus injuries, fractures of tubular bones, and neonatal asphyxia . For the mother, childbirth may be accompanied by complications such as prolonged labor and operative vaginal delivery or may result in a cesarean section. Postpartum complications may include trauma to the birth canal and uterine atony with an increased risk of postpartum hemorrhage . Neonatal morbidity rates are higher among newborns with birth weights greater than 4000 g. Fetal macrosomia is associated with electrolyte and metabolic disturbances in the early neonatal period. For example, the incidence of severe hypoglycemia and neonatal hyperbilirubinemia is 5 and 2 times higher, respectively, compared to infants born to healthy mothers . There are several theories regarding the pathogenesis of fetal macrosomia. The pathogenesis of fetal macrosomia is based on disturbances in carbohydrate and lipid metabolism within the “fetus–maternal” complex. These disruptions lead to excess glucose and lipids reaching the fetus, which contributes to excessive growth. Maternal hyperglycemia results in fetal hyperinsulinemia and an increase in fetal adipose tissue (Pedersen’s hypothesis) . Changes in blood lipid profiles in pregnant women with gestational diabetes and macrosomia are currently a widely studied topic, though comprehensive and systematic data are yet to be established. A study on the serum lipid composition of healthy pregnant women in the first trimester showed that those with macrosomia had lower levels of lipids (phospholipids, lysophospholipids, monoacylglycerides) compared to women without macrosomia . In mothers with metabolic disorders, the transplacental transfer of lipids is altered, which promotes accelerated differentiation of adipocytes in the fetus and leads to fetal macrosomia, excess weight, and obesity in children . Transported fatty acids are thought to initiate the conversion of mesenchymal stem cells into adipocytes by activating transcription factors in the fetus. Excessive adipocyte formation leads to the development of excess weight. The source of these transferred lipids includes “free” fatty acids, triglyceride hydrolysis products, and components of the mother’s phospholipid polyunsaturated fatty acids . Most of the fatty acids required by the fetus must be produced de novo, with approximately 20% being transferable through the placenta . In pregnant women with diabetes, excess weight, and obesity, prolonged high plasma lipid concentrations (hyperlipidemia) result in excessive lipid transfer across the placenta. The fetus’s subcutaneous adipose tissue can synthesize fatty acids stimulated by insulin starting from the 10th week. The initial differentiation of adipocytes occurs between the 14th and 24th weeks of pregnancy . During this period, the primary transplacental transfer of fatty acids takes place, and their increased intake is critical for the excessive formation of adipocytes in the fetus, leading to the birth of a large infant and the development of metabolic complications in the future. Thus, the exponential growth of adipose tissue in the fetus occurs simultaneously with the rise of lipids in the mother’s blood. However, the mechanisms behind the development of fetal macrosomia and its relationship with maternal hyperlipidemia constitute a complex, multifactorial process that requires further in-depth analysis. Additionally, high lipid levels in the mother’s blood correlate with maternal hypertension and preeclampsia . A key challenge remains the development of early methods for predicting fetal macrosomia to reduce the occurrence of this complication both among pregnant women with metabolic disorders such as gestational diabetes and obesity and women without these risk factors. The aim of this study was to develop a system for predicting fetal macrosomia based on the lipidomic profiles of pregnant women’s blood serum.
2.1. Analysis of Clinical Characteristics A comparative analysis of the clinical characteristics of the study groups was conducted . The groups were designated as GDM+ (the main group) and GDM− (the comparative group) for patients with and without GDM, respectively. The subgroups within each group were designated as Macrosomia+ and Macrosomia− for patients with and without macrosomia. The results of the analysis between the groups of women examined are presented in . The analysis of the main and comparative groups did not reveal significant differences in the age and anthropometric data of the patients before pregnancy. The median age of the patients did not differ significantly between the groups and subgroups, with an average age of 32 years ( p = 0.56). However, within-group analysis showed a difference between the Macrosomia+ and Macrosomia− subgroups. Pregnant women of the Macrosomia+ subgroups had a significantly higher pre-pregnancy body weight, especially those with GDM (84 (66;96), p = 0.006). Pre-pregnancy BMI significantly differed between the main GDM+ (22.6 (20.1; 26.5)) and comparative GDM− groups (21.2 (19.5; 22.9)) with p = 0.03). Patients with gestational diabetes and macrosomia started pregnancy with a significantly higher BMI compared to patients in other subgroups (27 (23; 30), p = 0.006). Weight gain by the time of delivery differed significantly among the patients with macrosomia, regardless of the presence of gestational diabetes (14 (11; 16) and 16 (13; 18), p = 0.006). Those with fetal macrosomia without gestational diabetes had the greatest weight gain by delivery. Additionally, women in this subgroup had a significantly higher birth weight (3.8 (3.5; 4.1), p = 0.004) compared to other patients. It is likely that genetic and behavioral factors like family diet and eating habits may equally contribute to the development of macrosomia in women without carbohydrate metabolism disorders. Patients in the GDM+ group were significantly more likely to undergo cesarean delivery compared to the GDM− group (18 (60%), 27 (34%), p = 0.02), reaching 71% in cases of fetal macrosomia. Cesarean deliveries in patients with gestational diabetes were mostly planned (13 (43%), 14 (18%), p = 0.01). Meanwhile, in patients with macrosomia without gestational diabetes, the nature of the cesarean delivery was mostly emergency (33%), which is known to be associated with more complications for both the mother and the fetus. The length of hospital stay after delivery varied considerably, with longer stays in the GDM+ group, regardless of fetal weight at birth. The terms were 5 (3; 5) for GDM+ and 4 (3; 5) for GDM− ( p = 0.02). Neonatal outcomes including early neonatal complications, 1 and 5 min Apgar scores, and length of hospital stay did not differ significantly by GDM. However, when analyzing the Macrosomia+ and Macrosomia− subgroups, newborns with macrosomia remained in the hospital for significantly longer (4 (4; 5) and 4 (3; 5), p = 0.009). Patients with macrosomia were significantly more likely to have a history of delivering a large baby, regardless of the presence of gestational diabetes (43% and 25% vs. 0% and 7%, p = 0.002). In addition, type 1 and type 2 diabetes in close relatives was more common in the Macrosomia+ subgroups for both the GDP+ (33%) and GDP− (57%) groups. A family history of diabetes among close relatives was found in 21% of cases in the GDM− group and 26% in the GDM+ group in the subgroups without macrosomia. A comparative analysis was conducted on the frequency of excessive, insufficient, and recommended total weight gain during pregnancy among the women studied, using the criteria from the U.S. Institute of Medicine (2009). These criteria are based on the woman’s pre-pregnancy BMI to better differentiate between normal and pathological weight gain. The analysis showed that excessive weight gain during pregnancy was significantly more common among patients in the GDM+ group compared to those in the GDM− group ( p < 0.001), while insufficient weight gain was significantly more frequent in the group of patients with fetal normosomia ( p = 0.005). There were no significant differences in the frequency of recommended weight gain between the two groups ( p = 0.16). Excessive weight gain in women with macrosomia also significantly exceeded that in patients with normal fetal weight at delivery (57% and 75% vs. 22% and 25%, p < 0.001). Diagnoses of carbohydrate metabolism disorders were conducted in the first trimester of pregnancy using fasting venous blood glucose levels, in the second trimester using oral glucose tolerance test results, and in the third trimester based on fasting venous blood glucose levels. GDM was diagnosed in women with fetal macrosomia in 10.0% of cases in the first trimester, 70.0% in the second trimester, and 20.0% in the third trimester. Dynamic analysis of glycemia levels during the oral glucose tolerance test did not show significant postprandial glycemia in women with fetal macrosomia. GDM was diagnosed based on a single fasting venous blood glucose test in all patients with fetal macrosomia. The primary therapeutic method for correcting hyperglycemia was dietary therapy, which involved excluding easily digestible carbohydrates (patients who required a switch to insulin therapy were excluded from further study). The recommended diet for gestational diabetes was followed by patients in 80.0% of cases. Adherence to dietary therapy was lower among patients with gestational diabetes who developed fetal macrosomia: only 50.0% complied with the diet. 2.2. Lipidomic Analysis of Blood Serum in Pregnant Women at 11–13, 24–26, and 30–32 Weeks of Pregnancy A shotgun lipidomics analysis was conducted on the blood serum of 110 patients at 11–13, 24–26, and 30–32 weeks of pregnancy (330 samples in total). Using the OPLS-DA method, the samples were separated into Macrosomia− and Macrosomia+ subgroups based on lipid profiles at three stages of pregnancy: 11–13, 24–26, and 30–32 weeks . The molecular profiles of blood serum could distinguish between samples from patients who later developed or did not develop fetal macrosomia. The lipid profiles in the serum of patients with fetal macrosomia were characterized by increased levels of SM 32:7 and SM 33:5 and decreased levels of phosphatidylcholines, plasmalogen, and SM 34:1 . In the next phase of this study, the potential of mass spectrometry to differentiate the blood serum of pregnant women with fetal macrosomia based on the presence or absence of GDM was evaluated. Lipidomic analyses of blood serum were conducted at 11–13 weeks, 24–26 weeks, and 30–32 weeks of pregnancy. OPLS-DA models were constructed for this purpose: one for patients with GDM and one for patients without carbohydrate metabolism disorders . The greatest differences in the blood lipidome between patients with and without macrosomia in the GDM+ group were found at 24 weeks of pregnancy, which coincided with the onset of GDM and the peak weight gain for most patients. For patients without GDM, the best separation of blood serum samples was observed at 11–13 weeks and 30–32 weeks of pregnancy. At 11–13 weeks of pregnancy, the serum lipidome of patients with fetal macrosomia and GDM was characterized by elevated levels of PC 38:4, PC 38:3, and SM 32:7, along with a decrease in PC 36:5 ( a). At 24–26 weeks, an increase in LPC and PC levels was found in patients with macrosomia and GDM ( b). By 30–32 weeks, a decrease in PC and SM levels was observed in patients with macrosomia and GDM ( c). Similarly, lipids associated with fetal macrosomia in the GDM− group were identified at three time points. At 11–13 weeks of pregnancy, levels of PC O-30:5, SM 32:7, SM 33:5, and SM 32:6 were elevated, while levels of SM 34:1, LPC 16:0, and several PCs were decreased in the macrosomia group ( a). At 24–26 weeks, an increase in LPC and SM levels along with a decrease in PC levels was observed in the macrosomia group ( b). At 30–32 weeks, levels of SM 32:7 and SM 33:5 were elevated, while levels of several PCs and SM 34:1 were decreased in the macrosomia group ( c). The obtained data suggest that the serum lipid profiles of women with fetal macrosomia and those with normal fetal size showed significant differences throughout pregnancy . Moreover, the predictive value of the lipid profile increased when identifying patients with GDM. The characteristics of the developed OPLS-DA models that could be used to predict the occurrence of fetal macrosomia throughout pregnancy and the results of their ROC analyses are presented in . The Q 2 values above 0.4 for the GDM+ group models indicated good predictive ability at all stages of pregnancy. Models based on serum lipid levels at 24 and 30 weeks of pregnancy, when fasting glucose data were unavailable, as well as those at 24 weeks in women without GDM were characterized by lower Q 2 values. shows the ROC curves for each of the developed models along with the areas under the curves (AUCs). All AUC values were above 0.89.
A comparative analysis of the clinical characteristics of the study groups was conducted . The groups were designated as GDM+ (the main group) and GDM− (the comparative group) for patients with and without GDM, respectively. The subgroups within each group were designated as Macrosomia+ and Macrosomia− for patients with and without macrosomia. The results of the analysis between the groups of women examined are presented in . The analysis of the main and comparative groups did not reveal significant differences in the age and anthropometric data of the patients before pregnancy. The median age of the patients did not differ significantly between the groups and subgroups, with an average age of 32 years ( p = 0.56). However, within-group analysis showed a difference between the Macrosomia+ and Macrosomia− subgroups. Pregnant women of the Macrosomia+ subgroups had a significantly higher pre-pregnancy body weight, especially those with GDM (84 (66;96), p = 0.006). Pre-pregnancy BMI significantly differed between the main GDM+ (22.6 (20.1; 26.5)) and comparative GDM− groups (21.2 (19.5; 22.9)) with p = 0.03). Patients with gestational diabetes and macrosomia started pregnancy with a significantly higher BMI compared to patients in other subgroups (27 (23; 30), p = 0.006). Weight gain by the time of delivery differed significantly among the patients with macrosomia, regardless of the presence of gestational diabetes (14 (11; 16) and 16 (13; 18), p = 0.006). Those with fetal macrosomia without gestational diabetes had the greatest weight gain by delivery. Additionally, women in this subgroup had a significantly higher birth weight (3.8 (3.5; 4.1), p = 0.004) compared to other patients. It is likely that genetic and behavioral factors like family diet and eating habits may equally contribute to the development of macrosomia in women without carbohydrate metabolism disorders. Patients in the GDM+ group were significantly more likely to undergo cesarean delivery compared to the GDM− group (18 (60%), 27 (34%), p = 0.02), reaching 71% in cases of fetal macrosomia. Cesarean deliveries in patients with gestational diabetes were mostly planned (13 (43%), 14 (18%), p = 0.01). Meanwhile, in patients with macrosomia without gestational diabetes, the nature of the cesarean delivery was mostly emergency (33%), which is known to be associated with more complications for both the mother and the fetus. The length of hospital stay after delivery varied considerably, with longer stays in the GDM+ group, regardless of fetal weight at birth. The terms were 5 (3; 5) for GDM+ and 4 (3; 5) for GDM− ( p = 0.02). Neonatal outcomes including early neonatal complications, 1 and 5 min Apgar scores, and length of hospital stay did not differ significantly by GDM. However, when analyzing the Macrosomia+ and Macrosomia− subgroups, newborns with macrosomia remained in the hospital for significantly longer (4 (4; 5) and 4 (3; 5), p = 0.009). Patients with macrosomia were significantly more likely to have a history of delivering a large baby, regardless of the presence of gestational diabetes (43% and 25% vs. 0% and 7%, p = 0.002). In addition, type 1 and type 2 diabetes in close relatives was more common in the Macrosomia+ subgroups for both the GDP+ (33%) and GDP− (57%) groups. A family history of diabetes among close relatives was found in 21% of cases in the GDM− group and 26% in the GDM+ group in the subgroups without macrosomia. A comparative analysis was conducted on the frequency of excessive, insufficient, and recommended total weight gain during pregnancy among the women studied, using the criteria from the U.S. Institute of Medicine (2009). These criteria are based on the woman’s pre-pregnancy BMI to better differentiate between normal and pathological weight gain. The analysis showed that excessive weight gain during pregnancy was significantly more common among patients in the GDM+ group compared to those in the GDM− group ( p < 0.001), while insufficient weight gain was significantly more frequent in the group of patients with fetal normosomia ( p = 0.005). There were no significant differences in the frequency of recommended weight gain between the two groups ( p = 0.16). Excessive weight gain in women with macrosomia also significantly exceeded that in patients with normal fetal weight at delivery (57% and 75% vs. 22% and 25%, p < 0.001). Diagnoses of carbohydrate metabolism disorders were conducted in the first trimester of pregnancy using fasting venous blood glucose levels, in the second trimester using oral glucose tolerance test results, and in the third trimester based on fasting venous blood glucose levels. GDM was diagnosed in women with fetal macrosomia in 10.0% of cases in the first trimester, 70.0% in the second trimester, and 20.0% in the third trimester. Dynamic analysis of glycemia levels during the oral glucose tolerance test did not show significant postprandial glycemia in women with fetal macrosomia. GDM was diagnosed based on a single fasting venous blood glucose test in all patients with fetal macrosomia. The primary therapeutic method for correcting hyperglycemia was dietary therapy, which involved excluding easily digestible carbohydrates (patients who required a switch to insulin therapy were excluded from further study). The recommended diet for gestational diabetes was followed by patients in 80.0% of cases. Adherence to dietary therapy was lower among patients with gestational diabetes who developed fetal macrosomia: only 50.0% complied with the diet.
A shotgun lipidomics analysis was conducted on the blood serum of 110 patients at 11–13, 24–26, and 30–32 weeks of pregnancy (330 samples in total). Using the OPLS-DA method, the samples were separated into Macrosomia− and Macrosomia+ subgroups based on lipid profiles at three stages of pregnancy: 11–13, 24–26, and 30–32 weeks . The molecular profiles of blood serum could distinguish between samples from patients who later developed or did not develop fetal macrosomia. The lipid profiles in the serum of patients with fetal macrosomia were characterized by increased levels of SM 32:7 and SM 33:5 and decreased levels of phosphatidylcholines, plasmalogen, and SM 34:1 . In the next phase of this study, the potential of mass spectrometry to differentiate the blood serum of pregnant women with fetal macrosomia based on the presence or absence of GDM was evaluated. Lipidomic analyses of blood serum were conducted at 11–13 weeks, 24–26 weeks, and 30–32 weeks of pregnancy. OPLS-DA models were constructed for this purpose: one for patients with GDM and one for patients without carbohydrate metabolism disorders . The greatest differences in the blood lipidome between patients with and without macrosomia in the GDM+ group were found at 24 weeks of pregnancy, which coincided with the onset of GDM and the peak weight gain for most patients. For patients without GDM, the best separation of blood serum samples was observed at 11–13 weeks and 30–32 weeks of pregnancy. At 11–13 weeks of pregnancy, the serum lipidome of patients with fetal macrosomia and GDM was characterized by elevated levels of PC 38:4, PC 38:3, and SM 32:7, along with a decrease in PC 36:5 ( a). At 24–26 weeks, an increase in LPC and PC levels was found in patients with macrosomia and GDM ( b). By 30–32 weeks, a decrease in PC and SM levels was observed in patients with macrosomia and GDM ( c). Similarly, lipids associated with fetal macrosomia in the GDM− group were identified at three time points. At 11–13 weeks of pregnancy, levels of PC O-30:5, SM 32:7, SM 33:5, and SM 32:6 were elevated, while levels of SM 34:1, LPC 16:0, and several PCs were decreased in the macrosomia group ( a). At 24–26 weeks, an increase in LPC and SM levels along with a decrease in PC levels was observed in the macrosomia group ( b). At 30–32 weeks, levels of SM 32:7 and SM 33:5 were elevated, while levels of several PCs and SM 34:1 were decreased in the macrosomia group ( c). The obtained data suggest that the serum lipid profiles of women with fetal macrosomia and those with normal fetal size showed significant differences throughout pregnancy . Moreover, the predictive value of the lipid profile increased when identifying patients with GDM. The characteristics of the developed OPLS-DA models that could be used to predict the occurrence of fetal macrosomia throughout pregnancy and the results of their ROC analyses are presented in . The Q 2 values above 0.4 for the GDM+ group models indicated good predictive ability at all stages of pregnancy. Models based on serum lipid levels at 24 and 30 weeks of pregnancy, when fasting glucose data were unavailable, as well as those at 24 weeks in women without GDM were characterized by lower Q 2 values. shows the ROC curves for each of the developed models along with the areas under the curves (AUCs). All AUC values were above 0.89.
The main risk factors for macrosomia include maternal age, pre-pregnancy obesity, excessive weight gain before and during pregnancy, and GDM without insulin use . In 2023, the Chinese Medical Association (JCMA) published data on maternal factors associated with fetal macrosomia in the Taiwanese population based on 4262 cases of full-term singleton births. According to the study, the significant risk factors identified were GDM, weight gain during the first six months of pregnancy (6 months GWG), and maternal BMI. The odds ratio (OR) for macrosomia was 3.1 in newborns of mothers with a 6 months GWG ≥ 15 kg, 6.3 for those born to mothers with GDM, and 4.1 for those born to mothers with a BMI ≥ 30 kg/m 2 , respectively. The authors emphasized the importance of counseling mothers to control weight both before and during pregnancy . The analysis of our own clinical data yielded similar results in identifying significant risk factors for fetal macrosomia. These factors included higher pre-pregnancy body weight, with the highest values seen in patients with GDM; pre-pregnancy BMI; and total weight gain by the time of delivery (11–18 kg), regardless of GDM status. Notably, the greatest weight gain at delivery was observed in women with fetal macrosomia who did not have GDM. Additionally, it was interesting to find that women who delivered macrosomic babies had significantly higher birth weights themselves compared to other patients. Other factors included a history of delivering a large baby (regardless of GDM status) among multiparous women and a family history of type 1 or type 2 diabetes in close relatives (33–57%). These findings suggest that both genetic and behavioral factors, such as family diet and eating habits, may contribute equally to the development of macrosomia in women without carbohydrate metabolism disorders. Despite extensive research on macrosomia, predicting which women are at risk remains challenging . Antenatal risk factors are important for predicting macrosomia, but the outcome for both the fetus and the mother depends on the management of labor . In our study, delivery by cesarean section reached 71% in cases of fetal macrosomia. Among patients with GDM, cesarean delivery was often planned, while, in cases of macrosomia without GDM, most cesarean deliveries (33%) were performed on an emergency basis, leading to a higher number of complications for both mother and baby. The length of hospital stay was significantly longer for patients with GDM, regardless of the infant’s birth weight. However, neonatal outcomes, such as early neonatal complications, Apgar scores at 1 and 5 min, and discharge time, did not differ significantly based on GDM status. The outcomes for newborns, however, were significantly different when accounting for birth weight. Infants with macrosomia had a notably longer hospital stay. A retrospective cohort study by Dana Vitner and colleagues, which included 3098 mothers and children with macrosomia and spanned 15 years of observation (2000–2015), allowed for a comparison of management and outcomes between women with predicted fetal macrosomia and those whose fetal weight was unknown at the time of delivery. Primary outcomes included the frequency of cesarean sections (CSs) and postpartum hemorrhage, while secondary outcomes included combined maternal and neonatal outcomes and birth injuries. Macrosomia was predicted in 601 (19.4%) women, while, in 2497 (80.6%) cases, macrosomia was unknown. The rate of CS was more than 3.5 times higher in the predicted macrosomia group (47.2% vs. 12.7%, p < 0.001), consistent with our study results, where macrosomia was a predicted factor in GDM cases. The authors also noted a reduced risk of postpartum hemorrhage with an adjusted odds ratio (aOR) of 0.5 and a 95% confidence interval (95% CI) of 0.2–1.0 with planned CSs in the predicted macrosomia group, as well as reductions in other maternal complications (aOR 0.3, 95% CI 0.2–0.5) and adverse combined neonatal outcomes (aOR 0.7, 95% CI 0.6–0.9). Thus, the authors concluded that planned CSs, when macrosomia is predicted, lead to reduced risks of postpartum hemorrhage and improved maternal and neonatal outcomes, even for infants with a birth weight of less than 4500 g . In cases where fetal macrosomia is suspected, patients should be thoroughly counseled about the delivery plan, and cesarean section should be considered when indicated. Methods for estimating fetal weight and predicting macrosomia include clinical measurements, ultrasound, and magnetic resonance imaging. However, current prediction strategies, such as clinical assessments and ultrasound, are inaccurate. Therefore, the search for new methods to predict and diagnose fetal macrosomia early is highly relevant today. The molecular mechanisms of dyslipidemia in fetal macrosomia remain largely unexplored, making research in this area highly promising. In this study, we attempted to predict fetal macrosomia regardless of the presence or absence of GDM, starting as early as the first screening (11–13 weeks of pregnancy). Analysis of the lipid spectrum in the blood serum of women with fetal macrosomia and normal fetal weight revealed significant differences throughout pregnancy (at 11–13, 24–26, and 30–32 weeks). In a 2023 study by Yingdi Yuan et al., a predictive model for fetal macrosomia was developed based on maternal clinical and laboratory blood biomarkers that differed between women with GDM and macrosomia (GDM-M) and women with GDM and normal birth weight (GDM-N). The model included parameters such as pre-pregnancy BMI, weight gain by 24 weeks, parity, blood glucose levels two hours after an oral glucose tolerance test with 75 g of glucose at 24 weeks, HDL and LDL levels at 24 weeks, and the expression of CLUL1, VCAN, and RNASE3 in plasma at 24 weeks. This model showed good predictive efficiency for forecasting macrosomia in women with GDM . Another 2023 study described comprehensive metabolite profiles in the serum of pregnant mothers and fetuses with normoglycemic macrosomia in a Chinese population. A total of 203 metabolites were identified, with lipids and lipid-like molecules predominating. Among them, 53 metabolites showed significant differences between samples of maternal venous blood and umbilical cord blood. These differences were observed in both the serum of pregnant women and in the fetuses with macrosomia . In our study, lipid spectra in cases of fetal macrosomia showed significant differences throughout pregnancy in both patients with and without GDM while consistently highlighting the presence of phosphatidylcholines and sphingomyelins. It can be assumed that these lipid groups play a critical role in the pathogenesis of fetal macrosomia. Elevated levels of sphingomyelins may indicate decreased cellular sensitivity to insulin. Studies have found a positive correlation between sphingomyelin levels in adipocytes, insulin levels, and the insulin resistance index (HOMA-IR) in individuals with excess body weight . An attempt to characterize and compare placental sphingolipid metabolism in type 1 diabetes (T1D), type 2 diabetes (T2D), and a control group without diabetes was conducted by Miira M. Klemetti and colleagues. Placental samples from T1D, T2D, and the control group were processed for sphingolipid analysis using tandem mass spectrometry. Western blotting, enzyme activity assays, and immunofluorescence were employed to study the enzymes regulating sphingolipids. The levels of ceramide in the placenta were found to be lower in T1D and T2D compared to the control group, which was associated with the increased expression of the enzyme that breaks down ceramide, acid ceramidase (ASAH1). Elevated ceramide levels in the placenta were observed in T1D complicated by preeclampsia. Similarly, higher ceramide levels were noted in pregnancies with poorly controlled glycemia in both T1D and T2D. The protein levels and activity of sphingosine kinase (SPHK), which produces sphingosine-1-phosphate (S1P), were highest in T2D. Additionally, SPHK levels were elevated in pregnancies with T1D and T2D associated with fetal macrosomia. In vitro experiments using JEG3 trophoblast cells demonstrated increased expression and activity of SPHK following glucose and insulin treatment. Specific alterations in the placental sphingolipids characterize placentas with T1D and T2D, depending on the type of diabetes and associated complications in both the fetus and mother. Increased insulin and glucose exposure is likely a contributing factor to the upregulation of the SPHK-S1P axis in placentas affected by diabetes . In a study by Michal Ciborowski and colleagues, metabolic profiles of serum from healthy pregnant women were evaluated to identify early biomarkers of macrosomia and understand the mechanisms leading to abnormal fetal growth, independent of maternal body mass index or the presence of gestational diabetes. Lower levels of phospholipids, lysophospholipids, and monoacylglycerols; low metabolites of vitamin D3; and elevated bilirubin levels were associated with macrosomia. Since most of the changes were related to lipids, levels of the adipocyte fatty acid-binding protein (A-FABP) were measured as a validation concept, revealing a correlation with the studied lipids and birth weight. Serum fingerprinting in early pregnancy may predict the risk of macrosomia. Serum levels of A-FABP and several lipids are promising prognostic markers for macrosomia in healthy pregnancies . According to a review published in 2021, a comprehensive summary of metabolomics studies in gestational diabetes (GDM) was conducted. The authors found that the pathways most commonly disrupted in GDM include amino acids (glutathione, alanine, valine, and serine), carbohydrates (2-hydroxybutyrate and 1,5-anhydroglycitol), and lipids (phosphatidylcholines and lysophosphatidylcholines). They also highlighted the potential use of certain metabolites as predictive markers for the development of GDM using highly stratified modeling methods . Lysophosphatidylcholines are formed as a result of the partial hydrolysis of phosphatidylcholines by phospholipase A2. In their study, Patel N. et al. demonstrated a positive correlation between the levels of phosphatidylcholines and lysophosphatidylcholines in umbilical blood, the birth weight of infants born to obese mothers, and weight gain in children during the first six months of life. High levels of LPC 16:1 and 18:1 exhibited a linear relationship with hyperglycemia in women at 28 weeks of pregnancy . In a study by Hellmuth et al. (2017), a positive correlation was found between newborn body weight and LPC 14:0, LPC 16:1, and LPC 18:1 in umbilical blood . Another study demonstrated a positive correlation between LPC 14:0 and childhood obesity . In our study, increased levels of LPC 16:0 were noted in the group of patients with fetal macrosomia, suggesting the potential prognostic significance of this lipid in this condition. Xiuli Su et al. applied untargeted metabolomic analysis to identify blood metabolites with high predictive potential for detecting type 2 diabetes (T2D) over a follow-up period of approximately 16 years. The metabolic profiles revealed significant disturbances in metabolomics even before the clinical onset and diagnosis of T2D. Overall metabolic shifts were closely related to insulin resistance rather than β-cell dysfunction. Additionally, 188 out of 578 annotated metabolites were associated with insulin resistance. Bidirectional mediational analysis revealed potential causal relationships between metabolites, insulin resistance, and the risk of T2D. Metabolomic analysis has potential clinical utility in predicting T2D . The molecular mechanisms of dyslipidemia in fetal macrosomia remain poorly understood, and research in this area may not only enhance our understanding of the role of metabolic diseases but also contribute to the development of effective preventive measures aimed at reducing the incidence of this condition.
4.1. Study Design A case–control study was conducted at the National Medical Research Center for Obstetrics, Gynecology, and Perinatology named after V.I. Kulakov in Moscow from January to September 2024 . Out of 1200 women who were monitored in the Scientific and Outpatient Department of the Kulakov Center starting from the first-trimester prenatal screening procedure (11–13.6 weeks), 110 patients were selected after delivery. Two groups were formed based on the presence (GDM+) or absence (GDM−) of gestational diabetes mellitus. The main GDM+ group included 30 patients, while the comparison GDM− group included 80. To address this study’s objectives, patients were further stratified into subgroups based on the presence (Macrosomia+) or absence (Macrosomia−) of fetal macrosomia. The Macrosomia+ subgroup of the GDM+ group included 7 patients with newborns weighing ≥ 4000 g and/or above the 90th percentile with GDM and the Macrosomia− subgroup included 23 patients with newborns weighing 2501 to 3999 g and GDM. The Macrosomia+ subgroup of the GDM-group included 24 patients with newborns weighing ≥ 4000 g and/or above the 90th percentile without GDM, while the Macrosomia− subgroup (control) included 56 patients with newborns weighing 2501 to 3999 g without GDM. All patients signed voluntary informed consent to participate in this study. This work was approved by the Ethical Committee of the National Medical Research Center for Obstetrics, Gynecology, and Perinatology named after academician V.I. Kulakov (protocol no. 9, dated 22 November 2018). A semi-quantitative assessment of serum lipid levels was performed using mass spectrometry. Inclusion criteria for the main group were Caucasian race, singleton pregnancy, newborn weight between 2501 g and 4999 g with GDM, and patient consent to participate in this study. Inclusion criteria for the comparison group were Caucasian race, absence of GDM, singleton pregnancy, newborn weight between 2501 g and 4999 g, and patient consent to participate in this study. A mandatory condition for all patients’ inclusion in this study was participation in a comprehensive examination, which included three screening ultrasounds, venous blood collection, an oral glucose tolerance test at 24–28 weeks of pregnancy, and delivery at the center. Exclusion criteria were type 1 and type 2 diabetes, any somatic pathology in the stage of decompensation, oncological diseases, autoimmune diseases, bronchial asthma in the stage of medical compensation, and multiple pregnancies. Ultrasound examinations were conducted for all pregnant women at the designated times (11–14 weeks, 18–21 weeks, and 30–32 weeks of pregnancy). GDM was diagnosed using a glucose tolerance test with 75 g after 8–14 h of overnight fasting. After the first blood sample was taken, the plasma glucose level was measured within 30 min. If the glucose level exceeded 5.1 mmol/L after the first blood draw, the test was discontinued. If the test continued, the patient drank a glucose solution consisting of 75 g of dry glucose dissolved in 250–300 mL of warm (37–40 °C), non-carbonated water within 5 min. Subsequent blood draws for the determination of plasma glucose levels were performed 1 and 2 h after the glucose load. The threshold values for plasma glucose for diagnosing GDM and manifest diabetes are presented in and . 4.2. Sample Collection Blood samples for lipid analysis were collected at three points: the first at 11–13 weeks, the second at 24–26 weeks, and the third at 30–32 weeks of pregnancy. The samples were collected using a vacuum method into a sterile 9 mL S-Monovette tube containing a clot activator and separation granules, following a 12 h fasting diet. The serum was then centrifuged for 10 min at 700 g and 4 °C. The supernatant was carefully pipetted into a sterile tube, frozen, and stored at −80 °C. Blood was drawn from each patient at the specified times during this study: 11–13 weeks, 24–26 weeks, and 30–32 weeks of pregnancy. 4.3. Sample Preparation for Shotgun MS/MS Lipid extracts were obtained using a modified Folch method. To 40 μL of serum, 480 μL of a chloroform–methanol mixture (2:1, v / v ) was added. The mixture was incubated for 10 min and then filtered using filter paper, and 150 μL of a 1 mol/L NaCl aqueous solution was added to the resulting solution. The mixture was centrifuged at 3000 rpm for 5 min at room temperature. The organic lower layer containing the lipids was collected and dried under a nitrogen stream and then re-dissolved in a mixture of acetonitrile–isopropanol (1:1, v / v ) for subsequent mass spectrometric analysis. 4.4. Shotgun MS Analysis of Serum Lipid Extracts The molecular composition of serum samples was determined using flow injection analysis (FIA) electrospray ionization mass spectrometry with a Maxis Impact qTOF mass spectrometer (Bruker Daltonics, Bremen, Germany) . Mass spectra were acquired over the m / z range of 400–1000 with the following parameters: capillary voltage at 4.1 kV, nebulizer gas pressure at 0.7 bar, drying gas flow rate at 6 L/min, and drying gas temperature at 200 °C. A constant flow of a methanol/water mixture (9:1, v / v ) was supplied at a rate of 10 µL/min using a Dionex UltiMate 3000 binary pump (ThermoScientific, Bremen, Germany), and 20 µL of sample was injected via a Dionex UltiMate 3000 autosampler (ThermoScientific, Bremen, Germany). Mass spectra were recorded in positive ion mode, achieving a resolution of 50,000 within the mass range of m / z 400–1000. For compound identification, tandem mass spectrometry (MS/MS) was performed in data-dependent mode. After a full mass scan, the five most abundant peaks were selected for MS/MS analysis, utilizing collision-induced dissociation with 35 eV of collision energy, a 1 Da isolation window, and a 1 min mass exclusion time. After the mass spectrometric analysis, 100 mass spectra obtained during sample elution were averaged, normalized by total ion current (TIC), and converted into an abundance– m / z table for further processing. The spray remained stable throughout the analysis due to the constant flow and unchanged ion source parameters. The relative standard deviation of TIC during the integration time did not exceed 10%, while the intragroup relative standard deviation ranged from 10% to 20%. 4.5. Statistical Analysis Statistical data processing was conducted using RStudio version 2023.06.1 with custom scripts written in R version 4.1.1. The Shapiro–Wilk test was employed to assess the normality of data distribution. For quantitative data that did not follow a normal distribution, median values (Me) and quartiles (Q1, Q3) were reported. Qualitative data were expressed as absolute values . Comparative analysis of qualitative data was performed using Fisher’s exact test and the Chi-square (χ 2 ) test. For quantitative data, the Mann–Whitney test was used for pairwise comparisons between groups, while the Kruskal–Wallis test was applied for comparisons involving more than two groups. Bonferroni correction was utilized for multiple comparisons. The significance threshold was set at 0.05. Mass spectrometry data analysis was carried out using multivariate analysis through Orthogonal Projections to Latent Structures Discriminant Analysis (OPLS-DA) , which allows for the construction of statistical models using multidimensional data to differentiate between samples. In building the linear regression model, the variable influence on projection (VIP) was calculated to assess the impact of individual X-variables (lipids) on the model. This helped to identify the most significant lipids and assess the statistical significance of differences in their levels between the study groups. Lipids were identified based on their accurate mass using the Lipid Maps database and by characteristic tandem mass spectra . The lipid nomenclature corresponds to LipidMaps . To determine the prognostic significance of the features, ROC analysis was performed, generating ROC curves and calculating the AUC.
A case–control study was conducted at the National Medical Research Center for Obstetrics, Gynecology, and Perinatology named after V.I. Kulakov in Moscow from January to September 2024 . Out of 1200 women who were monitored in the Scientific and Outpatient Department of the Kulakov Center starting from the first-trimester prenatal screening procedure (11–13.6 weeks), 110 patients were selected after delivery. Two groups were formed based on the presence (GDM+) or absence (GDM−) of gestational diabetes mellitus. The main GDM+ group included 30 patients, while the comparison GDM− group included 80. To address this study’s objectives, patients were further stratified into subgroups based on the presence (Macrosomia+) or absence (Macrosomia−) of fetal macrosomia. The Macrosomia+ subgroup of the GDM+ group included 7 patients with newborns weighing ≥ 4000 g and/or above the 90th percentile with GDM and the Macrosomia− subgroup included 23 patients with newborns weighing 2501 to 3999 g and GDM. The Macrosomia+ subgroup of the GDM-group included 24 patients with newborns weighing ≥ 4000 g and/or above the 90th percentile without GDM, while the Macrosomia− subgroup (control) included 56 patients with newborns weighing 2501 to 3999 g without GDM. All patients signed voluntary informed consent to participate in this study. This work was approved by the Ethical Committee of the National Medical Research Center for Obstetrics, Gynecology, and Perinatology named after academician V.I. Kulakov (protocol no. 9, dated 22 November 2018). A semi-quantitative assessment of serum lipid levels was performed using mass spectrometry. Inclusion criteria for the main group were Caucasian race, singleton pregnancy, newborn weight between 2501 g and 4999 g with GDM, and patient consent to participate in this study. Inclusion criteria for the comparison group were Caucasian race, absence of GDM, singleton pregnancy, newborn weight between 2501 g and 4999 g, and patient consent to participate in this study. A mandatory condition for all patients’ inclusion in this study was participation in a comprehensive examination, which included three screening ultrasounds, venous blood collection, an oral glucose tolerance test at 24–28 weeks of pregnancy, and delivery at the center. Exclusion criteria were type 1 and type 2 diabetes, any somatic pathology in the stage of decompensation, oncological diseases, autoimmune diseases, bronchial asthma in the stage of medical compensation, and multiple pregnancies. Ultrasound examinations were conducted for all pregnant women at the designated times (11–14 weeks, 18–21 weeks, and 30–32 weeks of pregnancy). GDM was diagnosed using a glucose tolerance test with 75 g after 8–14 h of overnight fasting. After the first blood sample was taken, the plasma glucose level was measured within 30 min. If the glucose level exceeded 5.1 mmol/L after the first blood draw, the test was discontinued. If the test continued, the patient drank a glucose solution consisting of 75 g of dry glucose dissolved in 250–300 mL of warm (37–40 °C), non-carbonated water within 5 min. Subsequent blood draws for the determination of plasma glucose levels were performed 1 and 2 h after the glucose load. The threshold values for plasma glucose for diagnosing GDM and manifest diabetes are presented in and .
Blood samples for lipid analysis were collected at three points: the first at 11–13 weeks, the second at 24–26 weeks, and the third at 30–32 weeks of pregnancy. The samples were collected using a vacuum method into a sterile 9 mL S-Monovette tube containing a clot activator and separation granules, following a 12 h fasting diet. The serum was then centrifuged for 10 min at 700 g and 4 °C. The supernatant was carefully pipetted into a sterile tube, frozen, and stored at −80 °C. Blood was drawn from each patient at the specified times during this study: 11–13 weeks, 24–26 weeks, and 30–32 weeks of pregnancy.
Lipid extracts were obtained using a modified Folch method. To 40 μL of serum, 480 μL of a chloroform–methanol mixture (2:1, v / v ) was added. The mixture was incubated for 10 min and then filtered using filter paper, and 150 μL of a 1 mol/L NaCl aqueous solution was added to the resulting solution. The mixture was centrifuged at 3000 rpm for 5 min at room temperature. The organic lower layer containing the lipids was collected and dried under a nitrogen stream and then re-dissolved in a mixture of acetonitrile–isopropanol (1:1, v / v ) for subsequent mass spectrometric analysis.
The molecular composition of serum samples was determined using flow injection analysis (FIA) electrospray ionization mass spectrometry with a Maxis Impact qTOF mass spectrometer (Bruker Daltonics, Bremen, Germany) . Mass spectra were acquired over the m / z range of 400–1000 with the following parameters: capillary voltage at 4.1 kV, nebulizer gas pressure at 0.7 bar, drying gas flow rate at 6 L/min, and drying gas temperature at 200 °C. A constant flow of a methanol/water mixture (9:1, v / v ) was supplied at a rate of 10 µL/min using a Dionex UltiMate 3000 binary pump (ThermoScientific, Bremen, Germany), and 20 µL of sample was injected via a Dionex UltiMate 3000 autosampler (ThermoScientific, Bremen, Germany). Mass spectra were recorded in positive ion mode, achieving a resolution of 50,000 within the mass range of m / z 400–1000. For compound identification, tandem mass spectrometry (MS/MS) was performed in data-dependent mode. After a full mass scan, the five most abundant peaks were selected for MS/MS analysis, utilizing collision-induced dissociation with 35 eV of collision energy, a 1 Da isolation window, and a 1 min mass exclusion time. After the mass spectrometric analysis, 100 mass spectra obtained during sample elution were averaged, normalized by total ion current (TIC), and converted into an abundance– m / z table for further processing. The spray remained stable throughout the analysis due to the constant flow and unchanged ion source parameters. The relative standard deviation of TIC during the integration time did not exceed 10%, while the intragroup relative standard deviation ranged from 10% to 20%.
Statistical data processing was conducted using RStudio version 2023.06.1 with custom scripts written in R version 4.1.1. The Shapiro–Wilk test was employed to assess the normality of data distribution. For quantitative data that did not follow a normal distribution, median values (Me) and quartiles (Q1, Q3) were reported. Qualitative data were expressed as absolute values . Comparative analysis of qualitative data was performed using Fisher’s exact test and the Chi-square (χ 2 ) test. For quantitative data, the Mann–Whitney test was used for pairwise comparisons between groups, while the Kruskal–Wallis test was applied for comparisons involving more than two groups. Bonferroni correction was utilized for multiple comparisons. The significance threshold was set at 0.05. Mass spectrometry data analysis was carried out using multivariate analysis through Orthogonal Projections to Latent Structures Discriminant Analysis (OPLS-DA) , which allows for the construction of statistical models using multidimensional data to differentiate between samples. In building the linear regression model, the variable influence on projection (VIP) was calculated to assess the impact of individual X-variables (lipids) on the model. This helped to identify the most significant lipids and assess the statistical significance of differences in their levels between the study groups. Lipids were identified based on their accurate mass using the Lipid Maps database and by characteristic tandem mass spectra . The lipid nomenclature corresponds to LipidMaps . To determine the prognostic significance of the features, ROC analysis was performed, generating ROC curves and calculating the AUC.
Identifying at-risk groups for fetal macrosomia early in pregnancy allows healthcare providers to implement preventive measures focusing on lifestyle and dietary changes as early as the first trimester. This proactive approach can help tailor obstetric management to ensure the timely selection of the best delivery method, ultimately reducing adverse birth outcomes for both mothers and their babies. Our research has identified clinical and experimental predictive markers for fetal macrosomia based on mass spectrometry analysis of blood samples. Key clinical risk factors include maternal age, pre-pregnancy obesity, excessive weight gain before and during pregnancy, and gestational diabetes (GDM) that has not progressed to insulin therapy. Our findings indicate that cesarean deliveries, often necessitated by fetal macrosomia, are typically performed in emergency situations, which increases the risk of complications for both mother and child. Additionally, newborn outcomes correlate strongly with their birth weight. In this study, we aimed to predict fetal macrosomia regardless of whether GDM is present, starting as early as the first trimester. Our analysis of the lipid profiles in the serum of women with fetal macrosomia, compared to those with normal fetal weights, revealed significant differences throughout pregnancy (at 11–13 weeks, 24–26 weeks, and 30–32 weeks). We hypothesize that the lipid groups we identified, particularly phosphatidylcholines and sphingomyelins, play a critical role in the development of fetal macrosomia and could serve as laboratory markers for this complication. We developed pilot OPLS-DA models that demonstrate high sensitivity and specificity for predicting fetal macrosomia during pregnancy. These models can be applied regardless of GDM status, making them useful for women with unknown GDM status as well as those with confirmed positive or negative diagnoses. The molecular mechanisms that contribute to fetal macrosomia are largely uncharted, and further research in this area could enhance our understanding of the impact of metabolic diseases. This knowledge could lead to effective preventive strategies, improved prognostic techniques, and early diagnosis methods to help reduce the incidence of fetal macrosomia.
|
Scaling Down Large-Scale Thawing of Monoclonal Antibody Solutions: 3D Temperature Profiles, Changes in Concentration, and Density Gradients | b826f2d9-7063-489f-ad40-9f384312c214 | 8688388 | Pharmacology[mh] | Therapeutic proteins, especially monoclonal antibodies (mAbs), are of significant importance in a constantly growing market and number of indications . During large-scale production of biopharmaceuticals freezing is a commonly used processing step to enhance physical and chemical stability and to increase shelf life . In comparison to liquid storage, frozen storage and transport offers flexibility for the manufacturing process by decoupling drug substance and drug product processing, decreased risk of microbial growth and reduced foaming, shaking and agitation . While freezing and thawing (FT) is intentional during production of bulk drug substances and drug products , commercial products may also be frozen multiple times accidentally by mishandling. Despite its apparent simplicity and undoubted advantages, freezing can be associated with a variety of possible drawbacks. Cold denaturation , cryoconcentration on a microscopic and macroscopic scale , crystallisation of cryoprotectants or buffer salts, associated with significant pH shifts , interactions at the ice interface , and ultimately conformational and colloidal instability are well known, although not fully understood. In addition, Authelin et al. address enhanced oxidation, the formation of air bubbles, local pressure, and mechanical stresses as further considerable occurrences that are so far not well examined . The FT process becomes even more complex as the described phenomena are not only relevant during freezing but also during thawing. Especially crystallisation during and concentration gradients after thawing can lead to denaturation, aggregation and precipitation of proteins . During freezing, growing ice crystals exclude proteins and other excipients. Consequently, the concentration of the remaining unfrozen solution increases. The microenvironment of the protein between the ice crystals can be described as a freeze-concentrated matrix (FCM). Under ideal equilibrium conditions its maximum concentration is characterised by state diagrams . The glass transition temperature of the maximally freeze concentrated solution (T g ′) is the temperature of complete solidification of the FCM. Macroscopically, the cryoconcentrated solution near the freezing front is transported away into unfrozen regions by diffusion and convection. This leads to a substantial heterogeneity throughout the frozen bulk . During the thawing process the FCM melts out of the frozen bulk leaving nearly pure ice behind. Due to the high density of the FCM and the low density of the ice, ice floats on top and dilutes the top region as it melts . Several studies report a tremendous concentration gradient in post-thawed large-scale bottles . These gradients can be associated with decreased protein stability. Protein precipitation can occur in the top layer due to lack of stabilisers or at the bottom because of a salting out effect . The complex and interdependent processes have been studied using small-scale experiments and specific scale-down devices (SDDs) . Such experiments support finding of the optimal formulation, revealing limitations during processing and storage, and unveiling aggregation mechanisms under the chosen FT setup. However, the experiments do not reflect the large-scale FT rates, cryoconcentration, surface area of ice, and exposure time of protein to the numerous stresses and therefore cannot be straightly extrapolated to large-scale . Using Computational Fluid Dynamics (CFD) the heat transfer can be adapted so that mass fractions in small- and large-scale containers experience equivalent stress. Based on this approach SmartFreeZ developed a SDD assisted by CFD following a novel and innovative approach to representatively scale down FT behaviour in a 2 L bottle. The scale-down strategy by CFD has been described previously . In contrast to matching FT profiles at the last point to freeze (LPTF) or the last point to thaw (LPTT), simulations were used to divide the bulk into control volumes of approximately 1 mm 3 . The time between the beginning of freezing in the container and reaching T g ′ in the control volume can be calculated. This time span is most detrimental for protein molecules as they become exposed to the ice liquid interface and concentrated together with the other solutes. When passing T g ′, the viscosity of the FCM increases considerably and the decreased mobility prevents aggregation . The thermal history of a protein solution during large-scale FT was translated into the cumulative thermal history in the SDD. The SDD, which matches the thermal history of a large-scale 2 L bottle, consists of a 125 mL bottle surrounded by a unique holder that controls heat exchange. Previously, we characterised the performance of this innovative SDD during freezing, revealing a good agreement between the SDD and large-scale 2 L bottles . It still needs to be evaluated whether this SDD also reflects thawing at large-scale. Large-scale thawing can affect protein stability. Several studies highlighted significant concentration gradients that evolve during thawing and its negative impact on mAb stability . The thawing process is more time-consuming than freezing under similar conditions. Thus, the protein is exposed to an unfavourable environment significantly longer. If the bottle is not homogenised immediately after thawing, this time span extends even further. Until now no SDD is marketed that can adequately mimic the thawing process in widely used rectangular bottles. While our previous study focused on the validation of the SDD during freezing , this study aims to validate the SDD in respect of thawing. Therefore, we compared temperature profiles at several locations in the SDD to a 125 mL and a 2 L bottle. We found significant changes in protein and excipient concentrations in the 2 L bottle that were predicted by the SDD. Excipients play a critical role in ensuring protein stability during freezing and subsequent thawing. Therefore, we did not only compare a model mAb but also the buffer species histidine and the surfactant polysorbate 80 (PS80) at different locations in the SDD to the commercially utilised 2 L bottle. We used these results to assess the density gradient that evolves during thawing and, to our knowledge, has not been examined. Finally, we underline the importance of diffusion, which leads to a rapid equilibration of histidine in the SDD.
Materials Polyethersulfone bottle top and syringe filters (0.2 µm) were purchased from VWR International GmbH (Darmstadt, Germany). Cellon S.A. (Bascharage, Luxembourg) provided 2 L and 125 mL PharmaTainer™ polyethylene terephthalate (PET) bottles. Dipotassium hydrogen phosphate and potassium dihydrogen phosphate needed for the preparation of the mobile phase for the high-performance liquid chromatography (HPLC) were obtained from Merck KGaA (Darmstadt, Germany). Novartis AG (Basel, Switzerland) provided a 185 mg/mL IgG1 mAb stock solution in a 20 mM histidine buffer at pH 5.5. For the dilution l -histidine monochloride monohydrate and l -histidine were purchased from Merck KGaA (Darmstadt, Germany). Super Refined™ PS80 from Croda International plc (Snaith, UK) was used. Dulbecco’s phosphate buffered saline 1× (DPBS) without calcium and magnesium chloride, needed for the quantification of PS80, were obtained from Gibco™ (Thermo Fisher Scientific Inc., Waltham, MA, USA). The fluorescence probe 4,4′-dianilino-1,1′-binaphthyl-5,5′-disulfonic acid dipotassium salt (bis-ANS) was purchased from Invitrogen™ (Thermo Fisher Scientific Inc., Waltham, MA, USA). Scale-Down Device SmartFreeZ (Porto Salvo, Portugal) provided the SDDs used in this study. A detailed description of the development assisted by CFD can be found elsewhere . The specific SDD was designed to predict thawing in a rectangular 2 L PharmaTainer™ bottle. The 3D printed SDD covers a 125 mL PharmaTainer™ bottle and uses 1% ethanol as a phase change liquid to insulate approximately two walls (Fig. ). Thereby, a 125 mL bottle shall be utilised to mimic FT processes in a large-scale 2 L bottle. A soft polymer insert prevents circulation of air between the SDD and the bottle. To avoid radiation, a top cover shields the bottle from above. Two SDDs were used simultaneously in back-to-back orientation during measurements. One SDD, filled with highly purified water, was needed for shielding. A second one containing the sample was used for experiments. Preparation of Protein Samples The stock solution was diluted to a final concentration of 5 mg/mL mAb in 20 mM histidine at pH 5.5. The concentration was determined via UV absorption at 280 nm with a NanoDrop One by Thermo Fisher Scientific Inc. (Waltham, MA, USA). Samples containing 0.4 mg/mL PS80 were prepared by spiking a 10 mg/mL PS80 stock solution. All solutions were filtered through a 0.2 µm bottle top or syringe filter. 2 L and 125 mL bottles were 80% filled with 1.6 L and 100 mL, respectively. Temperature Mapping During Thawing The temperature measurements were performed as previously described . Briefly, five type T thermocouples (TCs) connected to an HH520 handheld data logger thermometer (OMEGA Engineering GmbH, Deckenpfronn, Germany) were positioned at half liquid height in the edges and the centre of the 125 mL bottle using stainless steel capillaries (Acufirm Ernst Kratz GmbH, Berlin, Germany) for reproducible placement (Fig. ). A sixth TC (TC 6) was placed at the exact position of TC 1 but at 75% liquid level. Six TCs were arranged at equivalent positions in the 2 L bottle. After acclimatisation at 20 °C for 1 h in an MKF 240 air-blast climate chamber (Binder GmbH, Tuttlingen, Germany), the chamber was cooled at maximum rate to − 40 °C and the temperature held for 10 h. Subsequently, the temperature was set to 20 °C with maximum heating rate and the solution thawed until the set temperature was reached at all positions. All temperature measurements were executed in triplicates in independent runs. Analysis of Concentration Gradients After Thawing Concentration gradients after FT were analysed, once in the 2 L bottle and in triplicates in the SDD and the 125 mL bottle. Samples were taken from nine (2 L bottle) or five (SDD and 125 mL bottle) layers. For each layer five 1 mL samples were taken, four in the edges and a fifth from the centre. Samples were taken with a 1 mL serological glass pipette or 1 mL syringes (B. Braun Melsungen AG, Melsungen, Germany) equipped with a Sterican® 0.80 × 120 mm needle (Braun Melsungen AG, Melsungen, Germany). Quantification of mAb and Histidine Size-exclusion chromatography (SEC) on an Agilent 1200 HPLC with a diode array detector (Agilent Technologies, Santa Clara, CA, USA) allowed the separation and simultaneous quantification of mAb and histidine. Therefore, samples were diluted 1:4 or 1:10 with mobile phase. After centrifugation for 2 min at 25,700× g with a Heraeus™ Megafuge™ 16R (Thermo Fisher Scientific Inc., Waltham, MA, USA) 5 µL of each sample were injected. As stationary phase a TSKgel G3000 SWxl column (Tosoh Bioscience GmbH, Griesheim, Germany) and as mobile phase a 150 mM potassium phosphate buffer pH 6.5 at a flow rate of 0.4 mL/min were used. Histidine was quantified at 210 nm and mAb at 280 nm by comparing the areas under the curve to standard curves (R 2 = 0.9999 and R 2 = 0.9994). Quantification of PS80 The method for the PS80 quantification was adapted from Zheng et al. . Samples were diluted 1:4 or 1:10 with DPBS and subsequently heated for 5 min at 99 °C. Afterwards, samples were centrifuged for 5 min at 25,700× g . 190 µL supernatant were mixed with 10 µL of 1 mM bis-ANS and vortexed for 5 s. 60 µL of each sample were analysed in a Varian Cary Eclipse fluorescence spectrophotometer (Agilent Technologies, Santa Clara, CA, USA) using a quartz cuvette at 380 nm excitation and 500 nm emission with both slits set to 5 nm. A calibration curve of PS80 in DPBS allowed the quantification of PS80 between 0.005 and 0.15 mg/mL (R 2 = 0.9988). Diffusion of Solution Components After Thawing The diffusion of mAb and histidine after complete thawing of the solution in the SDD was mapped. Samples were taken after complete thawing of the solution (16 h at 20 °C) as well as after additional 24 h and 48 h. To minimize any possible influence of the removed volume on subsequent results, only 0.25 mL were taken per sample. Samples were obtained from the edges in the top layer, the middle layer and at the bottom. Mixing was avoided by tightly attaching the SDD to the grid in the air-blast climate chamber and taking samples slowly with a 1 mL syringe equipped with a needle. The DynaPro Plate Reader III (Wyatt Technology, Dernbach, Germany) was used to determine diffusion coefficients of mAb and PS80 via dynamic light scattering. Samples with 10 mg/mL mAb or PS80, respectively, were prepared and filtered. 100 µL of each sample was pipetted in triplicates into a 96-well clear bottom plate (Corning Inc., Corning, NY, USA) and ten acquisitions of 5 s at 25 °C taken. The Dynamics V7.8.2.18 software was used for all calculations. Analysis of Density Gradients The changes in density after FT were assessed using a portable density meter DMA 35 Standard (Anton Paar Group AG, Graz, Austria). 15 mL samples were prepared according to concentrations found for each layer during the analysis of the concentration gradients after thawing (Table ). After a pre-rinse, density was measured in triplicates at room temperature. The density meter extrapolated results to 20 °C.
Polyethersulfone bottle top and syringe filters (0.2 µm) were purchased from VWR International GmbH (Darmstadt, Germany). Cellon S.A. (Bascharage, Luxembourg) provided 2 L and 125 mL PharmaTainer™ polyethylene terephthalate (PET) bottles. Dipotassium hydrogen phosphate and potassium dihydrogen phosphate needed for the preparation of the mobile phase for the high-performance liquid chromatography (HPLC) were obtained from Merck KGaA (Darmstadt, Germany). Novartis AG (Basel, Switzerland) provided a 185 mg/mL IgG1 mAb stock solution in a 20 mM histidine buffer at pH 5.5. For the dilution l -histidine monochloride monohydrate and l -histidine were purchased from Merck KGaA (Darmstadt, Germany). Super Refined™ PS80 from Croda International plc (Snaith, UK) was used. Dulbecco’s phosphate buffered saline 1× (DPBS) without calcium and magnesium chloride, needed for the quantification of PS80, were obtained from Gibco™ (Thermo Fisher Scientific Inc., Waltham, MA, USA). The fluorescence probe 4,4′-dianilino-1,1′-binaphthyl-5,5′-disulfonic acid dipotassium salt (bis-ANS) was purchased from Invitrogen™ (Thermo Fisher Scientific Inc., Waltham, MA, USA).
SmartFreeZ (Porto Salvo, Portugal) provided the SDDs used in this study. A detailed description of the development assisted by CFD can be found elsewhere . The specific SDD was designed to predict thawing in a rectangular 2 L PharmaTainer™ bottle. The 3D printed SDD covers a 125 mL PharmaTainer™ bottle and uses 1% ethanol as a phase change liquid to insulate approximately two walls (Fig. ). Thereby, a 125 mL bottle shall be utilised to mimic FT processes in a large-scale 2 L bottle. A soft polymer insert prevents circulation of air between the SDD and the bottle. To avoid radiation, a top cover shields the bottle from above. Two SDDs were used simultaneously in back-to-back orientation during measurements. One SDD, filled with highly purified water, was needed for shielding. A second one containing the sample was used for experiments.
The stock solution was diluted to a final concentration of 5 mg/mL mAb in 20 mM histidine at pH 5.5. The concentration was determined via UV absorption at 280 nm with a NanoDrop One by Thermo Fisher Scientific Inc. (Waltham, MA, USA). Samples containing 0.4 mg/mL PS80 were prepared by spiking a 10 mg/mL PS80 stock solution. All solutions were filtered through a 0.2 µm bottle top or syringe filter. 2 L and 125 mL bottles were 80% filled with 1.6 L and 100 mL, respectively.
The temperature measurements were performed as previously described . Briefly, five type T thermocouples (TCs) connected to an HH520 handheld data logger thermometer (OMEGA Engineering GmbH, Deckenpfronn, Germany) were positioned at half liquid height in the edges and the centre of the 125 mL bottle using stainless steel capillaries (Acufirm Ernst Kratz GmbH, Berlin, Germany) for reproducible placement (Fig. ). A sixth TC (TC 6) was placed at the exact position of TC 1 but at 75% liquid level. Six TCs were arranged at equivalent positions in the 2 L bottle. After acclimatisation at 20 °C for 1 h in an MKF 240 air-blast climate chamber (Binder GmbH, Tuttlingen, Germany), the chamber was cooled at maximum rate to − 40 °C and the temperature held for 10 h. Subsequently, the temperature was set to 20 °C with maximum heating rate and the solution thawed until the set temperature was reached at all positions. All temperature measurements were executed in triplicates in independent runs.
Concentration gradients after FT were analysed, once in the 2 L bottle and in triplicates in the SDD and the 125 mL bottle. Samples were taken from nine (2 L bottle) or five (SDD and 125 mL bottle) layers. For each layer five 1 mL samples were taken, four in the edges and a fifth from the centre. Samples were taken with a 1 mL serological glass pipette or 1 mL syringes (B. Braun Melsungen AG, Melsungen, Germany) equipped with a Sterican® 0.80 × 120 mm needle (Braun Melsungen AG, Melsungen, Germany). Quantification of mAb and Histidine Size-exclusion chromatography (SEC) on an Agilent 1200 HPLC with a diode array detector (Agilent Technologies, Santa Clara, CA, USA) allowed the separation and simultaneous quantification of mAb and histidine. Therefore, samples were diluted 1:4 or 1:10 with mobile phase. After centrifugation for 2 min at 25,700× g with a Heraeus™ Megafuge™ 16R (Thermo Fisher Scientific Inc., Waltham, MA, USA) 5 µL of each sample were injected. As stationary phase a TSKgel G3000 SWxl column (Tosoh Bioscience GmbH, Griesheim, Germany) and as mobile phase a 150 mM potassium phosphate buffer pH 6.5 at a flow rate of 0.4 mL/min were used. Histidine was quantified at 210 nm and mAb at 280 nm by comparing the areas under the curve to standard curves (R 2 = 0.9999 and R 2 = 0.9994). Quantification of PS80 The method for the PS80 quantification was adapted from Zheng et al. . Samples were diluted 1:4 or 1:10 with DPBS and subsequently heated for 5 min at 99 °C. Afterwards, samples were centrifuged for 5 min at 25,700× g . 190 µL supernatant were mixed with 10 µL of 1 mM bis-ANS and vortexed for 5 s. 60 µL of each sample were analysed in a Varian Cary Eclipse fluorescence spectrophotometer (Agilent Technologies, Santa Clara, CA, USA) using a quartz cuvette at 380 nm excitation and 500 nm emission with both slits set to 5 nm. A calibration curve of PS80 in DPBS allowed the quantification of PS80 between 0.005 and 0.15 mg/mL (R 2 = 0.9988).
Size-exclusion chromatography (SEC) on an Agilent 1200 HPLC with a diode array detector (Agilent Technologies, Santa Clara, CA, USA) allowed the separation and simultaneous quantification of mAb and histidine. Therefore, samples were diluted 1:4 or 1:10 with mobile phase. After centrifugation for 2 min at 25,700× g with a Heraeus™ Megafuge™ 16R (Thermo Fisher Scientific Inc., Waltham, MA, USA) 5 µL of each sample were injected. As stationary phase a TSKgel G3000 SWxl column (Tosoh Bioscience GmbH, Griesheim, Germany) and as mobile phase a 150 mM potassium phosphate buffer pH 6.5 at a flow rate of 0.4 mL/min were used. Histidine was quantified at 210 nm and mAb at 280 nm by comparing the areas under the curve to standard curves (R 2 = 0.9999 and R 2 = 0.9994).
The method for the PS80 quantification was adapted from Zheng et al. . Samples were diluted 1:4 or 1:10 with DPBS and subsequently heated for 5 min at 99 °C. Afterwards, samples were centrifuged for 5 min at 25,700× g . 190 µL supernatant were mixed with 10 µL of 1 mM bis-ANS and vortexed for 5 s. 60 µL of each sample were analysed in a Varian Cary Eclipse fluorescence spectrophotometer (Agilent Technologies, Santa Clara, CA, USA) using a quartz cuvette at 380 nm excitation and 500 nm emission with both slits set to 5 nm. A calibration curve of PS80 in DPBS allowed the quantification of PS80 between 0.005 and 0.15 mg/mL (R 2 = 0.9988).
The diffusion of mAb and histidine after complete thawing of the solution in the SDD was mapped. Samples were taken after complete thawing of the solution (16 h at 20 °C) as well as after additional 24 h and 48 h. To minimize any possible influence of the removed volume on subsequent results, only 0.25 mL were taken per sample. Samples were obtained from the edges in the top layer, the middle layer and at the bottom. Mixing was avoided by tightly attaching the SDD to the grid in the air-blast climate chamber and taking samples slowly with a 1 mL syringe equipped with a needle. The DynaPro Plate Reader III (Wyatt Technology, Dernbach, Germany) was used to determine diffusion coefficients of mAb and PS80 via dynamic light scattering. Samples with 10 mg/mL mAb or PS80, respectively, were prepared and filtered. 100 µL of each sample was pipetted in triplicates into a 96-well clear bottom plate (Corning Inc., Corning, NY, USA) and ten acquisitions of 5 s at 25 °C taken. The Dynamics V7.8.2.18 software was used for all calculations.
The changes in density after FT were assessed using a portable density meter DMA 35 Standard (Anton Paar Group AG, Graz, Austria). 15 mL samples were prepared according to concentrations found for each layer during the analysis of the concentration gradients after thawing (Table ). After a pre-rinse, density was measured in triplicates at room temperature. The density meter extrapolated results to 20 °C.
Comparison of 3D Temperature Profiles During Thawing The SDD represents the 2 L bottle by achieving an equivalent cumulative thermal history, although the number of control volumes in the CFD simulations was different. Consequently, temperature measurements at equivalent specific position do not necessarily match. Nonetheless, temperature measurements are important to understand the influence of the SDD during thawing and to characterise the thawing behaviour in comparison to the 2 L and the 125 mL PharmaTainer™ bottles. Within this work, the term thawing time is used and defined as the time needed until a TC, placed inside the bottle, reaches 1 °C after the beginning of heating. At this point in time, ice is completely melted at this location and the intermediate plateau ends as no further heat of melting is needed. The term process time is used to describe the time required to reach 17 °C after the beginning of heating. Temperature profiles during thawing were recorded at six positions. In the 125 mL bottle the TCs in the edges (TC 1 and TC 3–6) recorded similar profiles (Fig. ). The thawing and process times at these positions were 1.4–1.9 h and 4.1–4.6 h, respectively. The LPTT was the geometrical centre of the bottle (TC 2). Thawing time at this position was 3.4 h, process time was comparable to the TCs in the edges with 4.5 h. The time needed for thawing was significantly prolonged by the use of the SDD. In addition, the direction of thawing was changed. Thawing started at the exposed edge (TC 3) after 2.8 h and proceeded towards the geometrical centre (TC 2), where the material was thawed after 6.7 h. Similar thawing profiles were recorded for TC 4 and TC 5 near the walls of the SDD with slightly faster thawing within 5.4 h and 6.1 h, respectively. The LPTT was no longer the geometrical centre of the bottle but instead the insulated region (TC 1 and TC 6). After 8.7 h the ice completely melted at both positions. The difference in height between TC 1 and TC 6 did not influence the outcome. The process time was similar for all mapped positions with approximately 12 h. The results for the 2 L bottle in comparison to the SDD are shown in Fig. . Thawing started at the exposed edge (TC 3) and took 1.6 h. At the walls thawing needed 2.3 h (TC 4) and 4.0 h (TC 5) and thereby already exceeded the maximum thawing time at the LPTT in the 125 mL bottle. At half height and half distance thawing was completed after 7.2 h (TC 2). The LPTT was the centre of the 2 L bottle, where complete thawing took 9.3 h (TC 6) and 10.0 h (TC 1). In contrast to the SDD, the difference in height between TC 1 and TC 6 changed thawing time by about 40 min. In agreement with the 125 mL bottle and the SDD, temperature profiles at the various locations merged after complete thawing. Therefore, process times were similar for all position with approximately 13 h. In the 125 mL bottle, thawing started simultaneously at the exposed edges. The walls were heated by the air in the chamber and consequently the edges experienced heat exchange from both adjacent walls. The LPTT was the geometrical centre, the region furthest away from the walls. The latent heat of melting led to a plateau in the observed profiles. As long as the ice was not completely melted, energy was removed from the system to thaw the remaining ice. As soon as no ice was left, the different profiles started to merge. In comparison to the 2 L bottle, thawing time was significantly shorter due to the smaller volume. The 2 L bottle also started to thaw in the region with highest heat exchange, the exposed edge, followed by the walls. Thawing proceeded towards the geometrical centre. The significantly higher volume increased the time for complete thawing at the LPTT by a factor of three in comparison to the smaller bottle. While the absolute difference in height between TC 1 und TC 6 is small in the SDD and the 125 mL bottle, the higher positioning of TC 6 led to a reduced thawing time in the 2 L bottle. Though more or less stagnant air in the bottle’s headspace provides insulation, heat exchange was still high enough to thaw the bottle also from top towards the centre. Thawing the same volume in the SDD took significantly longer than in the 125 mL bottle. The SDD provided sufficient insulation to reach adiabatic conditions at two walls. The thawing direction and the heat exchange could be controlled through the exposed window. Consequently, the LPTT was no longer the geometric centre but instead the insulated edge. The temperature profiles were overall similar to those of the 2 L bottle. It took more time to thaw in the exposed regions in the edges and near the walls in the SDD compared to large-scale, whereas there was no difference at half distance. Only slightly less time was needed to completely thaw and fully process the mAb solution in the SDD despite the enormous difference in fill volume by a factor of 16. As the SDD is designed to match the overall thermal history, we expected some regions to thaw slightly faster and others more slowly. In the SDD thawing was strongly delayed as compared to the 125 mL bottle. Furthermore, the SDD changed the thawing direction so that the LPTT was found in the insulated region and no longer in the geometrical centre. Although thawing and process time were marginally reduced in the SDD, we consider the SDD as representative for the thawing process in large-scale 2 L bottles. In contrast to material consuming large-scale experiments, the SDD needs only a fraction of the volume. Comparison of Concentration Gradients After Thawing Significant changes in concentration throughout the container can be expected after large-scale FT of mAb solutions, potentially affecting mAb stability . A screening study revealed marginal relative concentration gradients upon thawing highly concentrated mAb solution. Thus, higher mAb concentrations would level off changes in concentration upon thawing and impede the validation of the SDD. Consequently, we used a diluted solution to validate the SDD under maximally challenging conditions. We quantified the mAb, the buffering agent histidine, and the surfactant PS80 to picture the distribution of the protein versus its stabilisers. Changes in concentration were expressed as the concentration factor (CF), which is the ratio of the concentration of the sample to the initial concentration. For the mAb CF values between 0.53 and 3.17 were found in the 2 L bottle (Fig. ). These results are in good agreement with previous studies, showing a highly diluted top region and a strongly increased protein concentration at the bottom . The CF steadily decreased with increasing height. In contrast to the three-dimensional cryoconcentration after freezing , changes in concentration after thawing can be reduced to a two-dimensional behaviour. CF values were reproducible within each layer. The only exception was the bottom layer with CF values up to 3.17 in the edges, while the centre was significantly less concentrated (CF 2.15). This deviation in the centre is due to the shape of the bottle’s bottom. The bottom is slightly elevated in the centre and reflects a region between the bottom layer and the layer directly above. Because of this deviation, centre samples were excluded from further data evaluation. Figure shows mean and standard deviation of the CF values detected in the edges of the 2 L bottle. The CF values ranged between 0.54 and 2.86 for mAb, 0.61 and 2.79 for PS80 and 0.58 and 2.54 for histidine. All three components behaved similarly in the top and middle region. Near the bottom, maximum mAb and PS80 concentrations were identical. The histidine CF was slightly higher at the 200 mL liquid level and slightly lower at the bottom compared to the mAb and PS80 CF values. In the 125 mL bottle, CF values between 0.62 and 1.93 for mAb, 0.65 and 1.99 for PS80 and 0.61 and 1.86 for histidine were found (Fig. ). For all three components dilution in the top region was only slightly less in comparison to the 2 L bottle. In contrast, the maximum concentrations at the bottom were significantly underestimated. In the second last layer deviations in histidine concentration compared to mAb and PS80 were still noticeable but less pronounced. CF values between 0.24 and 3.22 for mAb, 0.26 and 3.23 for PS80 and 0.30 and 2.26 for histidine were detected in the SDD (Fig. ). As previously shown, mAb and PS80 concentrations changed identically throughout the entity. Near the top, similar CF values for mAb, histidine and PS80 were detected. In contrast, histidine concentration was lower at the bottom and higher at 25 mL liquid level. In comparison to the 125 mL bottle, the SDD had a significant impact on concentration changes. The SDD increased the concentration gradient for all three components. Compared to the 2 L bottle, the dilution in the SDD was enhanced. The maximum mAb and PS80 concentrations at the bottom were insignificantly higher, revealing a good predictive power for this critical region. Deviations between the buffer and the remaining components were emphasized in the SDD. Maity et al. studied the mechanism of gradient formation during thawing and concluded that the FCM, containing proteins and excipients, melts out of the ice, leaving more or less pure ice behind . Due to its lower density, this ice floats on top and dilutes the top region during melting. At the same time, fractions with higher solute concentrations exhibit increased densities and sink to the bottom of the bottle. The supplementary information illustrates this draining of the FCM out of the ice during thawing of a dyed sucrose solution in a video. The difference between histidine CF and both the PS80 and the mAb CF in the second last and the bottom layer may be related to the difference in diffusion between the small histidine molecules and the larger mAb and PS80 micelles during and after thawing. The diffusion coefficients of mAb and PS80 micelles with 4.7 × 10 –11 m 2 /s and 5.0 × 10 –11 m 2 /s are significantly lower compared to that of histidine with 7.3 × 10 –10 m 2 /s . In order to substantiate this effect, we compared the mAb and histidine concentration in the SDD directly after a thawing period of 16 h and an additional 24 h and 48 h storage period at 20 °C (Fig. ). After 16 h of thawing a strong concentration gradient was evident for the mAb with CF values between 0.36 and 3.32. This outcome is in good agreement with the results described previously. After additional 24 h and 48 h at 20 °C, the gradient only minimally diminished and showed CF values between 0.39 and 3.00 (+ 24 h) and between 0.41 and 2.49 (+ 48 h), indicating that the significant changes of mAb concentration compared to the initial CF value of 1 (representing 5 mg/mL mAb) sustain for more than 2 days. The dilution of histidine in the top region was similar after 16 h of thawing (CF 0.40), but, as described, the maximum CF was lower (CF 2.26). CF values between 0.60 and 1.58 as well as 0.78 and 1.24 after 24 h and 48 h, respectively, emphasized that the concentration gradient of histidine constantly declined. The results demonstrate that the larger mAb diffuses much slower into the top region of the SDD than the smaller histidine. This is substantiated by the mAb to histidine ratio. Immediately after thawing, slight differences were already detectable and the ratio progressively shifted towards histidine near the top and towards mAb near the bottom (Fig. ). In the 125 mL bottle similar diffusion as in the SDD can be assumed. In the larger 2 L bottle, histidine diffusion, as described previously, only became noticeable in the two layers near the bottom due to significantly larger height. Strong changes in concentration of mAb, histidine and PS80 were detected in the 2 L bottle after thawing with a diluted top region and a concentrated bottom region. Faster diffusion of histidine during and after thawing was the driving force for the different behaviour of mAb and PS80 in comparison to the buffering agent. The 125 mL bottle matched concentrations in the top region, but it significantly underestimated the maximum concentrations at the bottom. This outcome can be deduced from a faster thawing in the smaller bottle. Faster thawing allows less time for the FCM to melt out of the ice. Thus, the ice floating on top incorporates a larger fraction of solutes, resulting in a less diluted top and consequently a less concentrated bottom region. In contrast, the thawing behaviour in the SDD is very similar to the 2 L bottle, enabling the release of the solutes from the ice. The slightly stronger gradient that developed in the SDD reflects a worst-case scenario in respect of changes in concentration after FT in the 2 L bottle. Comparison of Density Gradients After Thawing During thawing the solutes melt out of the ice, leaving ice behind that floats on top . At the same time, fractions with higher solute concentrations and increased density sink to the bottom. Although changes in density are often mentioned as an important factor for cryoconcentration after freezing , to our knowledge density gradients after thawing in large-scale bottles have not been analysed yet. Therefore, we assessed density gradients in the 2 L bottle, the SDD and the 125 mL bottle. The large volume that is needed for density measurements would inevitably entail interference and mixing of the different layers in the containers. Consequently, we did not take samples directly from the bottle, but prepared samples according to the concentrations found after thawing in the different devices (see Table ). In the 2 L bottle the density steadily increased with each layer from 1.0003 g/cm −3 at the top to 1.0066 g/cm −3 at the bottom (Fig. ). In the 125 mL bottle the density was 1.0005 g/cm −3 at the top and 1.0042 g/cm −3 at the bottom. The density values in the SDD ranged between 0.9995 and 1.0066 g/cm −3 . The slight density differences between the three systems reflect the differences in the concentrations of the different components. The minimum density in the 2 L bottle was mimicked by the 125 mL bottle, which had a similarly diluted top region. Density near the bottom was markedly underestimated in the 125 mL bottle, reflecting the lower CF values. The SDD showed a more diluted top layer of lower density. Although mAb and PS80 concentrations in the SDD were slightly higher at the bottom, the densities of the bottom layers were identical in the 2 L bottle and the SDD. We assume that the lower histidine concentration balanced the impact of the increased protein and surfactant concentrations in respect of the density. Perspective The SDD used in this study was designed to be universally utilised, regardless of the applied FT process. While our previous study validated the performance of the SDD in respect of freezing , the presented study proved the reliability of the SDD during thawing in an air-blast chamber. However, further studies are needed to validate that the SDD can be used regardless of the thawing method. The design of the specific SDD led to a match of the thermal history compared to a large-scale 2 L bottle. The heat exchange can be adapted by a manipulation of the insulation via adaption of the window of the SDD. SmartFreeZ used this to generate SDDs for rectangular 5 L, 10 L, and 20 L bottles. Studies similar to ours with the 2 L bottle are necessary to proof the reliability of the CFD based concept for the larger bottles.
The SDD represents the 2 L bottle by achieving an equivalent cumulative thermal history, although the number of control volumes in the CFD simulations was different. Consequently, temperature measurements at equivalent specific position do not necessarily match. Nonetheless, temperature measurements are important to understand the influence of the SDD during thawing and to characterise the thawing behaviour in comparison to the 2 L and the 125 mL PharmaTainer™ bottles. Within this work, the term thawing time is used and defined as the time needed until a TC, placed inside the bottle, reaches 1 °C after the beginning of heating. At this point in time, ice is completely melted at this location and the intermediate plateau ends as no further heat of melting is needed. The term process time is used to describe the time required to reach 17 °C after the beginning of heating. Temperature profiles during thawing were recorded at six positions. In the 125 mL bottle the TCs in the edges (TC 1 and TC 3–6) recorded similar profiles (Fig. ). The thawing and process times at these positions were 1.4–1.9 h and 4.1–4.6 h, respectively. The LPTT was the geometrical centre of the bottle (TC 2). Thawing time at this position was 3.4 h, process time was comparable to the TCs in the edges with 4.5 h. The time needed for thawing was significantly prolonged by the use of the SDD. In addition, the direction of thawing was changed. Thawing started at the exposed edge (TC 3) after 2.8 h and proceeded towards the geometrical centre (TC 2), where the material was thawed after 6.7 h. Similar thawing profiles were recorded for TC 4 and TC 5 near the walls of the SDD with slightly faster thawing within 5.4 h and 6.1 h, respectively. The LPTT was no longer the geometrical centre of the bottle but instead the insulated region (TC 1 and TC 6). After 8.7 h the ice completely melted at both positions. The difference in height between TC 1 and TC 6 did not influence the outcome. The process time was similar for all mapped positions with approximately 12 h. The results for the 2 L bottle in comparison to the SDD are shown in Fig. . Thawing started at the exposed edge (TC 3) and took 1.6 h. At the walls thawing needed 2.3 h (TC 4) and 4.0 h (TC 5) and thereby already exceeded the maximum thawing time at the LPTT in the 125 mL bottle. At half height and half distance thawing was completed after 7.2 h (TC 2). The LPTT was the centre of the 2 L bottle, where complete thawing took 9.3 h (TC 6) and 10.0 h (TC 1). In contrast to the SDD, the difference in height between TC 1 and TC 6 changed thawing time by about 40 min. In agreement with the 125 mL bottle and the SDD, temperature profiles at the various locations merged after complete thawing. Therefore, process times were similar for all position with approximately 13 h. In the 125 mL bottle, thawing started simultaneously at the exposed edges. The walls were heated by the air in the chamber and consequently the edges experienced heat exchange from both adjacent walls. The LPTT was the geometrical centre, the region furthest away from the walls. The latent heat of melting led to a plateau in the observed profiles. As long as the ice was not completely melted, energy was removed from the system to thaw the remaining ice. As soon as no ice was left, the different profiles started to merge. In comparison to the 2 L bottle, thawing time was significantly shorter due to the smaller volume. The 2 L bottle also started to thaw in the region with highest heat exchange, the exposed edge, followed by the walls. Thawing proceeded towards the geometrical centre. The significantly higher volume increased the time for complete thawing at the LPTT by a factor of three in comparison to the smaller bottle. While the absolute difference in height between TC 1 und TC 6 is small in the SDD and the 125 mL bottle, the higher positioning of TC 6 led to a reduced thawing time in the 2 L bottle. Though more or less stagnant air in the bottle’s headspace provides insulation, heat exchange was still high enough to thaw the bottle also from top towards the centre. Thawing the same volume in the SDD took significantly longer than in the 125 mL bottle. The SDD provided sufficient insulation to reach adiabatic conditions at two walls. The thawing direction and the heat exchange could be controlled through the exposed window. Consequently, the LPTT was no longer the geometric centre but instead the insulated edge. The temperature profiles were overall similar to those of the 2 L bottle. It took more time to thaw in the exposed regions in the edges and near the walls in the SDD compared to large-scale, whereas there was no difference at half distance. Only slightly less time was needed to completely thaw and fully process the mAb solution in the SDD despite the enormous difference in fill volume by a factor of 16. As the SDD is designed to match the overall thermal history, we expected some regions to thaw slightly faster and others more slowly. In the SDD thawing was strongly delayed as compared to the 125 mL bottle. Furthermore, the SDD changed the thawing direction so that the LPTT was found in the insulated region and no longer in the geometrical centre. Although thawing and process time were marginally reduced in the SDD, we consider the SDD as representative for the thawing process in large-scale 2 L bottles. In contrast to material consuming large-scale experiments, the SDD needs only a fraction of the volume.
Significant changes in concentration throughout the container can be expected after large-scale FT of mAb solutions, potentially affecting mAb stability . A screening study revealed marginal relative concentration gradients upon thawing highly concentrated mAb solution. Thus, higher mAb concentrations would level off changes in concentration upon thawing and impede the validation of the SDD. Consequently, we used a diluted solution to validate the SDD under maximally challenging conditions. We quantified the mAb, the buffering agent histidine, and the surfactant PS80 to picture the distribution of the protein versus its stabilisers. Changes in concentration were expressed as the concentration factor (CF), which is the ratio of the concentration of the sample to the initial concentration. For the mAb CF values between 0.53 and 3.17 were found in the 2 L bottle (Fig. ). These results are in good agreement with previous studies, showing a highly diluted top region and a strongly increased protein concentration at the bottom . The CF steadily decreased with increasing height. In contrast to the three-dimensional cryoconcentration after freezing , changes in concentration after thawing can be reduced to a two-dimensional behaviour. CF values were reproducible within each layer. The only exception was the bottom layer with CF values up to 3.17 in the edges, while the centre was significantly less concentrated (CF 2.15). This deviation in the centre is due to the shape of the bottle’s bottom. The bottom is slightly elevated in the centre and reflects a region between the bottom layer and the layer directly above. Because of this deviation, centre samples were excluded from further data evaluation. Figure shows mean and standard deviation of the CF values detected in the edges of the 2 L bottle. The CF values ranged between 0.54 and 2.86 for mAb, 0.61 and 2.79 for PS80 and 0.58 and 2.54 for histidine. All three components behaved similarly in the top and middle region. Near the bottom, maximum mAb and PS80 concentrations were identical. The histidine CF was slightly higher at the 200 mL liquid level and slightly lower at the bottom compared to the mAb and PS80 CF values. In the 125 mL bottle, CF values between 0.62 and 1.93 for mAb, 0.65 and 1.99 for PS80 and 0.61 and 1.86 for histidine were found (Fig. ). For all three components dilution in the top region was only slightly less in comparison to the 2 L bottle. In contrast, the maximum concentrations at the bottom were significantly underestimated. In the second last layer deviations in histidine concentration compared to mAb and PS80 were still noticeable but less pronounced. CF values between 0.24 and 3.22 for mAb, 0.26 and 3.23 for PS80 and 0.30 and 2.26 for histidine were detected in the SDD (Fig. ). As previously shown, mAb and PS80 concentrations changed identically throughout the entity. Near the top, similar CF values for mAb, histidine and PS80 were detected. In contrast, histidine concentration was lower at the bottom and higher at 25 mL liquid level. In comparison to the 125 mL bottle, the SDD had a significant impact on concentration changes. The SDD increased the concentration gradient for all three components. Compared to the 2 L bottle, the dilution in the SDD was enhanced. The maximum mAb and PS80 concentrations at the bottom were insignificantly higher, revealing a good predictive power for this critical region. Deviations between the buffer and the remaining components were emphasized in the SDD. Maity et al. studied the mechanism of gradient formation during thawing and concluded that the FCM, containing proteins and excipients, melts out of the ice, leaving more or less pure ice behind . Due to its lower density, this ice floats on top and dilutes the top region during melting. At the same time, fractions with higher solute concentrations exhibit increased densities and sink to the bottom of the bottle. The supplementary information illustrates this draining of the FCM out of the ice during thawing of a dyed sucrose solution in a video. The difference between histidine CF and both the PS80 and the mAb CF in the second last and the bottom layer may be related to the difference in diffusion between the small histidine molecules and the larger mAb and PS80 micelles during and after thawing. The diffusion coefficients of mAb and PS80 micelles with 4.7 × 10 –11 m 2 /s and 5.0 × 10 –11 m 2 /s are significantly lower compared to that of histidine with 7.3 × 10 –10 m 2 /s . In order to substantiate this effect, we compared the mAb and histidine concentration in the SDD directly after a thawing period of 16 h and an additional 24 h and 48 h storage period at 20 °C (Fig. ). After 16 h of thawing a strong concentration gradient was evident for the mAb with CF values between 0.36 and 3.32. This outcome is in good agreement with the results described previously. After additional 24 h and 48 h at 20 °C, the gradient only minimally diminished and showed CF values between 0.39 and 3.00 (+ 24 h) and between 0.41 and 2.49 (+ 48 h), indicating that the significant changes of mAb concentration compared to the initial CF value of 1 (representing 5 mg/mL mAb) sustain for more than 2 days. The dilution of histidine in the top region was similar after 16 h of thawing (CF 0.40), but, as described, the maximum CF was lower (CF 2.26). CF values between 0.60 and 1.58 as well as 0.78 and 1.24 after 24 h and 48 h, respectively, emphasized that the concentration gradient of histidine constantly declined. The results demonstrate that the larger mAb diffuses much slower into the top region of the SDD than the smaller histidine. This is substantiated by the mAb to histidine ratio. Immediately after thawing, slight differences were already detectable and the ratio progressively shifted towards histidine near the top and towards mAb near the bottom (Fig. ). In the 125 mL bottle similar diffusion as in the SDD can be assumed. In the larger 2 L bottle, histidine diffusion, as described previously, only became noticeable in the two layers near the bottom due to significantly larger height. Strong changes in concentration of mAb, histidine and PS80 were detected in the 2 L bottle after thawing with a diluted top region and a concentrated bottom region. Faster diffusion of histidine during and after thawing was the driving force for the different behaviour of mAb and PS80 in comparison to the buffering agent. The 125 mL bottle matched concentrations in the top region, but it significantly underestimated the maximum concentrations at the bottom. This outcome can be deduced from a faster thawing in the smaller bottle. Faster thawing allows less time for the FCM to melt out of the ice. Thus, the ice floating on top incorporates a larger fraction of solutes, resulting in a less diluted top and consequently a less concentrated bottom region. In contrast, the thawing behaviour in the SDD is very similar to the 2 L bottle, enabling the release of the solutes from the ice. The slightly stronger gradient that developed in the SDD reflects a worst-case scenario in respect of changes in concentration after FT in the 2 L bottle.
During thawing the solutes melt out of the ice, leaving ice behind that floats on top . At the same time, fractions with higher solute concentrations and increased density sink to the bottom. Although changes in density are often mentioned as an important factor for cryoconcentration after freezing , to our knowledge density gradients after thawing in large-scale bottles have not been analysed yet. Therefore, we assessed density gradients in the 2 L bottle, the SDD and the 125 mL bottle. The large volume that is needed for density measurements would inevitably entail interference and mixing of the different layers in the containers. Consequently, we did not take samples directly from the bottle, but prepared samples according to the concentrations found after thawing in the different devices (see Table ). In the 2 L bottle the density steadily increased with each layer from 1.0003 g/cm −3 at the top to 1.0066 g/cm −3 at the bottom (Fig. ). In the 125 mL bottle the density was 1.0005 g/cm −3 at the top and 1.0042 g/cm −3 at the bottom. The density values in the SDD ranged between 0.9995 and 1.0066 g/cm −3 . The slight density differences between the three systems reflect the differences in the concentrations of the different components. The minimum density in the 2 L bottle was mimicked by the 125 mL bottle, which had a similarly diluted top region. Density near the bottom was markedly underestimated in the 125 mL bottle, reflecting the lower CF values. The SDD showed a more diluted top layer of lower density. Although mAb and PS80 concentrations in the SDD were slightly higher at the bottom, the densities of the bottom layers were identical in the 2 L bottle and the SDD. We assume that the lower histidine concentration balanced the impact of the increased protein and surfactant concentrations in respect of the density.
The SDD used in this study was designed to be universally utilised, regardless of the applied FT process. While our previous study validated the performance of the SDD in respect of freezing , the presented study proved the reliability of the SDD during thawing in an air-blast chamber. However, further studies are needed to validate that the SDD can be used regardless of the thawing method. The design of the specific SDD led to a match of the thermal history compared to a large-scale 2 L bottle. The heat exchange can be adapted by a manipulation of the insulation via adaption of the window of the SDD. SmartFreeZ used this to generate SDDs for rectangular 5 L, 10 L, and 20 L bottles. Studies similar to ours with the 2 L bottle are necessary to proof the reliability of the CFD based concept for the larger bottles.
In a previous study, we characterised the performance of an innovate SDD during freezing, which requires only a fraction of material in order to mimic and understand the large-scale freezing behaviour of mAb solutions in widely used disposable bottles . In this work, we focused on the validation of the SDD in respect of the subsequently following thawing process of mAb solutions. We compared thawing and process times and quantified concentration gradients of mAb, histidine and PS80 in the SDD to 2 L and 125 mL PharmaTainer™ PET bottles. Furthermore, we assessed changes in density. Temperature profiles in the 2 L bottle revealed a directed thawing towards the centre of the bottle. LPTT was determined in the geometrical centre with 10.0 h for complete thawing. The 125 mL bottle showed significantly reduced thawing times due to the smaller volume, also progressing towards the centre. The SDD changed the direction of thawing so that no longer the centre, but the insulated edge was the LPTT. Hereby, the overall thawing time was strongly increased to 8.7 h. Thus, we consider the SDD as representative for thawing profiles in large-scale 2 L bottles. Strong changes in concentration built up during thawing of the large-scale 2 L bottle. The top layer displayed a dilution of approximately 50% compared to the initial mAb, PS80 and histidine concentrations, whereas the concentrations near the bottom increased by a factor of 2.8. The concentration gradients in the 125 mL bottle were significantly smaller for all three components. In contrast, the SDD showed a slightly more pronounced concentration gradient. For mAb and PS80, this gradient persisted for days after thawing due to their slow diffusion. In contrast, the gradient significantly levelled off for the smaller histidine, which can rapidly diffuse into the top region. The density gradients reflected the concentration gradients. The SDD showed an identical density at the bottom and an only slightly lower density near the top compared to the large-scale 2 L bottle. Thus, the SDD can predict thawing processes in large-scale 2 L bottles using only a fraction of the material.
Below is the link to the electronic supplementary material. Supplementary file1 (MP4 7286 kb)
|
How COVID-19 Is Testing and Evolving Our Communication Skills | 4b911307-87bd-46f8-b743-6998b6c2adfe | 7316048 | Health Communication[mh] | Radiotherapy staff on treatment units began to work in personal protective equipment (PPE) such as masks and protective eyewear to minimise the risk of contracting or transmitting the virus. PPE is an uncomfortable, but necessary addition to our physical working environment that affects interactions with colleagues and patients. Communication with patients is much more difficult as voices behind masks are muffled. Patients who are hard of hearing or deaf are unable to read our lips. Interactions between team members are hampered, impeding workflow and adding to tensions. This literal barrier to effective communication is a moral injury we faced on a daily basis through this pandemic. Those of us in the radiotherapy review team have changed to telephone appointments to minimise the time spent by patients in a hospital environment. Telephone reviews are not a new process within health care ; oncology teams use telephone triage when patients report treatment-related side effects. Triage calls need to be efficient in finding out the problem while providing reassurance—they have a standard narrative: what is going on and how long for, what needs to be done, refer on appropriately, document and reflect . Radiotherapy reviews are usually face to face and are scheduled for each patient. They often cover much more than just physical side effects and help build up a rapport with the patient. They can cover healthy lifestyle advice, explore support networks, promote well-being and can help provide reassurance. Face-to-face reviews provide the opportunity to pick up on nonverbal communication cues that you would miss through a telephone. In my opinion, radiotherapy reviews are more in depth than triage calls as they cover all aspects of a patient's care and side effects to help them live with and beyond cancer. Despite the obvious barriers of PPE or a telephone, effective communication is vital to deliver patient-centred care. Communication is the cornerstone of life. It helps us build relationships, to find problems, and find solutions. We learn how to communicate from what our parents or carers teach us. This develops as we experience the world socially and eventually through work. We all have unique journeys and experiences in life to make sense of an ever-changing world. We all have plans and ambitions for the future. COVID-19 is a unique experience for all of us: plans and ambitions on hold, an ever-changing environment and uncertain vision of the future. The daily uncertainty, fear, and changes have been exhausting. We do not know what tomorrow might bring. People with cancer are dealing with all of these COVID-19 uncertainties on top of their diagnosis, treatments, side effects, and everything that goes along with that. To go one step further, patients with cancer can endure some form of lockdown during their cancer journey. The physical experience of lockdown to protect your health is nothing new to patients who have gone through systemic anticancer treatments that leave them with weakened immune systems. The experience of being locked down is new for other patients with cancer. The uncertainty of not knowing whether treatment will be successful or whether they will experience recurrence could be likened to the uncertainty surrounding a future coronavirus vaccine and effective treatments. We often talk to patients about adjusting to a ‘new normal’ after their treatment ends, but patients going through cancer treatment during a pandemic face a whole new set of challenges. In switching to telephone reviews, we were concerned we would not pick up as much as an in-person review. Addressing concerns face to face or through telephone should not really be any different, should it?
The British Association for Counselling and Psychotherapy recently released a set of competencies on telephone and e-counselling . Although not designed for use by radiographers, it has helped us transform how we approach telephone reviews. We have been letting patients know a day in advance to expect a phone call from the review team. On the day of the review, after having read up on the patient's history and treatment plan, I then prepare to call the patient. While I'm pulling on my call centre–style headset, I like to take a moment to look at their ID photo so I have their face in mind when I call them. “Hello …, my name is Naman, I am a member of the review team calling from … We're due to catch up for a treatment review over the phone this morning, is now a good time? Brilliant, this should take around 20–30 minutes. Before we get started, are you sitting comfortably or is there somewhere in your home you want to sit for the duration of this phone call? If you need a pen and paper to write down any notes or to grab a drink, then now would be the time to grab what you need too. … Is there anyone else with you joining in on this call that you feel comfortable having with you? Okay … let's continue …” Our new normal includes setting appropriate boundaries with time, surroundings, who can listen in on the review and awareness. We have to remember to summarise after important aspects of the conversation and give the patient time to write some of the information down. It is important to wait for anyone else in the background to ask any questions. It can be a real flowing conversation over the phone perhaps easier than face-to-mask, but it's not without its challenges…
Keeping the conversation on track can be more difficult; for many patients who are isolating, we are tackling their loneliness as well as their treatment side effects. You might be the only person this patient has spoken to in a while. One patient I recall had no family in touch and no children. The only friends he had were the ones he would meet at the pub which was now off limits. During our chat he began to tell me about his garden. I remember thinking briefly, I have a long list of tasks to get through today and live in a flat with no garden but this is important to him. I think “should I quickly google something about gardening?” Then the patient says his wife used to love gardening and she made the garden look colourful and tidy when she was around…then I heard him crying through the phone. At this point it was difficult not to feel guilty that I could not do more. The use of silence and listening has become even more important; I wonder whether this patient could have opened up like this had we been speaking face to face? Grief is difficult to overcome and stereotypes challenge all of us. The stereotype is that men do not cry. Especially men of a certain generation and especially not to other men. We are all having to live in the present, stop and notice what is around us. There are people like him everywhere, whether they have had cancer treatment or not. It is not my job to talk about a patient's garden but by not cutting him off, he got to talk about someone he loves and misses. It was sad and it upset me, but it was an absolute privilege that while he was in his safe space, at home, he trusted me enough to confide in me and cry over the phone. I'm very lucky to have at least one person in my phone book who would answer and listen to me. The traditional view of face-to-face consultations by a health care professional has changed during this pandemic, potentially forever. Change is inevitable, however uninvited it is. We all have had to adapt at what feels like breakneck speed. Patients and staff have been forced to accept technology. Some of our more ‘mature’ staff have been looking to younger staff to guide them. For some patients, seeing someone “in person” is their way of feeling looked after. It gives them the space to open up and engage with you. Our new normal is a virtual review room where boundaries need to be drawn with the patient. Distractions need to be put aside. There are no more visual cues for you to pick up on and alter your approach. It is no longer eye contact and body language that we look for, but subtle changes in the tone of the patient's voice. It is difficult to empathise and console using just your voice. Use of silence and listening are really important. However, I found that being silent on the call for too long, waiting for the patient to answer a probing question, led a patient to think I was not there anymore and she hung up on me. This was embarrassing and amusing but at least I have learnt that on the phone you cannot be silent for too long. Sporadic murmurs of agreement or support are often just as important as a fully formed answer. Lockdown has meant many of our patients are less busy and are being forced to slow down and refocus on themselves. There is more time to sit and talk. Perhaps the telephone helps erode some of the hesitancy and awkwardness of talking about openly themselves. For example, explaining how many times you open your bowels seems less intimate over the phone than face to face. I would argue we are uncovering more about side effects than we have before. Patients have more time to think about what they feel, what has changed, and what is on their mind. As we navigate the moral injuries that this pandemic deals us, it is possible to also sense some of the positive outcomes. For me, the bond between patients with cancer and radiation therapists has grown stronger. Patients with cancer are risking their lives to come to our department, and when they arrive, they are treated by staff risking theirs. We have changed how we work, what skills we use to work, but behind a mask or telephone, we are still here to help our patients.
|
An Update on Molecular Diagnostics for COVID-19 | fa42baab-0751-4a84-a957-cbcec74ff8fb | 7683783 | Pathology[mh] | Diagnosis is a major aspect in tackling the consequences of any deadly contagious diseases. Diagnostic tests demonstrate the presence or absence of an infectious agent. Early and better diagnosis has helped in limiting fatalities due to highly infectious and contagious diseases in the past . Diagnosis has an empirical role in such diseases that are caused by any novel pathogen for which the population is not pre-immune. COVID-19 is one of such infectious diseases, highly contagious and deadly . COVID-19 was found to be caused by a novel viral pathogen named as SARS-CoV-2. COVID-19 originated in Wuhan city of China and spread to almost the entire world . In January 2020, China reported the outbreak of SARS-CoV-2 that was further declared as pandemic owing to indiscriminate and rapid spread to different parts of world. Among the first 41 cases of COVID-19 reported by Huang et al., most of the patients had a history of exposure to Wuhan wild animal market. Patients had several symptoms like cough, fever, dyspnea, fatigue and radiographic indication of pneumonia . Whole genome sequencing of SARS-CoV-2 led scientists to design testing protocols to detect the pathogen in the affected people and also provided an insight in the phylogenetic study of the virus. It was elucidated that SARS-CoV-2 belongs to the family of beta-coronavirus, which include SARS-CoV and Middle East Respiratory Syndrome (MERS) viruses . The origin of this virus is still a matter of debate; however, bats are thought to be its primary source. SARS-CoV-2 was identified on 3 rd January 2020 in the bronchioalveolar lavage fluid samples taken from a patient in Wuhan, China . As the understanding regarding COVID-19 disease evolved, it was observed that SARS-CoV-2 infection spreads via asymptomatic carriers , which further prompted health workers to increase the diagnostic frequency among the masses. As evident from the previous outbreaks of SARS-CoV in 2003 and MERS in 2012 , a sensitive, specific and rapid diagnosis for COVID-19 are crucial in identifying positive cases, tracing its contacts, finding the source of virus and finally rationalizing the measure of controlling infection. During the early period of the epidemic, complete sequencing of SARS-CoV-2 facilitated the specific primer-designing and laboratory diagnosis of COVID-19 . On 23 rd January 2020 the first protocol of RT-PCR for COVID-19 diagnosis was published. This assay targeted viral genes related to Nucleocapsid (N), RNA-dependent RNA polymerase (RdRp), and Envelope (E) protein. RdRp was found to be most sensitive among these variants. This assay was using two probes; one probe called ‘Pan Sarbeco- probe’ detects SARS-CoV-2, bat-related SARS coronaviruses and other coronaviruses. However, another probe called RdRp-p2 was specific to SARS-CoV-2 only . Jasper Fuk-woo Chen et al. had developed a novel RT-PCR based diagnostic method for COVID-19 called RdRp/Hel, which was found to be more sensitive and specific than the previously published RdRp-p2 method RT-PCR is a highly specific and sensitive diagnostic method for infectious disease detection like COVID-19. As this method is based on the detection of nucleic acids, it enables an early diagnosis of COVID-19 by detecting SARS-CoV-2 RNA in the patient sample . Despite being specific, sensitive and rapid, PCR-based methods lie restricted to a clinical laboratory with sophisticated equipment and trained personnel that limits its use as a Point of care diagnosis . There is a dire need of alternative diagnostic techniques to use as point of care tests in the current pandemic. A dramatic burden on public health and society due to sudden outbreak of SARS-CoV-2 virus limited the use of sequencing and RT-PCR as a tool of rapid diagnosis because both these techniques take much time to generate test result and are affordable to a very limited section of population, thus making it imperative to develop new point of care diagnostic techniques that can compete with RT-PCR in terms of sensitivity and are easily affordable to common man. Many techniques can be exploited to make rapid and affordable testing available to common people that include Point of Care (POC) also known as bed side testing. POC testing is developing as a portable and promising tool in pathogen diagnosis , a major example is the discovery of CRISPR COVID that has been shown to deliver comparable specificity and sensitivity as that of RT-PCR and sequencing based meta genomic diagnostic approaches . Moreover, lateral flow assay (LFA) is one of the latest POC techniques, which works on the principle of fluid flow, and it is employed in immune assays targeting protein, DNA or RNA based samples . According to a study conducted by Banerjee et al. in 2018, these POC based techniques can be utilized for viral detection in patient samples . Other LFA-based techniques like SHERLOCK (Specific High Sensitivity Enzymatic Reporter unLOCKing) developed by Feng Zheng’s lab is aiming to be a fast, cost-effective and sensitive technique for pathogen detection . All-In-One Dual CRISPR-Cas12a (AIOD-CRISPR) is another promising technique known to be fast, sensitive and a very efficient system for pathogen detection . These techniques can be employed to lower the burden of costly diagnostic procedures as well as lowering time consumption during any pandemic disease outbreak. This review will elaborate the diagnostic procedures, which are currently in use and will also highlight other new techniques that can be utilized for rapid and cost-effective diagnosis of COVID-19. High level of viral loads in the upper and lower respiratory tracts have been demonstrated in COVID-19 patients within 5-6 days of the onset of symptoms . For an early diagnosis of COVID-19, nasopharyngeal or oropharyngeal swabs are recommended , however, a single nasopharyngeal swab is a method of choice for health practitioners because patients can easily tolerate it and is safe for handling. To obtain a proper nasopharyngeal swab specimen, the swab must go deep into the nasal cavity eliciting tears in the patient . Collected swabs should be immediately transported using transport media to the diagnostic laboratory, ideally in refrigerated conditions . Patients with severe COVID-19 pneumonia have shown high viral loads in bronchoalveolar lavages, however, nasopharyngeal swabs were not compared in the particular study . These patients have also shown high viral RNA in fecal samples as well . Thus the preferred method of collecting samples from advanced COVID-19 patients is from the stool or the rectal swabs . For the safety of health practitioners and the proper processing of samples, it becomes imperative to take utmost precautions while collecting, transporting and processing the COVID-19 samples. In response to this, the health practitioners must use goggles, N95 respirators, gloves, full sleeve gowns or PPE kits in order to minimize direct contact with the COVID-19 positive patients . During the early time of COVID-19 pandemic, a simple and convenient approach of collecting patient sample was a dare need to replace the painful nasopharyngeal swab collection process. In that scenario, Rutgers Clinical Genomics Laboratory came up with an idea of developing an RT-PCR based technique, which can detect SARS-CoV-2 RNA in self collected saliva samples. They developed an assay kit, which was commercialized as TaqPath™ COVID-19 combo kit. This strategy lowered the risk of contracting infection during sample collection by the health practitioners . Sample collection for protein-based diagnosis like IgG/IgM and LFA, requires patients’ blood samples. shows the schematics of specimen/sample collection for COVID-19 diagnosis as well as various nucleic acid and protein-based diagnostics approach. Molecular diagnostic approaches are appropriate as compared to other syndromic testing approaches because molecular diagnosis targets the genome or proteome of the pathogen thus making it a specific and reliable method of diagnosis . For a novel pathogen sequencing and diagnosis becomes imperative to recognize the nature of the pathogen and its genomic composition. Random amplification and deep sequencing strategies played a critical role in early identification of the SARS-CoV-2, which was further confirmed to be a member of the coronavirus family via different bioinformatics approaches . Using metagenomic sequencing, the first genomic sequencing was conducted for SARS-CoV-2 . On 10 th January 2020, the findings were made public and the sequences submitted to the sequence repository of GenBank . Release of whole genome sequence of SARS-CoV-2 to public databases made it easy for scientists to design primers and probes for conducting laboratory diagnosis of COVID-19 . After the identification of this virus, WHO recommended real time reverse transcription polymerase chain reaction (real time RT-PCR), which is a nucleic acid-based technique, as the frontline diagnostic approach to detect SARS-CoV-2 infection in suspected patients. RT-PCR is highly sensitive and can detect infection at minute levels of pathogen present in the patient sample. It is a nucleic acid-based technique used to amplify a target gene/nucleotide present in a sample, which helps in detecting a specific pathogen and discriminating it from other related pathogens. There are usually two possible ways of performing RT-PCR including one-step assay or two-step assay. One step assay consolidates reverse transcription and PCR amplification in a single tube thus making the process of detection rapid and reproducible; however, this assay provides a lower target amplicon generation. In case of two-step assay the reactions are carried out sequentially in two separate tubes making it time-consuming, but a sensitive assay compared to the one-step assay format . Although eleven nucleic acid-based protocols and eight antibody detection kits have been approved by the National Medical Product Administration (NMPA) in China, PCR was considered as a preferred diagnostic technique. The US Centers for Disease Control and Prevention (CDC) uses a one-step PCR format to diagnose COVID-19 ( https://www.fda.gov/media/134922/download ). The assay is carried out by isolating RNA from the sample and adding to the master mix containing forward and reverse primers, nuclease-free water, reaction mixture (reverse transcriptase, polymerase, nucleotides, magnesium and other additives). A PCR thermocycler is loaded with extracted RNA and mastermix, and the temperature is set to run the PCR reaction ( https://www.fda.gov/media/134922/download ). Cleavage of a fluorophore quencher probe during this reaction generates a fluorescence signal that is detected by the thermocycler, and the progress of amplification is recorded. Positive and negative controls must be included whenever running any RT-PCR reaction, which makes the interpretation of results easy and stringent . RT-PCR and some biosensor based diagnostic kits can detect SARS-CoV-2 nucleotides in fecal samples or sewage water that can be a warning of an infectious disease outbreak in the particular area. SARS-CoV-2 can survive from hours to days in the untreated sewage water . RT-PCR is a sensitive and rapid detection tool in molecular diagnostics. It can detect and amplify even a few copies of specific genomic sequence in a variety of samples, but it depends upon certain aspects to deliver reliable results like proper collection, transport, storage, and processing of samples . It has been used for detection of diverse viruses like Adenovirus, Rotavirus, Astroviruses and many enteric viruses isolated from fecal samples . A Major drawback of this technique is the need for a well-equipped laboratory and technical personnel for handling the experiment, which cannot mitigate the increased demand of rapid testing during pandemic situations like COVID-19 . The RT-PCR based kits are highly expensive and take much time to deliver results thus making it essential to look for other rapid and reliable diagnostic methods . Nucleic Acid Sequence-Based Amplification (NASBA) NASBA is an in vitro amplification process conducted in isothermal conditions. It is a two-step amplification process where the first step is the denaturation and the second step is a polymerase dependent amplification conducted isothermally . Fluorochromes are also added to the reaction in order to make it a real time-based observation. This technique has been further modified as a multiplex process called multiplex real time nucleic acid sequence based amplification (RT-NASBA), which can help in the concurrent detection of different viral infections . RT-NASBA has been proven to be 10–100 times more sensitive than Multiplex RT- PCR, owing to the isothermal conditions where no time is consumed in heating and cooling and production of copies is faster than RT-PCR. RT-NASBA has been previously used for detection of SARS-CoVs infections, and their sensitivity and specificity was seen to be parallel with RT-PCR diagnosis . This technique can be a choice for the rapid diagnosis of COVID-19 during the current pandemic. Loop Mediated Isothermal Amplification (LAMP) LAMP is a diagnostic technique that is comparatively less expensive, much sensitive and rapid than RT-PCR. This technique involves the selective amplification of target nucleic acids at a constant temperature, usually 60°C. In this technique 4 to 6 specifically designed primers are used to detect distinct nucleic acid sequences, moreover there is no requirement of initial template denaturation and reaction time is minimized up to 30 minutes using strand-displacement polymerases . For a colorimetric based analysis LAMP reaction mixture is added with hydroxynepthol blue (HNB) prior to amplification, thus avoiding cross contamination in the future . Lin Yu et al. have used another approach that combines reverse transcription with LAMP diagnostic technique (RT-LAMP) allowing direct detection of SARS-CoV-2 RNA. This technique was further coupled with a pH indicator, which helped in visual readout of the amplification reaction via color change in the reaction mixture . RT-LAMP based assay that targets S gene of SARS-CoV-2 developed by Hu et al., which showed 88.89% sensitivity and high consistency compared with RT-PCR based diagnostic methods. Time required for this assay was around an hour, considerably lower than RT-PCR . LAMP technique avoids the use of costly reagents and instruments, thus helping in reducing the cost of diagnosis with rapid results . Various studies have highlighted the application of LAMP technique in detecting coronavirus infections in patient samples . It was further observed that 9 to 10 copies of viral RNA per reaction were sufficient to detect infection giving a 100 fold higher sensitivity than RT-PCR . Many other reports are also available on the use of LAMP as a diagnostic tool which includes the work carried out by where they demonstrated a two stage LAMP design called COVID-19 Penn-RAMP strategy. This method can be carried out in closed tubes with either colorimetric or fluorescence detection. Their results are not only comparable to RT-PCR but also show 10 fold higher sensitivity when using purified targets . Penn-RAMP is based on the preliminary reaction where outer LAMP primers bind to amplify all targets concurrently through recombinase polymerase. After the first step, a second highly specific reaction initiated. The first reaction specifically uses outer LAMP primers, while the second reaction combines four other RAMP primers. This enhances the sensitivity of this method compared to normal LAMP . For an improved diagnosis, Song’s group developed a two stage isothermal Penn-RAMP double stranded DNA amplification assay that combine LAMP and Recombinase Polymerase Amplification (RPA) in a single tube thus making the assay simple and less time consuming, further this method has been integrated to a smart phone application to make it highly accessible and point of care technique . The challenge related to the LAMP method is the primer optimization and reaction conditions. Point of Care Testing and COVID-19 Diagnosis of infectious diseases at the bed side of a patient where there is no need of sending patient samples to sophisticated labs is called as Point of Care testing (POC). POC has a very important role to play during community contagious infections because it enables communities to diagnose infection without the complex laboratory infrastructure. POC testing is the only option for the remote areas of any community. One of the Point-of-Care testing protocols is the LFA. Xiang J.; Yan M. et al. had used lateral flow antigen assay for COVID-19 diagnosis . LFA is carried out on a strip which is a paper like membrane. The membrane has two lines coated on it among which one line contains gold nanoparticle-antibody conjugate and the other line contains a capture antibody. Patient samples are deposited on the strip. The strip draws proteins from the sample via capillary action. As the sample runs over the first line of the strip, antigen binds the nanoparticle-antibody conjugate. This complex then moves to the second line where the capture antibody immobilizes the complex which makes a red or blue line visible on the strip. Individual gold nanoparticles on the strip are red, however clustered gold nanoparticles in solution show a blue color. For IgM, LFA has shown 57% clinical sensitivity, 69% accuracy, and 100% specificity however LFA has shown 81% sensitivity, 86% accuracy, and 100% specificity for IgG. Clinical sensitivity equal to 82% was observed in the tests that detected both IgG and IgM . These methods have comparatively less sensitivity than PCR and its variants. Researchers have tried to enhance the sensitivity of LFA-based methods by combining it with RT-LAMP and such related techniques. This combination has been used in MERS-CoV detection . Based on the principle of LFA, some techniques are specifically developed to detect viral pathogens. Specific High Sensitivity Enzymatic Reporter Unlocking (SHERLOCK) is one such technique developed by Feng Zheng’s lab. This technique is very cost-effective and rapid in diagnosing viral pathogens. It has been used for detecting Zika and Dengue virus in patient samples . For cold-chain and long-term storage SHERLOCK reagents are subjected to lyophilization. SHERLOCK can detect at least 200 copies/ml of serum/urine for Zika viral RNA . A protocol has been developed by Zhang et al. for diagnosis of COVID-19 using SHERLOCK. It is a three-step diagnostic process. Isothermal amplification, detection and visual readout are the three steps of diagnosis which take less than an hour for the final results . NASBA is an in vitro amplification process conducted in isothermal conditions. It is a two-step amplification process where the first step is the denaturation and the second step is a polymerase dependent amplification conducted isothermally . Fluorochromes are also added to the reaction in order to make it a real time-based observation. This technique has been further modified as a multiplex process called multiplex real time nucleic acid sequence based amplification (RT-NASBA), which can help in the concurrent detection of different viral infections . RT-NASBA has been proven to be 10–100 times more sensitive than Multiplex RT- PCR, owing to the isothermal conditions where no time is consumed in heating and cooling and production of copies is faster than RT-PCR. RT-NASBA has been previously used for detection of SARS-CoVs infections, and their sensitivity and specificity was seen to be parallel with RT-PCR diagnosis . This technique can be a choice for the rapid diagnosis of COVID-19 during the current pandemic. LAMP is a diagnostic technique that is comparatively less expensive, much sensitive and rapid than RT-PCR. This technique involves the selective amplification of target nucleic acids at a constant temperature, usually 60°C. In this technique 4 to 6 specifically designed primers are used to detect distinct nucleic acid sequences, moreover there is no requirement of initial template denaturation and reaction time is minimized up to 30 minutes using strand-displacement polymerases . For a colorimetric based analysis LAMP reaction mixture is added with hydroxynepthol blue (HNB) prior to amplification, thus avoiding cross contamination in the future . Lin Yu et al. have used another approach that combines reverse transcription with LAMP diagnostic technique (RT-LAMP) allowing direct detection of SARS-CoV-2 RNA. This technique was further coupled with a pH indicator, which helped in visual readout of the amplification reaction via color change in the reaction mixture . RT-LAMP based assay that targets S gene of SARS-CoV-2 developed by Hu et al., which showed 88.89% sensitivity and high consistency compared with RT-PCR based diagnostic methods. Time required for this assay was around an hour, considerably lower than RT-PCR . LAMP technique avoids the use of costly reagents and instruments, thus helping in reducing the cost of diagnosis with rapid results . Various studies have highlighted the application of LAMP technique in detecting coronavirus infections in patient samples . It was further observed that 9 to 10 copies of viral RNA per reaction were sufficient to detect infection giving a 100 fold higher sensitivity than RT-PCR . Many other reports are also available on the use of LAMP as a diagnostic tool which includes the work carried out by where they demonstrated a two stage LAMP design called COVID-19 Penn-RAMP strategy. This method can be carried out in closed tubes with either colorimetric or fluorescence detection. Their results are not only comparable to RT-PCR but also show 10 fold higher sensitivity when using purified targets . Penn-RAMP is based on the preliminary reaction where outer LAMP primers bind to amplify all targets concurrently through recombinase polymerase. After the first step, a second highly specific reaction initiated. The first reaction specifically uses outer LAMP primers, while the second reaction combines four other RAMP primers. This enhances the sensitivity of this method compared to normal LAMP . For an improved diagnosis, Song’s group developed a two stage isothermal Penn-RAMP double stranded DNA amplification assay that combine LAMP and Recombinase Polymerase Amplification (RPA) in a single tube thus making the assay simple and less time consuming, further this method has been integrated to a smart phone application to make it highly accessible and point of care technique . The challenge related to the LAMP method is the primer optimization and reaction conditions. Diagnosis of infectious diseases at the bed side of a patient where there is no need of sending patient samples to sophisticated labs is called as Point of Care testing (POC). POC has a very important role to play during community contagious infections because it enables communities to diagnose infection without the complex laboratory infrastructure. POC testing is the only option for the remote areas of any community. One of the Point-of-Care testing protocols is the LFA. Xiang J.; Yan M. et al. had used lateral flow antigen assay for COVID-19 diagnosis . LFA is carried out on a strip which is a paper like membrane. The membrane has two lines coated on it among which one line contains gold nanoparticle-antibody conjugate and the other line contains a capture antibody. Patient samples are deposited on the strip. The strip draws proteins from the sample via capillary action. As the sample runs over the first line of the strip, antigen binds the nanoparticle-antibody conjugate. This complex then moves to the second line where the capture antibody immobilizes the complex which makes a red or blue line visible on the strip. Individual gold nanoparticles on the strip are red, however clustered gold nanoparticles in solution show a blue color. For IgM, LFA has shown 57% clinical sensitivity, 69% accuracy, and 100% specificity however LFA has shown 81% sensitivity, 86% accuracy, and 100% specificity for IgG. Clinical sensitivity equal to 82% was observed in the tests that detected both IgG and IgM . These methods have comparatively less sensitivity than PCR and its variants. Researchers have tried to enhance the sensitivity of LFA-based methods by combining it with RT-LAMP and such related techniques. This combination has been used in MERS-CoV detection . Based on the principle of LFA, some techniques are specifically developed to detect viral pathogens. Specific High Sensitivity Enzymatic Reporter Unlocking (SHERLOCK) is one such technique developed by Feng Zheng’s lab. This technique is very cost-effective and rapid in diagnosing viral pathogens. It has been used for detecting Zika and Dengue virus in patient samples . For cold-chain and long-term storage SHERLOCK reagents are subjected to lyophilization. SHERLOCK can detect at least 200 copies/ml of serum/urine for Zika viral RNA . A protocol has been developed by Zhang et al. for diagnosis of COVID-19 using SHERLOCK. It is a three-step diagnostic process. Isothermal amplification, detection and visual readout are the three steps of diagnosis which take less than an hour for the final results . CRISPR (Clustered Regularly Interspaced Short Palindromic Repeats) has emerged as a game changer in the molecular biology experiments. It has also reshaped the diagnostics of the current time. CRISPR are the DNA sequences found in bacteria and archaea that have been extensively used in gene editing experiments. They play an important role in antiviral defense as the sequences are derived from bacteriophages which have previously infected bacteria . Lots of CRISPR based techniques are currently in use or have a potential to be an option of point of care testing in pathogen diagnosis like SARS-CoV-2. All-In-One Dual CRISPR Cas12a, also termed AIOD-CRISPR is believed to be an ultrasensitive and accurate visual diagnostic technique. In order to initiate a highly specific CRISPR- based detection of nucleic acids a dual crRNA (CRISPR-RNA) is introduced in the reaction mixture . All the components for nucleic acid amplification and CRISPR-based detection are mixed in a single tube and are subjected to isothermal incubation (37°C). This limits contamination because no separate pre-amplification and amplification is needed. AIOD-CRISPR assay has been engineered for detecting SARS-CoV-2 pathogen . AIOD-CRISPR system uses a pair of Cas 12a- crRNA complexes generated by two distinct crRNAs. These complexes bind to corresponding sites closer to the primer recognition sites in the target sequence. The reaction takes place in a single tube that contains: Cas12a-crRNA complex generated separately, recombinase polymerase amplification (RPA) primers, recombinase, ssDNA-FQ reporters, strand displacement DNA polymerase, ssDNA binding protein and target sequences. Binding of Cas12a-crRNA complex to the target sequence leads to the cleavage of nearby ssDNA-FQ via activated Cas 12a this process produces a fluorescence signal . In the study carried out by Xiong Ding et al. a plasmid containing 384 nucleotide N gene cDNA was used as the target to develop AIOD-CRISPR assay. This assay could detect 1.3 copies of SARS-CoV-2 N plasmid in both visual and real time detection within 40 minutes. AIOD-CRISPR has the benefit of being used as a Point of Care diagnosis for pathogenic diseases like COVID-19, owing to its simple use and visual detection by fluorescence or color change . Broughton and group had come up with a CRISPR Cas12 based detection method which they claim to be the most rapid (<40) minute technique among isothermal nucleic acid based POC technique. This CRISPR Cas12-based technique is used to detect SARS-CoV-2 from extracted RNA samples of suspected patients . This technique has been designed to perform simultaneous reverse transcription and isothermal amplification using RT-LAMP to the RNA extracted from nasopharyngeal swabs followed by Cas12-mediated detection of virus. The technique has been named DNA endonuclease targeted CRISPR trans reporter (DETECTR) assay . DETECTR assay showed a comparable accuracy related to RT-PCR. Some key advantages of this assay are isothermal amplification thus avoiding the need of thermocycling, easy to use systems like lateral flow strips, avoiding the use of complex laboratory infrastructure. This technique can be easily mobilized to the hot spots of COVID-19 transmission to ease the diagnosis process. A distinct approach in response to the increasing demand of global rapid testing for detecting SARS-CoV-2 infection was showcased by Rauch et al. which is CRISPR Cas13 based diagnostic method. The test has been named as CREST (Cas13-based, Rugged, equitable, scalable testing) . This method is sensitive and a field deployable procedure to combat diagnosis crisis in pandemic situations. The method is easy and can be used at minimal infrastructure sites. CREST uses easily available protein, low cost thermocyclers and easy to use fluorescent visualizers which makes it a very low cost and easily affordable diagnostic technique. An another portable, accurate and mobile phone based diagnostic assay was reported as an amplification free assay using Cas13a for directly detecting SARS-CoV-2. The assay utilizes patient nasal swab to detect SARS-CoV-2 infection which can be analyzed with the help of a smart phone, this technology has made testing portable and affordable in the low resource areas. It has exhibited limit of detection in 10 fM of RNA target. This technique is much advanced than CRISPR COVID as previously mentioned in this manuscript, which provides qualitative results using isothermal amplification. The sensitivity of other CRISPR based tests have been taken into consideration while developing this assay. Using best combination of crRNAs for entire viral genome makes this assay best fit for point of care diagnosis . A highly accurate, single nucleotide variant detection system developed by IGIB India employs Francisella novicida (Fn-Cas9) based enzymatic readout for nucleotide detection and nucleobase identification. They have named this approach as FnCas9 Editor Linked Uniform Detection Assay (FELUDA) . Viral proteins as antigens or antibodies generated in response to viral infection can serve as the means of diagnosis in viral infectious diseases like COVID-19. Relying on viral proteins for detection is cumbersome as the viral load fluctuates during the course of infection, so antibodies can better serve the diagnosis process . It has been an age old practice to detect specific antigens by antibodies that are directed against these antigenic epitopes using immunoblot assays . COVID-19 can be diagnosed indirectly by detecting antibodies generated in the patient’s blood in a certain window period. However, there is a challenge of cross-reactivity between antibodies generated against SARS-CoV-2 and antibodies against other coronaviruses. A high frequency of cross reactivity was observed by a study carried out by Lv et al. where they had tested plasma samples taken from fifteen COVID-19 patients . According to a study carried out by Bin Ju et al. where the antibody response was characterized in eight COVID-19 patients and around 206 mAbs specific to SARS-CoV-2 receptor binding domain were also isolated. They observed the diverse antibody generation in the set of patients and proposed that such antibodies can serve as prophylactic and therapeutic strategies against COVID-19 . Liu et al. used SARS-CoV-2 IgG/IgM antibody test kit that was manufactured by a Chinese Biotechnology company and further approved by the China Food and Drug Administration. The protocol used around 5 ml of fasting blood of every participant. After collecting serum, the SARS-CoV-2 IgG/IgM were detected. The kit consists of three detection lines, control (C) line/zone, G zone and M zone. C line appears when the sample is flushed over it. A red test line in G and M zones will indicate the SARS-CoV-2 IgG/IgM in the samples. The test is to be repeated if the C line doesn’t appear on the strip . A schematic diagram showing the general protocol of IgG/IgM rapid and Point of Care COVID-19 diagnosis is depicted in . Study conducted by Zhao et al. who had used enzyme linked immunosorbent assay (ELISA) kit developed by Baijing Wantai pharmaceutical company. The kit works on the principle of double antigen sandwich assay. A recombinant antigen containing receptor-binding domain (RBD) of the spike protein of SARS-CoV-2 expressed in mammalian cells was used as an immobilized horseradish peroxidase (HRP)-conjugate antigen. To detect IgM in the patient samples, IgM μ-chain capture method was used, however IgG was detected by using indirect ELISA kit based on recombinant nucleoproteins. They have claimed around 99% sensitivity of IgM and IgG antibodies for this assay . Elevated levels of C reactive-protein and D-dimer and low levels of lymphocytes, blood platelets and leukocytes were shown by Guan et al. in the SARS-CoV-2 infected patients. The challenge in using them as biomarkers is that such markers are also found in different other ailments and abnormalities . Based on different techniques, there are various FDA approved diagnostic kits or kits with an emergency approval in the market. summarizes various such diagnostic kits used for diagnosis of COVID-19. A biosensor-based point of care test strategy was also reported by some research groups where they are claiming rapid antigen detection in saliva of COVID-19 positive patients. The strategy includes developing of a U bent Fiber optic probe over which the gold nanoparticles are immobilized. The anti-N-protein antibody (monoclonal) covalently conjugates with immobilized nanoparticles via thiol-PEG-NHS based binding. This biofunctionalized system are used to detect COVID-19 by applying saliva samples. The device is named as Fiber -Optic Biosensor device and works on the principle of monitoring an optical power loss in light . Another setting of biosensor development for COVID-19 as highlighted by Pooja is an antibody immobilization on either polyaniline or gold nanoparticle coated optical fibers for specific detection of viral proteins in the sample, viral binding with the immobilized antibodies will change the refractive index in the local environment, thus causing change in the light intensity or absorption. Viral surface protein is also immobilized on the surface of optical fiber for the detection of IgG/IgM in the patient samples. The limit of detection claimed by the researchers in such setting was 100 U/ml in an hour . The sensitivity of such biosensor-based settings cannot be compared with RT-PCR sensitivity for a reason that antigen antibody interaction or immune response cannot be considered as precise indicators of pathogen infection or disease propagation. Another reason is that two types of pathogens can elicit same type of immune response which is a drawback of this setting in specificity. During the pandemic, the only way to tackle with the pathogen is to limit its spread which is only possible if the affected people get detected and separated at the earliest. This review has tried to compile the different diagnostic approaches used by academic labs and clinicians to diagnose COVID-19 disease since the identification of SARS-CoV-2 till now. Identification of SARS-CoV-2 in Wuhan, China using sequencing techniques was a major breakthrough because its identification led scientists to progress for its diagnosis and therapeutic studies. The first recommended technique for its diagnosis by the CDC China was RT-PCR. This technique is much used during the current COVID-19 pandemic . It is the same technique that was used to diagnose SARS-CoV in 2002. The lessons learned from SARS-CoV outbreak has guided the very early identification of SARS-CoV-2 infection using sequencing and RT-PCR techniques. During the course of time, since its identification, there has been an immense study on developing rapid nucleic acid-based tests to detect COVID-19 disease among which SHERLOCK, CRISPR and other lateral flow based diagnostic kits are important to mention. These diagnostic approaches are parallelly competing in diagnostic accuracy with the RT-PCR-based diagnosis, however the biosensor related diagnosis needs more understanding and optimization to make them fit for pathogen diagnosis without the need of further confirmations using RT-PCR based -tests. Another approach was the establishment of serological-based diagnostic tests which are comparatively easy to handle and don’t need sophisticated machines or trained personnel like RT-PCR and can be easily used in the home settings so as to decrease exposure of health practitioners who are at high risk of encountering an infection during this pandemic. This manuscript has tried to bring in light various serological-based diagnostic approaches as depicted in that is based on IgG/IgM antibodies. Asymptomatic spread of COVID-19 as reported by some research groups, made it crucial to develop multiplex and Point-of-Care techniques like isothermal amplification, CRISPR-based techniques and microfluidic techniques, so that they can be used to test the majority of the population and isolate infected persons mostly in remote areas, quarantine centers, in developing countries which lack enough resources and skills Data on COVID-19 is evolving very rapidly, and there is no doubt that some of the specifics of this article may change as more studies become available. This review will help a reader to understand the established and other promising techniques in managing pandemic diseases like COVID-19. JI and KUI conceived the idea. KUI wrote the review article. JI has read, corrected the write up, and finalized. All authors contributed to the article and approved the submitted version. The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. |
“I’d just love to hear what the community has to say”: Exploring the potential of community-driven vaccine messaging amongst ethnic minority communities | d04ad06d-a0d7-4458-b52f-7bf793c28d4d | 11542599 | Health Communication[mh] | Research consistently demonstrates community-led communication intervention effectiveness across various health domains. In the context of HIV prevention, a review of community-based health promotion projects suggested that involving community members in designing and implementing health interventions supported better participation in health initiatives. Other programs have used peer education approaches to improve health outcomes. For example, one study found that training lay persons with chronic conditions to deliver patient education programs achieved positive health outcomes, even among patients with diverse health experiences. These interventions often leverage the principles of the Social Ecological Model, Community-Based Participatory Research, and other theoretical frameworks, allowing for targeted and culturally sensitive approaches. The collaborative nature of community-based initiatives not only facilitates participation and dissemination of health information but also fosters a sense of collective ownership, contributing to sustained health improvements. In essence, community-based health communication interventions have emerged as powerful tools for promoting health, addressing disparities, and creating sustainable positive changes within communities. However, there is currently limited understanding of the impact of these community-led strategies regarding the promotion of vaccine readiness in the context of ethnic minority communities. While community-led communication efforts for vaccines, especially COVID-19, are not new, there is limited evidence regarding their role, what support mechanisms are needed, and the receptiveness of different communities. The need for culturally sensitive approaches to enhance vaccine acceptance within diverse communities has been well documented in the literature. In the context of COVID-19 vaccination programs, it has been suggested that culturally acceptable, community-led strategies are essential to combat vaccine hesitancy in regions like West Africa. In Australia, disparities exist in immunization coverage between those born in Australia and those who immigrated. While access factors are understood to contribute to this issue, previous studies have identified gaps in the awareness of ethnic minority community members regarding eligibility and understanding toward recommendations for life-long immunization processes, as well as concerns around trust in engaging with health systems. For some members of these communities, especially those that are from new and emerging communities (i.e. recent immigrants to the country or immigrants from ethnic communities with limited population numbers), working with people who not only share the language but may also appreciate the appropriate ways to share information within the community, may enhance how health promotion material is received. The intersectionality of culture, language, and community dynamics creates unique challenges and opportunities for public health initiatives, making it imperative to explore innovative strategies. To address this gap, this study explores the impact of a community-based intervention aimed at CaLD (Culturally and Linguistically Diverse) community members in Central Eastern Sydney by sampling participants of a community level health education initiative. In this intervention, ethnic minority community members participated in community health information sessions focused on immunization facilitated via the Ethnic Community Services Co-operative (ECSC) Community Centre in Sydney. This study investigates participants’ perceptions following this intervention about receiving and participating in community-led communication efforts regarding the COVID-19 vaccine and other recommended vaccines on the Australia Immunization Program. These insights offer an illustrative example of a culturally sensitive approach to health communication strategies within an ethnic minority community and contribute to a growing body of research in a domain where empirical studies remain limited. Setting and context The Ethnic Community Services Co-operative (ECSC) is a Sydney-based, not-for-profit organization, providing essential services to multicultural families, children, people with disabilities, and seniors, particularly those newly arrived in Australia or to established ethnic minority communities. The ECSC collaborates closely with culturally and linguistically diverse (CaLD) communities across Central and Eastern Sydney. The community it serves includes members from various ethnic backgrounds, such as Chinese, Japanese, Greek, Turkish, and Lebanese, among others. During the COVID-19 pandemic, there was an identified need for community organizations to develop health communication strategies that move beyond disseminating information to developing interventions tailored to the unique needs and contexts of localized people and communities. As one such intervention, the ECSC designed and implemented community-led information sessions, which were supported by a grant from the Central Eastern Sydney Primary Health Network. These sessions aimed to promote healthy living, with a particular focus on vaccination, addressing COVID-19, influenza, and other vaccine-preventable diseases. The workshops were designed and delivered by a dedicated team within the ECSC, with input from the study’s research team and specialized third party facilitators. Recruitment was facilitated through ECSC’s existing networks, relying on community connections to attract both participants and facilitators. The rationale for focusing on adult vaccination was the known gaps in vaccine coverage and the recognized likelihood that training session attendees were more likely to engage with other adults. Participants included people over the age of 18 who live in Central Eastern Sydney and who identify as being from a CaLD community. The sessions included attendees from Chinese, Japanese, Korean, Turkish, Greek, Italian, Lebanese, Egyptian, Indonesian, Filipino and Iraqi communities in Sydney. The sessions were delivered live (time range 90–120 minutes) and were conducted between August and November 2023. During the sessions, a GP or local doctor as well as other qualified health specialists presented information about the strategies to support healthy living, including the need for immunization (COVID-19, influenza, pneumococcal, shingles, etc) and answered questions from the audience. Study design This study employed an interpretivist qualitative approach to address the following question using the Community-Based Participatory Research model (CBPR): What are the perceptions, preferences, and willingness of this ethnic minority community to engage in community-driven vaccine messaging, particularly concerning COVID-19 booster vaccines. The CBPR was utilized to combine knowledge from stakeholders and action to create positive and lasting social change. In-depth, semi-structured interviews were conducted with participants from ethnic minority communities who attended community health information sessions focused on immunization at the Ethnic Community Services Co-operative (ECSC) Community Centre in Sydney. The study researchers were primarily involved in the planning phase of the information sessions, collaborating with the ECSC to ensure culturally sensitive and community-driven content, and assessing participants’ perceptions and experiences post-session through this study. The researchers did not directly participate in delivering the information sessions. The study adheres to the SPQR guidelines. Participants and sampling All information session attendees were invited into the study by the ECSC. Following each information session, an expression of interest form was distributed to attendees to participate in an interview for this study, which included the Participant Information Statement and Consent (PISCF). Eligible participants were aged 18 years and above, self-identified as being from an ethnic minority background, and could complete the consent form and interview in English. The study funding did not allow for the interviews to be conducted in any other language, which is acknowledged as a limitation. However, attempts were made to utilize clear and simplified English were possible. Efforts were made to ensure diversity in age, gender, and location to ensure adequate spread of representation, perspective, and therefore, richness of data. Participants were interviewed at the earliest possible time after attending the health community forum (range: one day to 4 weeks after attending a session) using a question guide. Data collection A trained data collector conducted interviews using a pretested semi-structured interview guide. All interviews occurred within 4 weeks after the participants attended the information session. Only those who attended the workshops in October and November 2023 were invited to participate to ensure consistency of the transpired time between their exposure to the intervention and the actual interview. The interviews focused on gathering information about how individuals typically access health-related information, particularly regarding immunization, including COVID-19 vaccines . It delved into various aspects such as sources of trusted information, potential barriers in accessing vaccine information within communities (such as language barriers), effective communication approaches, and the role of community members in passing on accurate information and combating misinformation during outbreaks. Interviews were audio-recorded with participant consent, and a $50 gift voucher was provided as reimbursement. Interviews were concluded when all questions had been satisfactorily answered. Since the study focused on participants who attended specific community-based health information sessions in a community center in Sydney, and given the quality of the interview retrieved data and corresponding inductive analysis, it was determined that the information power and data saturation were satisfactorily high to address the aims outlined in the study design. , The following table details the interview prompts utilized in every interview. Data analysis Thematic analysis, based on Braun and Clarke’s six-step framework, was utilized. Author 1 directed the analysis, revising transcripts for emerging concepts and themes. Author one harnessed Williams and Moser’s approach to coding and thematic exploration in qualitative research by engaging in open coding, axial coding, and selective coding to progress to theory and thematic development. Trustworthiness of the data and minimization of author subjectivity was ensured using methodological triangulation, member checking by clarifying the information obtained from the participants through summarizing and obtaining their validation during the interview itself, and the involvement of the senior author, HS, who, as an experienced qualitative researcher, to ensure confirmability by reviewing the coding of the interview transcripts. The early development of themes was guided by the answers to the questions asked in the interview. A preliminary coding scheme was then constructed iteratively, and NVivo 10 software facilitated coding, noding, data categorization, and comparison of participant views. The second and final investigators of the study independently reviewed in the data and thematic formulation to ensure consistency through member checking and triangulation of data. The investigators participated in the iterative process of data analysis and interpretation to identify and agree upon themes. Any inconsistencies were addressed in a final discussion and consensus. The Ethnic Community Services Co-operative (ECSC) is a Sydney-based, not-for-profit organization, providing essential services to multicultural families, children, people with disabilities, and seniors, particularly those newly arrived in Australia or to established ethnic minority communities. The ECSC collaborates closely with culturally and linguistically diverse (CaLD) communities across Central and Eastern Sydney. The community it serves includes members from various ethnic backgrounds, such as Chinese, Japanese, Greek, Turkish, and Lebanese, among others. During the COVID-19 pandemic, there was an identified need for community organizations to develop health communication strategies that move beyond disseminating information to developing interventions tailored to the unique needs and contexts of localized people and communities. As one such intervention, the ECSC designed and implemented community-led information sessions, which were supported by a grant from the Central Eastern Sydney Primary Health Network. These sessions aimed to promote healthy living, with a particular focus on vaccination, addressing COVID-19, influenza, and other vaccine-preventable diseases. The workshops were designed and delivered by a dedicated team within the ECSC, with input from the study’s research team and specialized third party facilitators. Recruitment was facilitated through ECSC’s existing networks, relying on community connections to attract both participants and facilitators. The rationale for focusing on adult vaccination was the known gaps in vaccine coverage and the recognized likelihood that training session attendees were more likely to engage with other adults. Participants included people over the age of 18 who live in Central Eastern Sydney and who identify as being from a CaLD community. The sessions included attendees from Chinese, Japanese, Korean, Turkish, Greek, Italian, Lebanese, Egyptian, Indonesian, Filipino and Iraqi communities in Sydney. The sessions were delivered live (time range 90–120 minutes) and were conducted between August and November 2023. During the sessions, a GP or local doctor as well as other qualified health specialists presented information about the strategies to support healthy living, including the need for immunization (COVID-19, influenza, pneumococcal, shingles, etc) and answered questions from the audience. This study employed an interpretivist qualitative approach to address the following question using the Community-Based Participatory Research model (CBPR): What are the perceptions, preferences, and willingness of this ethnic minority community to engage in community-driven vaccine messaging, particularly concerning COVID-19 booster vaccines. The CBPR was utilized to combine knowledge from stakeholders and action to create positive and lasting social change. In-depth, semi-structured interviews were conducted with participants from ethnic minority communities who attended community health information sessions focused on immunization at the Ethnic Community Services Co-operative (ECSC) Community Centre in Sydney. The study researchers were primarily involved in the planning phase of the information sessions, collaborating with the ECSC to ensure culturally sensitive and community-driven content, and assessing participants’ perceptions and experiences post-session through this study. The researchers did not directly participate in delivering the information sessions. The study adheres to the SPQR guidelines. All information session attendees were invited into the study by the ECSC. Following each information session, an expression of interest form was distributed to attendees to participate in an interview for this study, which included the Participant Information Statement and Consent (PISCF). Eligible participants were aged 18 years and above, self-identified as being from an ethnic minority background, and could complete the consent form and interview in English. The study funding did not allow for the interviews to be conducted in any other language, which is acknowledged as a limitation. However, attempts were made to utilize clear and simplified English were possible. Efforts were made to ensure diversity in age, gender, and location to ensure adequate spread of representation, perspective, and therefore, richness of data. Participants were interviewed at the earliest possible time after attending the health community forum (range: one day to 4 weeks after attending a session) using a question guide. A trained data collector conducted interviews using a pretested semi-structured interview guide. All interviews occurred within 4 weeks after the participants attended the information session. Only those who attended the workshops in October and November 2023 were invited to participate to ensure consistency of the transpired time between their exposure to the intervention and the actual interview. The interviews focused on gathering information about how individuals typically access health-related information, particularly regarding immunization, including COVID-19 vaccines . It delved into various aspects such as sources of trusted information, potential barriers in accessing vaccine information within communities (such as language barriers), effective communication approaches, and the role of community members in passing on accurate information and combating misinformation during outbreaks. Interviews were audio-recorded with participant consent, and a $50 gift voucher was provided as reimbursement. Interviews were concluded when all questions had been satisfactorily answered. Since the study focused on participants who attended specific community-based health information sessions in a community center in Sydney, and given the quality of the interview retrieved data and corresponding inductive analysis, it was determined that the information power and data saturation were satisfactorily high to address the aims outlined in the study design. , The following table details the interview prompts utilized in every interview. Thematic analysis, based on Braun and Clarke’s six-step framework, was utilized. Author 1 directed the analysis, revising transcripts for emerging concepts and themes. Author one harnessed Williams and Moser’s approach to coding and thematic exploration in qualitative research by engaging in open coding, axial coding, and selective coding to progress to theory and thematic development. Trustworthiness of the data and minimization of author subjectivity was ensured using methodological triangulation, member checking by clarifying the information obtained from the participants through summarizing and obtaining their validation during the interview itself, and the involvement of the senior author, HS, who, as an experienced qualitative researcher, to ensure confirmability by reviewing the coding of the interview transcripts. The early development of themes was guided by the answers to the questions asked in the interview. A preliminary coding scheme was then constructed iteratively, and NVivo 10 software facilitated coding, noding, data categorization, and comparison of participant views. The second and final investigators of the study independently reviewed in the data and thematic formulation to ensure consistency through member checking and triangulation of data. The investigators participated in the iterative process of data analysis and interpretation to identify and agree upon themes. Any inconsistencies were addressed in a final discussion and consensus. Fifteen interviews were conducted with participants with the following demographic characteristics (See ), who attended the Community Health Information Sessions . The thematic findings have been summarized into four categories: Trusted voices and sources, providing opportunities for conversations, the need for respectful dialogue and proclivity for passing on vaccine information. Trusted voices and sources Participants emphasized the importance of the information source over the mode of communication. Some explicitly noted that “there is a critical role trusted voices play in shaping public perception” and that they “should take precedence in any communication strategy.” Participants highlighted that while the communication approach is undoubtedly important and should be tailored to the diverse information-seeking habits identified, it should not overshadow the central focus on credible sources. Comments were made regarding the exposure to the health industry that community members face, which may influence their perspectives on the importance of immunization at a community level. Through the interviews, we identified a diverse landscape regarding information seeking, emphasizing the personalized nature of vaccine-related knowledge acquisition. While some participants gravitated toward authoritative health platforms, such as government websites and professional bodies, others were more open to hearing or receiving information about vaccines from sources. Participants consistently highlighted the importance of “personalized and genuine communication” in shaping public perceptions. The use of credible health professionals to endorse information emerged as a recurrent theme, emphasizing the significance of leveraging the expertise of reputable individuals to communicate key facts and steps in a clear and simple manner, thereby enhancing the effectiveness of the message. Furthermore, ad campaigns that provided visibility to vaccination details, side effects, and eligibility were identified as useful tools. Digital communication channels were underscored, particularly those catering to younger age groups. However, the need to go beyond generic messaging became evident, with participants advocating for identifying community champions and their education and leveraging their influence to ensure consistent messaging and minimize confusion. Another participant (47, M) demonstrated this concept when asked where they receive their vaccine information by answering, “Definitely not the government. I think I probably like to collect a range of opinions. Then I would say I value spiritual leaders and doctors that I know and respect and love.” Specific suggestions included targeting key professions like teachers and engaging respected figures, such as elders in religious communities. Additionally, the idea of establishing a central body responsible for organizing, funding, and executing vaccination campaigns for multicultural communities was proposed, aiming for efficient management of logistics and administration. Providing opportunities for conversations Participants stressed the essential nature of personal engagement and “opportunities for people to express their worries and concerns about vaccines” in community meetings, discussions, and small groups. The focus on peer-to-peer conversations and community-based discussions highlighted the importance of creating spaces for individuals to share experiences and hear from medical experts. While there was a consensus on the potential benefits of open discussions within churches, recognizing the topic’s sensitivity, participants also acknowledged these environments as trustworthy for facilitating dialogue. The suggestion for inclusive spaces in schools, universities, workplaces, and churches was made, emphasizing education and debate as effective tools for fostering understanding. The need for respectful dialogue Transparent communication about immunization’s purpose and realistic goals, coupled with sharing percentages of side effects, was highlighted as an effective strategy. Participants stressed the importance of “recognizing and addressing genuine concerns rather than dismissing them,” emphasizing that divergent opinions should not hinder productive discussions. These themes emerged with many participants, including some who self-identified COVID-19 vaccine hesitant, with one participant stating: It just never felt like there was space for open debate because whether that was at uni or whether that was at work or in any other institution, it felt like if you expose yourself as vaccine-hesitant, people were going to ostracise you. There was never an incentive or debate to discuss. I really feel like a lot of my opinions could be really well challenged if people were willing to nonjudgmentally debate it with me. (28, F) This comment, representing a collection of participants who felt disrespected during times of discourse regarding vaccines demonstrates that in some cases, individuals felt they were seeking opportunities to partake in a more balanced, respectful conversation about the topic, without fear of incidentally sending or receiving sentiments of insult or dismissal. “Transparent communication about vaccine risks” was advocated as another critical element, with participants expressing a desire for politicians and officials to be honest and forthright in addressing concerns. Participants also emphasized that this should include providing realistic assessments of risks and benefits and making side effects clear to the public. Another participant (47, M) stated “I guess I want open discussions and being respectful of all people’s opinions and their want to choose what they want to do, but also having open discussions about research and what sources of information are being used to base in a particular opinion, and whether that’s reliable or not.” Proclivity for passing on vaccine information Participants offered valuable insights into their motivations, challenges, and strategies when disseminating vaccine-related information. Many noted the nuanced, complex considerations taken when deliberating their desires and capabilities to discuss vaccines, with one participant (28, F) stating, “I think [discussing vaccines] has always been challenging, it has never seemed to be a very smooth process, but I’m trying to pinpoint where that comes from.” Some responses highlighted the complexities individuals face as they navigate the responsibility of being information conduits within their social circles. Some participants expressed a keen sense of duty and eagerness to share accurate information, driven by a desire to contribute to community well-being. Others disclosed challenges related to potential hesitations and societal norms, noting the pressures that their social circles place on them to decide to vaccinate, and fears of further misinformation spread, triggered by their attempts to dispel and debunk rumors. Participants emphasized the importance of the information source over the mode of communication. Some explicitly noted that “there is a critical role trusted voices play in shaping public perception” and that they “should take precedence in any communication strategy.” Participants highlighted that while the communication approach is undoubtedly important and should be tailored to the diverse information-seeking habits identified, it should not overshadow the central focus on credible sources. Comments were made regarding the exposure to the health industry that community members face, which may influence their perspectives on the importance of immunization at a community level. Through the interviews, we identified a diverse landscape regarding information seeking, emphasizing the personalized nature of vaccine-related knowledge acquisition. While some participants gravitated toward authoritative health platforms, such as government websites and professional bodies, others were more open to hearing or receiving information about vaccines from sources. Participants consistently highlighted the importance of “personalized and genuine communication” in shaping public perceptions. The use of credible health professionals to endorse information emerged as a recurrent theme, emphasizing the significance of leveraging the expertise of reputable individuals to communicate key facts and steps in a clear and simple manner, thereby enhancing the effectiveness of the message. Furthermore, ad campaigns that provided visibility to vaccination details, side effects, and eligibility were identified as useful tools. Digital communication channels were underscored, particularly those catering to younger age groups. However, the need to go beyond generic messaging became evident, with participants advocating for identifying community champions and their education and leveraging their influence to ensure consistent messaging and minimize confusion. Another participant (47, M) demonstrated this concept when asked where they receive their vaccine information by answering, “Definitely not the government. I think I probably like to collect a range of opinions. Then I would say I value spiritual leaders and doctors that I know and respect and love.” Specific suggestions included targeting key professions like teachers and engaging respected figures, such as elders in religious communities. Additionally, the idea of establishing a central body responsible for organizing, funding, and executing vaccination campaigns for multicultural communities was proposed, aiming for efficient management of logistics and administration. Participants stressed the essential nature of personal engagement and “opportunities for people to express their worries and concerns about vaccines” in community meetings, discussions, and small groups. The focus on peer-to-peer conversations and community-based discussions highlighted the importance of creating spaces for individuals to share experiences and hear from medical experts. While there was a consensus on the potential benefits of open discussions within churches, recognizing the topic’s sensitivity, participants also acknowledged these environments as trustworthy for facilitating dialogue. The suggestion for inclusive spaces in schools, universities, workplaces, and churches was made, emphasizing education and debate as effective tools for fostering understanding. Transparent communication about immunization’s purpose and realistic goals, coupled with sharing percentages of side effects, was highlighted as an effective strategy. Participants stressed the importance of “recognizing and addressing genuine concerns rather than dismissing them,” emphasizing that divergent opinions should not hinder productive discussions. These themes emerged with many participants, including some who self-identified COVID-19 vaccine hesitant, with one participant stating: It just never felt like there was space for open debate because whether that was at uni or whether that was at work or in any other institution, it felt like if you expose yourself as vaccine-hesitant, people were going to ostracise you. There was never an incentive or debate to discuss. I really feel like a lot of my opinions could be really well challenged if people were willing to nonjudgmentally debate it with me. (28, F) This comment, representing a collection of participants who felt disrespected during times of discourse regarding vaccines demonstrates that in some cases, individuals felt they were seeking opportunities to partake in a more balanced, respectful conversation about the topic, without fear of incidentally sending or receiving sentiments of insult or dismissal. “Transparent communication about vaccine risks” was advocated as another critical element, with participants expressing a desire for politicians and officials to be honest and forthright in addressing concerns. Participants also emphasized that this should include providing realistic assessments of risks and benefits and making side effects clear to the public. Another participant (47, M) stated “I guess I want open discussions and being respectful of all people’s opinions and their want to choose what they want to do, but also having open discussions about research and what sources of information are being used to base in a particular opinion, and whether that’s reliable or not.” Participants offered valuable insights into their motivations, challenges, and strategies when disseminating vaccine-related information. Many noted the nuanced, complex considerations taken when deliberating their desires and capabilities to discuss vaccines, with one participant (28, F) stating, “I think [discussing vaccines] has always been challenging, it has never seemed to be a very smooth process, but I’m trying to pinpoint where that comes from.” Some responses highlighted the complexities individuals face as they navigate the responsibility of being information conduits within their social circles. Some participants expressed a keen sense of duty and eagerness to share accurate information, driven by a desire to contribute to community well-being. Others disclosed challenges related to potential hesitations and societal norms, noting the pressures that their social circles place on them to decide to vaccinate, and fears of further misinformation spread, triggered by their attempts to dispel and debunk rumors. The results of this study shed light on the multifaceted nature of vaccine-related information seeking and dissemination, emphasizing the need for comprehensive strategies that extend beyond merely providing information. The findings suggest that although it is important to acknowledge the diverse sources individuals turn to for information from authoritative health platforms to interpersonal discussions and digital channels, it is equally important to recognize that having access to information alone is insufficient to drive behavior change. This is particularly true in the face of misinformation, which can be difficult to debunk and may be interpreted in line with long-held values. To address this, adequate support and communication from health authorities and governments is essential. Our study highlighted how social relationships play a significant role in shaping individuals’ vaccination decisions. The findings of this paper contribute to existing research that finds social influence, both from close social contacts and broader social networks, has a substantial impact on vaccination decision-making. , For instance, Bell et al. emphasized that vaccination decisions are often influenced by interpersonal relationships, especially between caregivers and extended family members, such as mothers-in-law or community elders. This suggests that interventions should adopt a “whole family” approach, fostering collaboration among family members to support vaccination decisions. Individuals often align their vaccination choices with those of their family members and social network members, indicating the strong influence of social norms on vaccination attitudes and behaviors. , Moreover, positive peer effects have been observed to influence individuals’ decisions to vaccinate, demonstrating the importance of peer groups undertaking training sessions at a community level, such as the one included in this study. Our research has reinforced the notion that social trust acts as a key factor in vaccination decisions. This aligns with previous studies which suggest that trust in the information provider and general trust in others have been linked to vaccination acceptance. , Generalized trust is expected to increase individuals’ willingness to accept vaccination, highlighting the importance of trust in promoting immunization. Community orientation, which encompasses prosocial attitudes and behaviors, has also been previously shown to impact vaccination decisions. Studies suggest that prosocial health attitudes contribute to individuals’ decisions regarding vaccination. This phenomenon was also observed in this study, suggesting that public health promotional strategies should consider the prosocial status of every population being administered an initiative, with such a bespoke approach ensuring that the needs and specific preferences of each group of recipients. Encouraging altruism in making vaccine decisions has been shown to improve influenza vaccination uptake. Framing vaccination as a prosocial act that benefits both the vaccinated individual and the society can elicit measures of altruism, reciprocity, and trust, further influencing vaccination decisions. Both our study and others have reinforced the notion that partnerships and organizations play a crucial role in creating trusted spaces for dialogue about vaccines, with the need for strategies to manage rumors and replicate successful community engagement interventions. Community organizations, such as the ECSC, maintain a unique position in facilitating and promoting community engagement. By leveraging the established relationships between the organization and the communities they support, coupled with their understanding of community dynamics, such organizations can effectively disseminate health information and resources, thereby addressing gaps in service provision. This is particularly crucial in contexts where formal health services may be met with skepticism or lack of engagement from community members. These community organizations can further support trust in health systems and drive demand, by partnering with local bilingual general practitioners (GPs), Practice Nurses (PNs) and other healthcare providers to present at forums, along with community connectors to facilitate discussions and address questions. The presence of healthcare providers in these settings can help address challenging questions about safety and effectiveness of vaccines, addressing other access concerns and provide accurate information. Moreover, the presence of community connectors can lead to improved health literacy among community members. Studies have shown that when community connectors are involved, there is a notable increase in the community’s ability to make informed health decisions. This is especially important in culturally diverse settings, where tailored health information can significantly impact health behaviors and outcomes. Organizations like the ECSC, who have established relationships with the community and experience with culturally and linguistically appropriate service provision, can play a pivotal role in ensuring that health information is not only accessible but also culturally relevant, thereby fostering better health practices within the community. These relationships and localized experiences are particularly important in culturally diverse communities, as they offer a more grounded perspective on the dynamics and nuances of cultural values and expectations of health education and care across community members. The health promotion sessions conducted within the parameters of this study were held at set times and locations, which may not necessarily suit all communities. Instead of this model, consideration may be given to the use of partnerships with mobile health clinics (MHCs) for the delivery of health promotion session around vaccination. In the past, these have been shown to be instrumental in bridging the gap between communities and the larger healthcare system. These clinics not only provide direct services but also serve as connectors to broader healthcare and social service systems, ensuring equitable access to care. By integrating vaccine mobile clinics into these forums organized by community organizations, accessibility barriers to vaccinations can be reduced. In this study, we focused on ethnic minority communities in Central Eastern Sydney, a region characterized by diverse cultural and linguistic backgrounds. The locality of this study played a critical role in shaping the findings, particularly in how trusted information sources and relationships with community leaders influenced vaccine messaging. For instance, several participants expressed a preference for hearing from spiritual leaders and doctors with whom they had an established relationship, highlighting the importance of trust and familiarity in health communication. This is a context-specific finding, as these community leaders already hold a prominent and respected role within their cultural networks in this area. Such a dynamic might differ in other communities where these relationships may not be as well-established or where different figures hold influence. Examples of these communities could include those that are mobile or newly emerging communities in a country. This study has demonstrated how altruism, social trust, and community orientation play crucial roles in shaping individuals’ decisions regarding vaccination. Research has shown that altruism, defined as accepting a net cost to oneself to benefit others, significantly influences vaccination decisions. Altruism shifts individuals’ focus from self-interest toward the well-being of the community, leading to increased acceptance of vaccination. Moreover, studies have indicated that altruistic motives are associated with higher vaccination acceptance rates, as they contribute to reducing the total cost, morbidity, and mortality for the community. More research is needed to determine more concretely how these naturally occurring positive sentiments can be generated and utilized to deliver positive health outcomes within communities. Numerous studies have established that vaccine hesitancy is often linked to distrust in governmental institutions, which can be exacerbated by cultural and socio-economic factors. For instance, Mohammed et al. emphasize the importance of building trust within communities to enhance vaccine acceptance, suggesting that intercultural health advocacy can bridge gaps in understanding and foster a supportive environment for vaccination, This aligns with findings from Chen et al., who argue that trust in government and healthcare professionals significantly influences individuals’ willingness to receive vaccines, particularly during health crises like the COVID-19 pandemic. Moreover, Chung et al. highlight that distrust in government correlates with lower vaccination rates, particularly among adolescents in Hong Kong, where social unrest has led to increased skepticism toward governmental measures. This phenomenon is not isolated; Juárez et al. found that trust in official information sources positively impacts vaccine uptake among Indigenous populations, accentuating the necessity for effective communication strategies tailored to diverse cultural contexts Similarly, Ben – Ezra et al. report that trust in government is a significant predictor of vaccine willingness across various countries, indicating a universal trend where governmental credibility directly affects public health outcomes. This study has demonstrated that leveraging community connectors and boundary spanners can be instrumental in effectively disseminating accurate health information at a community level. These individuals, such as Community Health Workers (CHWs), or, in our case, the ECSC, which service and fund the capacity for CHWs to engage with community members, play a vital role in bridging the gap between communities and healthcare systems, facilitating the flow of information and promoting health. By tapping into their unique position and utilizing their relationships within the community, these connectors can help ensure that health information is effectively communicated and understood. Additionally, community forums, such as the ECSC’s health sessions, can serve as valuable platforms for knowledge sharing and education. These forums provide opportunities for community members to engage in discussions, ask questions, and access reliable health information, ultimately empowering individuals to make informed decisions about their health. Through collaborative efforts involving community connectors, boundary spanners, and community forums, accurate health information can be effectively disseminated, leading to improved health outcomes and increased community engagement. By interviewing ethnic minority community members, we gained a rich understanding of the vaccine-focused communication approaches that resonate with them. The study’s findings indicate that participants valued Community-Led Vaccine Information Sessions and community-based communication strategies when making decisions about COVID Boosters and other vaccines. Participants stressed the value of credible health professionals in delivering information effectively, endorsing a personable approach to disseminate key facts. Participants also expressed a desire for detailed vaccine information, suggesting the creation of comprehensive resources such as infographics or spreadsheets to aid decision-making. The findings align with ongoing discussions at the WHO regarding prioritizing information sources over communication approaches. However, deeper exploration is needed to understand community members’ perceptions toward passing on information and to ensure that our findings contribute novel insights rather than replicating existing knowledge. The interviews were completed with a select group of participants from a certain population and geographic location, that being Central Eastern Sydney, and who were 18 years or older, identified as being part of a CaLD community and having attended one of the ECSC health sessions. It is, therefore, impossible to dismiss the scenario that other important themes were not included in the study. Due to the mildly contentious nature of the subject matter, individuals may potentially have shared false insights in order to convey perceived “acceptable” findings for the interviewer or otherwise place emphasis on topics that did not actually align with their actual thoughts on the topic. This study sought to address the explored topic’s sentiments for this particular population in a community context. Due to time constraints, the authors acknowledge that more robust socio-demographic and language assessment of interviewees and attendees could have been collected. We suggest that future research can address this limitation and extend the findings of this study by more closely attending to the specific multicultural, intercultural, and linguistic dynamics of this and other CaLD communities, including how these change over generations. Despite efforts to maintain objectivity, the characteristics and reflexivity of the researchers may have influenced the study. However, rigorous protocols were implemented to mitigate these biases and ensure the integrity of the data. The authors acknowledge that they may have inadvertently and without conscious knowledge brought their own biases into the analysis phase, noting that some of the authors are also members of CaLD communities across Sydney, although this was consciously avoided to the best of their abilities. Our analysis reveals a diverse landscape where individuals draw from various sources to form their understanding of vaccination. While authoritative health platforms are pivotal, our findings highlight the importance of tailored communication strategies that acknowledge and accommodate the multifaceted information ecosystem individuals navigate. Trusted voices, particularly well-informed healthcare professionals, and community connectors, emerge as crucial influencers in shaping public perceptions. Moreover, fostering open dialogue in inclusive spaces is key to addressing concerns and fostering understanding. Transparent communication, respectful dialogue, and emphasis on credible sources are paramount in effective communication strategies. It is prudent to prioritize the trustworthiness of information sources while adapting communication approaches to cater to diverse information-seeking habits. Ultimately, by embracing these insights and implementing targeted strategies, we can facilitate informed decision-making and enhance vaccine acceptance within our communities. |
European Pediatric Societies Call for an Implementation of Regular Vaccination Programs to Contrast the Immunity Debt Associated to Coronavirus Disease-2019 Pandemic in Children | f1722103-0d1e-4e87-9601-8116c09d1334 | 8626874 | Pediatrics[mh] | Since the beginning of the COVID-19 pandemic, children and adolescents have shown low rates of infection, as well as relatively low rates of severe or critical forms of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). , NPIs have limited the spread of COVID-19 in Europe. , NPIs imposed to reduce the SARS-CoV-2 spread are also credited for the decrease of viral and bacterial infections observed, particularly in pediatric population ( ; available at www.jpeds.com ). Studies performed in the European Union have shown that lockdown measures have led to a rapid decline in the infectious respiratory diseases, specifically, a sharper decrease in influenza and RSV infections in 2020 and 2021, compared with previous years. However, the implementation of NPIs among the various European countries was different with a negative correlation between the severity of the NPI imposed and the decline of the infectious diseases. For instance, the decline of influenza and RSV was stronger in Italy, compared with other European countries such as Sweden and Germany, where the NPIs were less stringent and the population less compliant. The efficacy of the measures established to control of COVID-19 transmission is a consolidated evidence, and their effectiveness is also recognized as a key factor in containing the spread of other respiratory infectious diseases. However, the long-term impact of an extensive adoption of useful NPI safety measures is currently unknown. In particular, it is unclear whether and to what extent infectious diseases that significantly declined during the pandemic will re-emerge after COVID-19 mitigation measures are lifted. Data on increasing numbers of RSV cases after the summer 2021 seems to suggest an imminent risk for children's health.
Because of the NPIs enforced during the pandemic, children have experienced extended periods of low exposure to pathogens. , Immune memory is a defining feature of the acquired immune system. Concerns have been raised about the lack of exposure to common infections and the possibility that prolonged periods of reduced contacts with pathogens may reduce protective immunity , , and influence the development of adaptive immunity against infectious agents. , , Activation of the innate immune system can also result in enhanced responsiveness to triggers. In the absence of exposure to seasonal infectious diseases, immunity may decrease while the susceptibility to infections increases. Therefore, the immunity debt caused by the NPIs adopted during the pandemic could expose children to a greater potential for an increase in outbreaks and possible epidemics because of new or re-emerging infectious agents. ,
A significant decline in infection rates was reported for pneumococcal diseases, , Neisseria meningitidis infections, , pertussis, , varicella, , measles, enterovirus infections, influenza, , , , and RSV infections. , , , , In Australia, RSV infections began to increase during the spring months and peaked in late September 2021 with the number of cases significantly greater than in previous years. , In Japan, a significant outbreak of RSV infections was reported starting in spring 2021, while in the US, RSV started to rise in some areas of the southern states before summer 2021. , In Texas, where the trend began at the end of June, nearly one-half of pediatric patients hospitalized in local hospitals in August 2021 were diagnosed with RSV. , In Europe, respiratory infections in young children have begun to rise in England following low infection rates in response to COVID-19 restrictions and positivity of samples tested for RSV rapidly increased over 5 consecutive weeks by the end of July 2021, well before the typical winter season. In France, a delayed bronchiolitis epidemic driven by RSV was reported in March 2021. In autumn 2021, epidemiology indicators showed an earlier and more rapid onset of bronchiolitis virus circulation compared to previous years. , , Israel has experienced a large outbreak of RSV bronchiolitis during the summer of 2021 following the decline of the COVID-19 third wave and the lifting of NPI. This outbreak resulted in a significant increase of hospitalizations and overloaded pediatric wards and pediatric intensive care units throughout the country. In response to the RSV outbreak, the Israeli Ministry of Health recommended routine prophylaxis programs from the month of November. In Italy, the Italian Society of Pediatrics reports that RSV infection rose earlier than the usual peak observed in December-January. , In some areas of Southern Italy, in October 2021 there was a 30% increase of RSV cases compared to previous years and similar data are reported from the Northern regions, where in Veneto, in the month of October 2021 alone, the hospitalizations related to RSV were one-half of those recorded throughout the entire 2020-2021 winter season.
COVID-19 mitigation measures led to shifts in typical annual respiratory virus patterns. It is unclear whether this global trend will continue and how co-infection of RSV and other respiratory viruses with SARS-CoV-2 will affect disease severity in children during the winter season. Appropriate monitoring of the ongoing RSV epidemic and adequate interventional measures , should be continued and regular vaccination programs for infectious diseases maintained to protect children against respiratory viruses and other infectious agents and to prevent community transmission of diseases. ,
|
Application of artificial intelligence in laryngeal lesions: a systematic review and meta-analysis | bd32ad96-dd93-44b0-999b-1a6d614239ff | 11890366 | Surgical Procedures, Operative[mh] | Laryngeal lesions, which include benign and malignant conditions, represent a significant concern due to their impact on patients’ voice, swallowing and airway function, decreasing overall quality of life and health . Currently, different methodologies are used to help differentiate between benign and malignant laryngeal lesions, including indirect laryngoscopy with or without narrow-band imaging, direct laryngoscopy, ultrasound (US), and computed tomography (CT) . Many of these tools can be effective—such as laryngoscopy—but they rely heavily on a physician’s expertise, which can vary drastically depending on the physician’s experience and training . Studies show that less experienced clinicians may miss critical findings or misclassify lesions . This variability underscores the need for more standardized diagnostic tools. Other imaging modalities such as CT and MRI are essential for evaluating laryngeal lesions, particularly for assessing tumor extent and staging, but they have limitations, especially in detecting early-stage cancers . CT imaging also raises concerns about radiation exposure. Considering these challenges, biopsy remains the gold standard for diagnosing laryngeal lesions, despite its inherent risks such as infection and bleeding. However, sampling errors during biopsy can lead to inaccurate or inconclusive results, further complicating the diagnostic process . Artificial Intelligence (AI) technology offers the potential to overcome many of these limitations by providing automated and objective analysis of diagnostic data. Unlike human operators, AI systems can apply consistent algorithms to interpret findings, reducing the variability that arises from subjective human judgment. This consistency enhances diagnostic accuracy and could lead to more standardized care across different clinical settings . AI has revolutionized healthcare, impacting disease diagnosis, treatment, and patient outcomes . Recent advancements in AI, particularly in machine learning and deep learning, have shown promise in enhancing the diagnostic accuracy of precancerous lesions. AI has shown significant potential in enhancing diagnostic accuracy across various medical fields. In breast cancer detection, AI systems have achieved high accuracy, often surpassing human radiologists in early tumor detection, with some studies reporting an AUC of 0.92 . In lung cancer screening, AI applied to low-dose CT scans has improved the detection of pulmonary nodules and enhanced the prediction of major cardiopulmonary outcomes, offering a powerful tool for early intervention . Additionally, AI systems for diabetic retinopathy have demonstrated high sensitivity and specificity, making them effective in primary care settings, leading to the authorization of such a system for widespread use . These advancements indicate AI's growing role in improving diagnostic precision and accessibility in healthcare. Similar to its diagnostic utility with breast cancer and lung cancer, AI poses great potential if applied to laryngeal lesions . AI has shown great promise in analyzing complex medical images and patient data to assist in early detection, classification, and treatment of laryngeal lesions . For instance, different AI models have been successfully applied in analyzing endoscopic images to differentiate between benign and malignant laryngeal lesions with an accuracy of up to 93% . Other applications of AI in this field include convolutional neural networks (CNNs). CNNs are a class of deep neural networks that have been successfully applied in analyzing voice changes, a possible sign of early laryngeal cancer, with a high sensitivity and specificity . Using these methods AI can detect subtle imaging features that may be overlooked by experienced clinicians, leading to earlier and more accurate detection of malignancies. A study by Parker showed that machine learning via CNN-based methods to add objectivity to laryngoscopy analysis identifying small changes with high specificity . The potential for AI applications to aid in the detection of laryngeal lesions is remarkable; however, the currently published literature presents a fragmented view, reporting only one AI model application per study. Previous studies focused on various individual AI application diagnostic tools, mainly on endoscopy . AI has demonstrated remarkable accuracy in detecting and classifying laryngeal lesions, making it a valuable tool for early diagnosis. For example, a study by Zhou et al. highlighted that AI models applied to endoscopic images achieved a classification accuracy of over 90% in distinguishing benign from malignant laryngeal lesions, significantly aiding clinical decision-making . Additionally, research by Ren et al. demonstrated that an AI-driven diagnostic system could accurately classify laryngeal lesions with an accuracy of 94%, showcasing its effectiveness in supporting physicians in making precise diagnoses . Furthermore, a review by Wu et al. from 2020 to 2022 reported the recent increase in AI technology applications in otolaryngology . A state-of-the-art review by Bensoussan et al. provides a comprehensive summary of current uses of AI, but it is limited by a lack of meta-analysis of their findings . Since most prior investigations focus on endoscopy, our systematic review and meta-analysis included studies that detect or differentiate benign versus malignant laryngeal lesions with endoscopy, histopathology, or voice changes. This is the first systematic review that meta-analyzes several tools using AI applications in this area. The aim is to expand on the results of previous reviews and provide a broader understanding of AI's potential in this area. In recent years, AI research has experienced exponential growth. This study addresses the need for a comprehensive meta-analysis of various AI techniques by providing an up-to-date systematic review. It serves as a valuable resource that consolidates findings across diverse AI technologies, offering clinicians timely and impactful insights while efficiently summarizing the most relevant data in a single source.
Information source and search strategy This study was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (“PRISMA”) guidelines . Detailed search strategies were developed using the following databases The COCHRANE Library (Wiley), PubMed (National Library of Medicine – National Institutes of Health), SCOPUS (Elsevier), and CINAHL (EBSCO). Databases were searched from inception through January 1, 2024, with the results limited to English. A combination of subject headings (e.g., Medical Subject Headings [MeSH] in PubMed) and the following keywords were used in the search: “Artificial Intelligence” or “Machine Learning,” “Laryngeal Neoplasm,” “laryngeal lesion” and “glottic neoplasm” The complete list of search terms is available in Table . Articles were screened using the review management software Covidence (Veritas Health Innovation Ltd). Titles and abstracts were screened for relevance, and full texts were reviewed to determine inclusion. References of all included articles were examined for additional studies. Selection criteria The study considered several types of research designs to ensure a robust and comprehensive analysis of AI applications in diagnosing and classifying laryngeal lesions. Included study designs were cohort studies, retrospective and prospective case series, randomized controlled trials (RCTs), and case–control studies, as these designs provide valuable longitudinal data, insights into real-world clinical applications, high levels of evidence. Studies were selected if they reported key performance metrics such as accuracy, sensitivity, and specificity, which are crucial for evaluating the diagnostic utility of AI technologies. To maintain the rigor of the meta-analysis, review articles, studies with incomplete data, case reports, studies with incorrect study designs, and studies that did not measure the outcomes of interest were excluded. This exclusion was necessary to ensure that only studies contributing fully to the quantitative synthesis were included, thereby maintaining the focus on clinically relevant outcomes and the best available evidence in the application of AI for laryngeal lesion diagnosis and classification (Table ). The study included a variety of AI approaches, specifically focusing on deep learning models such as CNNs, Recurrent Neural Networks (RNNs), and Autoencoders. Additionally, conventional machine learning models like Support Vector Machines (SVMs), Random Forests, and Naive Bayes were considered. Hybrid models, including Ensemble Learning and Transfer Learning models, were also included to capture a broad spectrum of AI applications in diagnosing and classifying laryngeal lesions. Two reviewers (A.R.M.G. and T.J.D.) independently conducted an initial screening of titles and abstracts to identify studies that met the broad eligibility criteria. This step was crucial for filtering out studies that were clearly irrelevant or did not align with the research focus. Following the initial screening, the reviewers conducted a more detailed full-text review of the remaining articles to determine their suitability for inclusion in the final analysis. This review process was guided by our PICOT framework, which focused on the following elements: Population (studies evaluating patients with laryngeal lesions using AI for diagnosis and classification), Intervention (AI technologies applied to endoscopy, voice analysis, and histopathology), Comparator (comparisons made against traditional diagnostic methods if applicable), Outcomes (sensitivity, specificity, and accuracy of AI in diagnosing and classifying laryngeal lesions), and Timing (inclusion of studies published up to January 1, 2024). To ensure best methods of the screening process, any disagreements between the two reviewers were addressed through discussion. If consensus could not be reached, a third reviewer (S.A.N.) was consulted to provide an objective assessment and resolve the conflict. Data collection process and data items The two reviewers independently extracted data from each included study to ensure accuracy and minimize the risk of bias. This process involved systematically gathering information on various study characteristics, including the author names, publication year, and participant demographics, such as age, gender, and clinical context. Additionally, the specific AI model utilized in each study (e.g., CNNs, RNNs, or SVMs) was documented, along with key performance metrics like accuracy, sensitivity, and specificity. After the initial data extraction, the reviewers compared their results to identify any discrepancies. In cases where disagreements arose, a third reviewer (S.A.N.) was consulted to provide an objective resolution. Critical appraisal The included articles were critically appraised to assess the level of evidence using the Oxford Center for Evidence-Based Medicine criteria . Since the included studies were a mix of randomized and non-randomized studies, QUADAS-2 was used. QUADAS-2 is a risk-of-bias tool that assesses four domains. These four domains are patient selection, index test, reference standard, and flow and timing. The patient selection domain examines how participants were chosen, ensuring they represent the typical clinical population. The index test domain evaluates whether the AI technology was applied and interpreted independently of the reference standard, maintaining objectivity. The reference standard domain assesses the validity and application of the "gold standard" diagnostic method without bias from the index test results. Finally, the flow and timing domain looks at the sequence and timing of tests to ensure consistency and reduce potential bias from variations in test administration. Each domain is assessed for risk of bias, and the first three domains are also evaluated for concerns about applicability . Two authors (A.R.M.G and T.J.D.) independently performed risk assessments on all studies. The risk of bias for each aspect was graded as low, unclear, or high. The third reviewer (T.C) resolved any conflicts between reviewers. Data analysis and synthesis of results (statistical analysis) Meta‐analysis of continuous measures (sensitivity, specificity, and accuracy) was performed with Cochrane Review Manager (RevMan) version 5.4 (The Cochrane Collaboration 2020, United Kingdom). A meta-analysis of proportions (gender) was performed using MedCalc 22.017 (MedCalc Software, Ostend, Belgium). Each measure (mean [log(mean0] / proportion and 95% confidence interval (CI)) was weighted according to the number of patients affected. Heterogeneity among studies was assessed using χ 2 and I 2 statistics with fixed effects (I 2 < 50%) and random effects (I 2 > 50%). We conducted a meta-regression where we examined factors influencing diagnostic accuracy, including the number of images, and AI model type using R (software 4.4.1). In addition, potential publication bias was evaluated by visual inspection of the funnel plot and Egger’s regression test, which statistically examines the asymmetry of the funnel plot . A p value of < 0.05 was considered to indicate a significant difference for all statistical tests.
This study was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (“PRISMA”) guidelines . Detailed search strategies were developed using the following databases The COCHRANE Library (Wiley), PubMed (National Library of Medicine – National Institutes of Health), SCOPUS (Elsevier), and CINAHL (EBSCO). Databases were searched from inception through January 1, 2024, with the results limited to English. A combination of subject headings (e.g., Medical Subject Headings [MeSH] in PubMed) and the following keywords were used in the search: “Artificial Intelligence” or “Machine Learning,” “Laryngeal Neoplasm,” “laryngeal lesion” and “glottic neoplasm” The complete list of search terms is available in Table . Articles were screened using the review management software Covidence (Veritas Health Innovation Ltd). Titles and abstracts were screened for relevance, and full texts were reviewed to determine inclusion. References of all included articles were examined for additional studies.
The study considered several types of research designs to ensure a robust and comprehensive analysis of AI applications in diagnosing and classifying laryngeal lesions. Included study designs were cohort studies, retrospective and prospective case series, randomized controlled trials (RCTs), and case–control studies, as these designs provide valuable longitudinal data, insights into real-world clinical applications, high levels of evidence. Studies were selected if they reported key performance metrics such as accuracy, sensitivity, and specificity, which are crucial for evaluating the diagnostic utility of AI technologies. To maintain the rigor of the meta-analysis, review articles, studies with incomplete data, case reports, studies with incorrect study designs, and studies that did not measure the outcomes of interest were excluded. This exclusion was necessary to ensure that only studies contributing fully to the quantitative synthesis were included, thereby maintaining the focus on clinically relevant outcomes and the best available evidence in the application of AI for laryngeal lesion diagnosis and classification (Table ). The study included a variety of AI approaches, specifically focusing on deep learning models such as CNNs, Recurrent Neural Networks (RNNs), and Autoencoders. Additionally, conventional machine learning models like Support Vector Machines (SVMs), Random Forests, and Naive Bayes were considered. Hybrid models, including Ensemble Learning and Transfer Learning models, were also included to capture a broad spectrum of AI applications in diagnosing and classifying laryngeal lesions. Two reviewers (A.R.M.G. and T.J.D.) independently conducted an initial screening of titles and abstracts to identify studies that met the broad eligibility criteria. This step was crucial for filtering out studies that were clearly irrelevant or did not align with the research focus. Following the initial screening, the reviewers conducted a more detailed full-text review of the remaining articles to determine their suitability for inclusion in the final analysis. This review process was guided by our PICOT framework, which focused on the following elements: Population (studies evaluating patients with laryngeal lesions using AI for diagnosis and classification), Intervention (AI technologies applied to endoscopy, voice analysis, and histopathology), Comparator (comparisons made against traditional diagnostic methods if applicable), Outcomes (sensitivity, specificity, and accuracy of AI in diagnosing and classifying laryngeal lesions), and Timing (inclusion of studies published up to January 1, 2024). To ensure best methods of the screening process, any disagreements between the two reviewers were addressed through discussion. If consensus could not be reached, a third reviewer (S.A.N.) was consulted to provide an objective assessment and resolve the conflict.
The two reviewers independently extracted data from each included study to ensure accuracy and minimize the risk of bias. This process involved systematically gathering information on various study characteristics, including the author names, publication year, and participant demographics, such as age, gender, and clinical context. Additionally, the specific AI model utilized in each study (e.g., CNNs, RNNs, or SVMs) was documented, along with key performance metrics like accuracy, sensitivity, and specificity. After the initial data extraction, the reviewers compared their results to identify any discrepancies. In cases where disagreements arose, a third reviewer (S.A.N.) was consulted to provide an objective resolution.
The included articles were critically appraised to assess the level of evidence using the Oxford Center for Evidence-Based Medicine criteria . Since the included studies were a mix of randomized and non-randomized studies, QUADAS-2 was used. QUADAS-2 is a risk-of-bias tool that assesses four domains. These four domains are patient selection, index test, reference standard, and flow and timing. The patient selection domain examines how participants were chosen, ensuring they represent the typical clinical population. The index test domain evaluates whether the AI technology was applied and interpreted independently of the reference standard, maintaining objectivity. The reference standard domain assesses the validity and application of the "gold standard" diagnostic method without bias from the index test results. Finally, the flow and timing domain looks at the sequence and timing of tests to ensure consistency and reduce potential bias from variations in test administration. Each domain is assessed for risk of bias, and the first three domains are also evaluated for concerns about applicability . Two authors (A.R.M.G and T.J.D.) independently performed risk assessments on all studies. The risk of bias for each aspect was graded as low, unclear, or high. The third reviewer (T.C) resolved any conflicts between reviewers.
Meta‐analysis of continuous measures (sensitivity, specificity, and accuracy) was performed with Cochrane Review Manager (RevMan) version 5.4 (The Cochrane Collaboration 2020, United Kingdom). A meta-analysis of proportions (gender) was performed using MedCalc 22.017 (MedCalc Software, Ostend, Belgium). Each measure (mean [log(mean0] / proportion and 95% confidence interval (CI)) was weighted according to the number of patients affected. Heterogeneity among studies was assessed using χ 2 and I 2 statistics with fixed effects (I 2 < 50%) and random effects (I 2 > 50%). We conducted a meta-regression where we examined factors influencing diagnostic accuracy, including the number of images, and AI model type using R (software 4.4.1). In addition, potential publication bias was evaluated by visual inspection of the funnel plot and Egger’s regression test, which statistically examines the asymmetry of the funnel plot . A p value of < 0.05 was considered to indicate a significant difference for all statistical tests.
Overview of search strategy A comprehensive literature search found 1713 unique articles related to AI and laryngeal pathology. Title and abstract review led to the exclusion of 1558 articles, resulting in 155 studies assessed in full text. After careful consideration, 18 articles were included in the systematic review and the meta-analysis. The search and screening process is in Fig. , which shows the PRISMA diagram [30]. The studies included non-randomized cohort studies, which were classified as level 3 based on the Oxford level of evidence and were published from the database's inception to January 2024. Risk of bias Critical appraisal of studies indicated an acceptably low risk of bias for all included studies. Potential sources of bias in non-randomized studies with QUADAS-2 (Fig. ) were most pronounced due to the index test, selection of participants, and reference standard. A funnel plot with Egger’s test (Fig. S2) (– 2.7, 95% CI – 9.6 to 4.1, p = 0.33) indicated that all studies were within the funnel, suggesting little publication bias. Overview of included studies The descriptive features of the included studies are summarized in Table . Eighteen studies comprising 18,944 patients were included, with the proportion of males being 92% (95% CI 85.6–96.6%). Of the 18 studies included in the meta-analysis, 12 (66.7%) reported AI-aided laryngeal endoscopy , 2 (11.1%) reported detection of lesions using voice changes , and 4 (22.2%) reported classification of benign versus malignant using histopathology . The number of samples/images used in the AI models ranges from 124 to 24,667, totaling 115,136. Laryngeal lesion detection Nine studies were identified that used AI application in laryngeal endoscopy to detect lesions and included sensitivity, specificity, and accuracy. Figure shows a pooled analysis of the outcomes of AI application in laryngeal endoscopy for detecting lesions versus healthy tissue. AI application in laryngeal endoscopy showed a pooled mean sensitivity and specificity were 89% (95% CI 85–93%) and 91% (95% CI 90–93%). The pooled accuracy between all studies was 92% (95% CI 89–96%). The mean number of images the AI-aided endoscopy used in this modality was 8765.66. This meta-analysis included only one other AI-assisted diagnostic tool, which was voice-analysis software, for detecting possible laryngeal lesions. Other investigations with diagnostic tools such as CT were identified in the search but were excluded due to their high risk of bias and lack of complete outcomes for meta-analysis. Voice analysis’ pooled mean sensitivity and specificity were 78% (95% CI 76–79%) and 82% (95% CI 72–94%). The pooled accuracy between all studies was 86% (95% CI 85–87%) (Fig. ). Classification of benign versus malignant lesions Seven studies were identified that used AI application in laryngeal endoscopy to differentiate benign versus malignant lesions, including sensitivity, specificity, and accuracy. Figure shows a pooled analysis of sensitivity, specificity, and accuracy for this diagnostic tool. The pooled mean sensitivity and specificity for laryngeal endoscopy distinguishing between lesions were 91% (95% CI 87–94%) and 91% (95% CI 88–95%), respectively. The pooled accuracy between all studies was 94% (95% CI 92–97%). The mean number of images the AI model used was 8133.33. AI-aided histopathology classification was the only diagnostic tool that used AI to differentiate benign versus malignant laryngeal lesions. Four studies reported using AI to aid in the histopathology classification, but the only outcome that could be analyzed in these studies was accuracy due to the lack of standard deviation or p-value in other outcomes, such as sensitivity and specificity. Figure , the pooled accuracy between these studies was 92% (95% CI 86–99%). The mean number of images the AI model used was 6,416, ranging from 414 to 18,750, with a total of 25,664. Meta-regression analysis The meta-regression analysis demonstrated a significant relationship between diagnostic accuracy and the AI technique used, with an R-squared value of 91.5%, indicating that 91.5% of the variability in diagnostic accuracy is explained by the type of AI modality (endoscopy, voice analysis, or histopathology). After adjusting for the number of predictors, the adjusted R-squared remained high at 87.2%, further supporting the strength of this relationship. The model was statistically significant (F-statistic = 21.49, p = 0.0435), suggesting that the choice of AI technique significantly influences diagnostic accuracy. Specifically, histopathology-based AI techniques were associated with a 24.46% higher diagnostic accuracy compared to endoscopy, as indicated by a coefficient of 0.2446 (95% CI 0.018 to 0.472, p = 0.044). However, due to limited data, it was not possible to determine the effect of voice analysis on diagnostic accuracy in this model.
A comprehensive literature search found 1713 unique articles related to AI and laryngeal pathology. Title and abstract review led to the exclusion of 1558 articles, resulting in 155 studies assessed in full text. After careful consideration, 18 articles were included in the systematic review and the meta-analysis. The search and screening process is in Fig. , which shows the PRISMA diagram [30]. The studies included non-randomized cohort studies, which were classified as level 3 based on the Oxford level of evidence and were published from the database's inception to January 2024.
Critical appraisal of studies indicated an acceptably low risk of bias for all included studies. Potential sources of bias in non-randomized studies with QUADAS-2 (Fig. ) were most pronounced due to the index test, selection of participants, and reference standard. A funnel plot with Egger’s test (Fig. S2) (– 2.7, 95% CI – 9.6 to 4.1, p = 0.33) indicated that all studies were within the funnel, suggesting little publication bias.
The descriptive features of the included studies are summarized in Table . Eighteen studies comprising 18,944 patients were included, with the proportion of males being 92% (95% CI 85.6–96.6%). Of the 18 studies included in the meta-analysis, 12 (66.7%) reported AI-aided laryngeal endoscopy , 2 (11.1%) reported detection of lesions using voice changes , and 4 (22.2%) reported classification of benign versus malignant using histopathology . The number of samples/images used in the AI models ranges from 124 to 24,667, totaling 115,136.
Nine studies were identified that used AI application in laryngeal endoscopy to detect lesions and included sensitivity, specificity, and accuracy. Figure shows a pooled analysis of the outcomes of AI application in laryngeal endoscopy for detecting lesions versus healthy tissue. AI application in laryngeal endoscopy showed a pooled mean sensitivity and specificity were 89% (95% CI 85–93%) and 91% (95% CI 90–93%). The pooled accuracy between all studies was 92% (95% CI 89–96%). The mean number of images the AI-aided endoscopy used in this modality was 8765.66. This meta-analysis included only one other AI-assisted diagnostic tool, which was voice-analysis software, for detecting possible laryngeal lesions. Other investigations with diagnostic tools such as CT were identified in the search but were excluded due to their high risk of bias and lack of complete outcomes for meta-analysis. Voice analysis’ pooled mean sensitivity and specificity were 78% (95% CI 76–79%) and 82% (95% CI 72–94%). The pooled accuracy between all studies was 86% (95% CI 85–87%) (Fig. ).
Seven studies were identified that used AI application in laryngeal endoscopy to differentiate benign versus malignant lesions, including sensitivity, specificity, and accuracy. Figure shows a pooled analysis of sensitivity, specificity, and accuracy for this diagnostic tool. The pooled mean sensitivity and specificity for laryngeal endoscopy distinguishing between lesions were 91% (95% CI 87–94%) and 91% (95% CI 88–95%), respectively. The pooled accuracy between all studies was 94% (95% CI 92–97%). The mean number of images the AI model used was 8133.33. AI-aided histopathology classification was the only diagnostic tool that used AI to differentiate benign versus malignant laryngeal lesions. Four studies reported using AI to aid in the histopathology classification, but the only outcome that could be analyzed in these studies was accuracy due to the lack of standard deviation or p-value in other outcomes, such as sensitivity and specificity. Figure , the pooled accuracy between these studies was 92% (95% CI 86–99%). The mean number of images the AI model used was 6,416, ranging from 414 to 18,750, with a total of 25,664.
The meta-regression analysis demonstrated a significant relationship between diagnostic accuracy and the AI technique used, with an R-squared value of 91.5%, indicating that 91.5% of the variability in diagnostic accuracy is explained by the type of AI modality (endoscopy, voice analysis, or histopathology). After adjusting for the number of predictors, the adjusted R-squared remained high at 87.2%, further supporting the strength of this relationship. The model was statistically significant (F-statistic = 21.49, p = 0.0435), suggesting that the choice of AI technique significantly influences diagnostic accuracy. Specifically, histopathology-based AI techniques were associated with a 24.46% higher diagnostic accuracy compared to endoscopy, as indicated by a coefficient of 0.2446 (95% CI 0.018 to 0.472, p = 0.044). However, due to limited data, it was not possible to determine the effect of voice analysis on diagnostic accuracy in this model.
Main findings Laryngeal cancer is one of the most common subtypes of head and neck cancer, with an estimated 12,380 new cases per year . According to the American Cancer Society, the 5-year relative survival rate for glottic cancer for all stages is 77%, but when stratified, it ranges from 84% for localized disease to 45% for distant disease . This decline in survival with distant disease highlights the importance of early detection and accurate diagnosis of laryngeal lesions. There has been an exponential increase in published research on AI applications in healthcare in the past decade . Most studies have shown high efficacy in the detection of lesions in other body sites, such as the colon and breast. This systematic review and meta-analysis supports that AI could be useful in laryngology, improving early detection and the accuracy of diagnosis of laryngeal lesions. Of the included studies, all AI models had very high sensitivity, specificity, and accuracy. The accuracy of the included studies ranges from 86 and 94%. AI application in laryngeal endoscopy had the highest accuracy with a pooled mean of 92% (95% CI 89–96%). This is supported by previous studies that reported the accuracy of AI used in laryngeal endoscopy was between 80 and 99% . These results show that AI applications have great efficacy and the potential to be introduced as diagnostic tools to aid physicians. Given the accessibility of laryngoscopy as an office procedure, implementing AI into this diagnostic tool may help improve diagnosis and early detection of malignant lesions, which could potentially lead to improved outcomes . Other techniques also show promising results, but a lack of standardized outcomes makes it challenging to compare diagnostic tools. The accuracy of reference standard tests, such as histopathology and expert-reviewed imaging, is generally high, but it is not without limitations. Variability in interpretation, especially in imaging, can introduce inconsistencies, which AI technologies aim to mitigate by offering more objective and standardized assessments. The finding of our meta regression suggests that while histopathology-based AI models are highly effective for diagnosing laryngeal lesions, further research is needed to better evaluate the impact of voice analysis and other contextual factors on diagnostic accuracy. A study by Kim et al. explored AI-assisted voice analysis programs to identify healthy tissue versus laryngeal cancer accurately and showed promising results, with 15% more accurate detection when compared with laryngologists . Our review of voice-analysis studies supports these findings, with meta-analysis showing a mean pooled accuracy of 86% Even though identifying laryngeal lesions is important, differentiating between benign and malignant lesions is crucial for patient outcomes. AI-aided technology has also shown great potential in this area. A study by Hu et al. reported that AI-aided ultrasound correctly classified benign versus malignant nodules in the liver with an accuracy of 91% . This AI model even outperformed the diagnostic accuracy of radiology residents and matched that of experts . Other studies support these findings in malignant thyroid, breast, and lung lesions . Our study also supports that AI application in diagnostic tools such as endoscopy is useful in differentiating benign versus malignant lesions with an accuracy of up to 94%. These results further support the utility of integrating AI into these commonly used diagnostic tools such as laryngeal endoscopy. A study by Ren et al., which used the largest dataset of images of all included studies, reported that compared to expert physicians, AI has better overall accuracy in the classification of leukoplakia (65% vs 91%) and glottic cancer (54% vs 90%) . This shows potential for integrating AI into clinical practice, especially for the future generation of physicians. In summary, our finding reported the diagnostic accuracy of AI technologies in laryngeal lesion detection varies across modalities. AI-assisted endoscopy demonstrated the highest diagnostic performance, with a pooled sensitivity of 91% (95% CI 87–94%) for differentiating benign from malignant lesions and a pooled accuracy of 94% (95% CI 92–97%) for lesion detection. Histopathology-based AI models, although less frequently studied, also achieved a high pooled accuracy of 92% (95% CI 86–99%). In contrast, voice analysis, while promising, showed a relatively lower pooled sensitivity of 78% (95% CI 76–79%) and an accuracy of 86% (95% CI 85–87%). These results indicate that while endoscopy and histopathology show strong diagnostic potential for lesion classification, voice analysis may require further refinement and validation to match the efficacy of imaging-based AI technologies. Applicability of AI in the clinical setting The results of this meta-analysis suggest that AI is a diagnostic tool that could be valuable for assessing laryngeal lesions. A previous systematic review reported that AI has high accuracy and clinical utility when assessing images of laryngeal lesions . This raises the question of who would benefit the most from AI in the clinical setting. The most significant utility of AI currently is in the classification of benign versus malignant lesions . A review by Sampieri et al. found that AI models perform better in binary classification of benign vs malignant but lose accuracy in multiclass classification . Of the included studies, Dunham et al. reported this finding, showing 93% accuracy when the model classified between benign and malignant but dropping to 83% when classifying between several different lesions . This is similar to humans; the more complicated the task, the lower the accuracy. However, it should be noted that the research studies presented so far have focused solely on the research-oriented setting . As real-world clinical applications of this technology have yet to be studied, it is difficult to assess the actual implications of this technology in a clinical setting. AI technologies could be particularly beneficial for patients in rural or underserved areas, where access to specialist care is limited. Moreover, busy clinical environments could also benefit from the enhanced efficiency and diagnostic consistency that AI tools offer. Several barriers exist before this technology is helpful in clinical practice, such as the AI software used is not commercially available. Significant obstacles to clinical implementation also include the current lack of regulatory clearance and commercial availability of AI tools. Additionally, the integration of these technologies will necessitate updated equipment and comprehensive training programs to ensure healthcare professionals can effectively utilize and interpret AI-generated results. In addition, AI implementation in healthcare must address significant privacy concerns, as these models depend on large datasets containing sensitive patient information. Additionally, potential biases in the data used to train AI models can result in inequitable outcomes, particularly for underrepresented patient populations. A study by Alowais et al. reported that integrating AI into healthcare improves disease diagnosis, but it does come with challenges. Incorporating AI into clinical practice could bring problems with data privacy and bias . More studies that show standardized outcomes are needed to assess the potential for AI-aided diagnostic and classification tools in laryngology. In addition, prospective studies that implement AI-aided laryngeal endoscopy systems in clinical practice are essential. Limitations It is important to consider several limitations when interpreting the results of this study. First, despite a comprehensive literature search, publication bias is possible. This could lead to overestimating AI’s effectiveness in detecting and classifying laryngeal lesions. Second, using different AI models and diagnostic tools across the included studies may limit the generalizability of our findings. Different AI algorithms, training datasets, and validation methods can significantly impact the results, making direct comparisons difficult. Finally, standardized reporting on the technical details of AI models, such as architecture, training parameters, and data augmentation techniques, needs to be improved. This limits the reproducibility of the studies and hinders the assessment of the utility of AI applications in laryngology. The study has a notable limitation due to the high heterogeneity among the included studies. Study heterogeneity poses a significant challenge in our meta-analysis, arising from differences in study design, patient populations, and AI model types. This variability limits the generalizability of our findings and underscores the importance of standardizing research protocols in future studies. Apart from heterogeneity, this study faced limitations due to the prevalence of retrospective designs and small sample sizes, which constrain the robustness of our conclusions. The potential for publication bias, especially the underreporting of negative outcomes, may further skew the perceived effectiveness of AI technologies. Additionally, the lack of sufficient data on various AI algorithms and architectures limited our ability to fully assess and compare their effectiveness. It may affect the applicability of the results and could affect the actual effectiveness of AI tools in diagnosing and classifying laryngeal lesions in various clinical settings. In summary, while our findings suggest that AI applications have the potential to diagnose and classify laryngeal lesions, these limitations highlight the need for further research. Although the AI models reviewed in this study exhibit high accuracy, further validation is needed before they can be widely adopted in clinical practice. Additional studies are required to assess the performance of these models in real-world settings to ensure their robustness and reliability. Future research should include large, multicenter studies to validate these findings across diverse patient populations. Comparative effectiveness studies are also crucial to directly compare AI-assisted diagnostics with traditional methods, thereby determining their true clinical value.
Laryngeal cancer is one of the most common subtypes of head and neck cancer, with an estimated 12,380 new cases per year . According to the American Cancer Society, the 5-year relative survival rate for glottic cancer for all stages is 77%, but when stratified, it ranges from 84% for localized disease to 45% for distant disease . This decline in survival with distant disease highlights the importance of early detection and accurate diagnosis of laryngeal lesions. There has been an exponential increase in published research on AI applications in healthcare in the past decade . Most studies have shown high efficacy in the detection of lesions in other body sites, such as the colon and breast. This systematic review and meta-analysis supports that AI could be useful in laryngology, improving early detection and the accuracy of diagnosis of laryngeal lesions. Of the included studies, all AI models had very high sensitivity, specificity, and accuracy. The accuracy of the included studies ranges from 86 and 94%. AI application in laryngeal endoscopy had the highest accuracy with a pooled mean of 92% (95% CI 89–96%). This is supported by previous studies that reported the accuracy of AI used in laryngeal endoscopy was between 80 and 99% . These results show that AI applications have great efficacy and the potential to be introduced as diagnostic tools to aid physicians. Given the accessibility of laryngoscopy as an office procedure, implementing AI into this diagnostic tool may help improve diagnosis and early detection of malignant lesions, which could potentially lead to improved outcomes . Other techniques also show promising results, but a lack of standardized outcomes makes it challenging to compare diagnostic tools. The accuracy of reference standard tests, such as histopathology and expert-reviewed imaging, is generally high, but it is not without limitations. Variability in interpretation, especially in imaging, can introduce inconsistencies, which AI technologies aim to mitigate by offering more objective and standardized assessments. The finding of our meta regression suggests that while histopathology-based AI models are highly effective for diagnosing laryngeal lesions, further research is needed to better evaluate the impact of voice analysis and other contextual factors on diagnostic accuracy. A study by Kim et al. explored AI-assisted voice analysis programs to identify healthy tissue versus laryngeal cancer accurately and showed promising results, with 15% more accurate detection when compared with laryngologists . Our review of voice-analysis studies supports these findings, with meta-analysis showing a mean pooled accuracy of 86% Even though identifying laryngeal lesions is important, differentiating between benign and malignant lesions is crucial for patient outcomes. AI-aided technology has also shown great potential in this area. A study by Hu et al. reported that AI-aided ultrasound correctly classified benign versus malignant nodules in the liver with an accuracy of 91% . This AI model even outperformed the diagnostic accuracy of radiology residents and matched that of experts . Other studies support these findings in malignant thyroid, breast, and lung lesions . Our study also supports that AI application in diagnostic tools such as endoscopy is useful in differentiating benign versus malignant lesions with an accuracy of up to 94%. These results further support the utility of integrating AI into these commonly used diagnostic tools such as laryngeal endoscopy. A study by Ren et al., which used the largest dataset of images of all included studies, reported that compared to expert physicians, AI has better overall accuracy in the classification of leukoplakia (65% vs 91%) and glottic cancer (54% vs 90%) . This shows potential for integrating AI into clinical practice, especially for the future generation of physicians. In summary, our finding reported the diagnostic accuracy of AI technologies in laryngeal lesion detection varies across modalities. AI-assisted endoscopy demonstrated the highest diagnostic performance, with a pooled sensitivity of 91% (95% CI 87–94%) for differentiating benign from malignant lesions and a pooled accuracy of 94% (95% CI 92–97%) for lesion detection. Histopathology-based AI models, although less frequently studied, also achieved a high pooled accuracy of 92% (95% CI 86–99%). In contrast, voice analysis, while promising, showed a relatively lower pooled sensitivity of 78% (95% CI 76–79%) and an accuracy of 86% (95% CI 85–87%). These results indicate that while endoscopy and histopathology show strong diagnostic potential for lesion classification, voice analysis may require further refinement and validation to match the efficacy of imaging-based AI technologies.
The results of this meta-analysis suggest that AI is a diagnostic tool that could be valuable for assessing laryngeal lesions. A previous systematic review reported that AI has high accuracy and clinical utility when assessing images of laryngeal lesions . This raises the question of who would benefit the most from AI in the clinical setting. The most significant utility of AI currently is in the classification of benign versus malignant lesions . A review by Sampieri et al. found that AI models perform better in binary classification of benign vs malignant but lose accuracy in multiclass classification . Of the included studies, Dunham et al. reported this finding, showing 93% accuracy when the model classified between benign and malignant but dropping to 83% when classifying between several different lesions . This is similar to humans; the more complicated the task, the lower the accuracy. However, it should be noted that the research studies presented so far have focused solely on the research-oriented setting . As real-world clinical applications of this technology have yet to be studied, it is difficult to assess the actual implications of this technology in a clinical setting. AI technologies could be particularly beneficial for patients in rural or underserved areas, where access to specialist care is limited. Moreover, busy clinical environments could also benefit from the enhanced efficiency and diagnostic consistency that AI tools offer. Several barriers exist before this technology is helpful in clinical practice, such as the AI software used is not commercially available. Significant obstacles to clinical implementation also include the current lack of regulatory clearance and commercial availability of AI tools. Additionally, the integration of these technologies will necessitate updated equipment and comprehensive training programs to ensure healthcare professionals can effectively utilize and interpret AI-generated results. In addition, AI implementation in healthcare must address significant privacy concerns, as these models depend on large datasets containing sensitive patient information. Additionally, potential biases in the data used to train AI models can result in inequitable outcomes, particularly for underrepresented patient populations. A study by Alowais et al. reported that integrating AI into healthcare improves disease diagnosis, but it does come with challenges. Incorporating AI into clinical practice could bring problems with data privacy and bias . More studies that show standardized outcomes are needed to assess the potential for AI-aided diagnostic and classification tools in laryngology. In addition, prospective studies that implement AI-aided laryngeal endoscopy systems in clinical practice are essential.
It is important to consider several limitations when interpreting the results of this study. First, despite a comprehensive literature search, publication bias is possible. This could lead to overestimating AI’s effectiveness in detecting and classifying laryngeal lesions. Second, using different AI models and diagnostic tools across the included studies may limit the generalizability of our findings. Different AI algorithms, training datasets, and validation methods can significantly impact the results, making direct comparisons difficult. Finally, standardized reporting on the technical details of AI models, such as architecture, training parameters, and data augmentation techniques, needs to be improved. This limits the reproducibility of the studies and hinders the assessment of the utility of AI applications in laryngology. The study has a notable limitation due to the high heterogeneity among the included studies. Study heterogeneity poses a significant challenge in our meta-analysis, arising from differences in study design, patient populations, and AI model types. This variability limits the generalizability of our findings and underscores the importance of standardizing research protocols in future studies. Apart from heterogeneity, this study faced limitations due to the prevalence of retrospective designs and small sample sizes, which constrain the robustness of our conclusions. The potential for publication bias, especially the underreporting of negative outcomes, may further skew the perceived effectiveness of AI technologies. Additionally, the lack of sufficient data on various AI algorithms and architectures limited our ability to fully assess and compare their effectiveness. It may affect the applicability of the results and could affect the actual effectiveness of AI tools in diagnosing and classifying laryngeal lesions in various clinical settings. In summary, while our findings suggest that AI applications have the potential to diagnose and classify laryngeal lesions, these limitations highlight the need for further research. Although the AI models reviewed in this study exhibit high accuracy, further validation is needed before they can be widely adopted in clinical practice. Additional studies are required to assess the performance of these models in real-world settings to ensure their robustness and reliability. Future research should include large, multicenter studies to validate these findings across diverse patient populations. Comparative effectiveness studies are also crucial to directly compare AI-assisted diagnostics with traditional methods, thereby determining their true clinical value.
The study reports high accuracy, sensitivity, and specificity of AI-aided tools, especially in laryngeal endoscopy, for classifying benign and malignant lesions. These findings indicate that AI could enhance early diagnosis and classification, potentially improving patient outcomes. However, several limitations highlight the need for further research to validate and refine AI applications in laryngology.
Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 1011 KB)
|
Early phase clinical trials in oncology: realising the potential of seamless designs | 0fa46d44-35c3-4f47-b750-7a75513644e1 | 7614750 | Internal Medicine[mh] | Introduction It is well-recognized drug development is risky with a very high rate of attrition . The development of oncology drugs is particularly challenging with low rates of approval for novel treatments when compared with other therapeutic areas . To accelerate the development and approval of oncology drugs with high clinical effectiveness, cost-effectiveness, and low attrition rate in the future, the US Food and Drug Administration (FDA) promotes innovation and modern technology to overcome corresponding challenges and only recently started to Project Optimus in order to reform dose optimization and dose selection. One important aspect of innovation in drug development are adaptive designs, in which seamless phase I/II designs are a particularly critical application in early phases of modern drug development . These seamless designs seek to establish the safe range of doses while also learning about the potential activity of a novel agent within a single study that forms the basis of identifying the best dose to use in later development. As a critical component of the expensive, time-consuming drug development pipeline, clinical trials are traditionally divided into sequential, separate phases, where drugs are evaluated for safety in phase I, early signals of efficacy in phase II, and then investigated against standard of care in large randomised phase III clinical trials. Decreasing the high attrition rates in drug development continues to be a primary challenge for the pharmaceutical industry. One of the main challenges in achieving this goal is to strike an appropriate balance between drug safety and efficacy particularly in the early-phase studies (i.e., phase I and phase II trials, ; ; ). Phase I clinical trials in oncology, wherein a new drug or drug combination is applied for the first time to humans, are the foundation of a successful clinical drug development programme. These studies are crucial because they define the drug dose and schedule to be further evaluated in subsequent trials. Conventionally, there is an underlying assumption that the dosage of a drug is related to undesirable (toxic) outcomes and that higher toxicity implies higher activity. This relationship could be assumed as monotonic or non-monotonic . Most methods for phase I trials are based on the assumption of monotonicity; namely, the probability of a dose-limiting toxicity (DLT) occurring in a patient is monotonically increasing with dose . Consequently, toxicity has traditionally and most widely been used as the primary endpoint for phase I trials evaluating cytotoxic agents. In light of the risk of severe side effects, such trials are performed as dose-escalation studies, wherein the dose of an investigational treatment is levelled up or down based on the DLT assessment from patients already treated. To target the primary objective of identifying the maximum tolerated dose (MTD) for cytotoxic therapies in oncology, a large number of dose-escalation approaches have been proposed: algorithmic approaches (including the popular 3 + 3 design, ), model-based designs (e.g. ) and curve-free methods (e.g. ). A comparison of these classes of trial designs can be found in . Note that the approaches developed during the era of cytotoxic agents assume that both toxicity and activity increase with higher doses. With the emergence of molecularly targeted anticancer agents, alternative endpoints to define optimal biological activity are starting to be used more commonly. Specifically, pharmacokinetic and pharmacodynamic effects, such as plasma drug concentration and target inhibition in tumour or surrogate tissues, have been advised as an alternative primary endpoint besides toxicity . As a result, a shift from traditional phase I designs to innovative seamless phase I/II designs, with added flexibility and special considerations, is currently ongoing. Even though it has been demonstrated that seamless designs are an efficient and flexible approach in simulation studies and some examples of recent applications of seamless phase I/II trials are available (e.g. ), the uptake of seamless phase I/II trials in clinical practice is still slow . In this work we seek to provide some insights in the benefits of modern seamless trial methodology and discuss key aspects based on real world examples.
Traditional dose-finding gone wrong Despite all efforts to ensure that the “best” dose is identified during a Phase I trial, sometimes this goal is not achieved. The example of fulvestrant is one such situation, highlighting that rational and rigorous dose-finding studies are required to avoid large phase III registration trials with a wrong or a suboptimal dose. The Phase I clinical trial of fulvestrant demonstrated that a short-acting formulation of fulvestrant administered daily for 7 days before primary breast surgery was well tolerated and had antiestrogenic and antiproliferative effects . Fulvestrant has very poor oral bioavailability so is administered as an intramuscular (i.m.) injection. A Phase II study with the current long-acting formulation, which was administered once monthly, showed that fulvestrant was effective in women who had breast carcinoma that progressed after tamoxifen therapy . In the Phase II trial, there were two long-term dose levels for 19 postmenopausal patients with tamoxifen-resistant advanced breast cancer enrolled. For appraisal of drug safety, patients 1-4 received escalating doses of fulvestrant, starting with 100 mg in the first month and increasing to 250 mg i.m. from the second month onwards, following confirmation of lack of local or systemic drug toxicity at the 100 mg dose. Patients 5-19 received 250 mg/m i.m. from the onset. The results demonstrated a clinical benefit, no serious drug-related and no effect on the pharmacokinetic results, illustrating the clinical activity and tolerability of the fulvestrant at the 250 mg/m dose regimen. Initially, fulvestrant was approved at a single 250 mg i.m. injection per month as a second-line therapy after progression/relapse on antiestrogen therapy. Its clinical efficacy was established in two phase III registration trials , for postmenopausal women with advanced breast cancer. Whilst 250 mg/month of fulvestrant yielded some clinical efficacy, experience and pharmacokinetic and clinical data from phase III trials had led to speculation that there may be scope to improve its clinical efficacy with higher-dose (HD) regimens based on evidence of dose-dependent effect . Subsequently, the recommended dose was revised to 500 mg/m following gains in efficacy with improved progression-free survival, based on evidence from the phase III Comparison of Faslodex in Recurrent or Metastatic Breast Cancer (CONFIRM, NCT00099437, ). The dose-response story of fulvestrant indicates that rigorous and reliable approaches are required to take all available information into account to guide dose-finding studies and to improve the success rate of Phase III clinical trials. In the case of fulvestrant determination of the optimal dose might have been improved by using two different ideas (ideally in conjunction): i by using modern dose-finding methods during the phase I study (see ) ii by seamlessly assessing safety and activity in one study (see ).
Modern model-based dose-finding based on toxicity Following the seminal work on the continual reassessment method (CRM, ), model-based designs have gained increasing popularity for planning phase I dose-finding trials . Such designs have the major advantage of estimating the dose-toxicity relationship based on a statistical model, fitted to the trial data that have been accrued on all doses investigated thus far. The risk of toxicity per dose can then be accurately assessed and, as a consequence, patients have a better chance of receiving a dose that is most likely to be the MTD . In a sequential phase I trial, the dose-toxicity relationship can be re-estimated by the inclusion of data from a new patient cohort. The consequence of using a typical model-based design in phase I trials is generally not a shortening of the study duration (or the number of patients required), but for more accurate estimation of the MTD. The result of this has been shown to yield improved success rates of development programmes as a whole . Despite this, there is room for further improvement to find an optimal dosage of a new treatment, that is the dose with the best safety/activity trade-off, in early drug development and model-based methods can readily be extended for this purpose. In the above and throughout this manuscript, we discuss ideas and principles on the basis of single-agent dose-escalation trials for ease of description and discussion. The general concepts do, however, also extend to dose-combination studies and schedule-finding studies.
How to seamlessly assess safety and efficacy under one protocol In light of the changing landscape of oncology drugs away from cytotoxic drugs, emphasis is now moving away from identifying the MTD but instead to determine the optimal biologic dose ( OBD ) which is viewed as the dose that (optimally) balances the risks of toxicity with the chance of seeing activity of the drug. Studies that seek to find this balance are called seamless Phase I/II trials and are characterised by the aim to identify the optimal dose, as well as the schedule, for late stage (Phase III) assessment within a single study protocol. We have identified three different types of seamless Phase I/II trials, namely: Dose-escalation followed by ad-hoc expansion cohort(s) Dose-escalation followed by expansion cohort(s) informed by statistical considerations Dose-determination with simultaneous evaluation of toxicity and efficacy The first of these is nowadays frequently used in some centers and could therefore be viewed as a traditional approach. We nevertheless included this approach here as it combines the Phase I (safety) and Phase II (activity) objectives in one protocol and hence is a seamless trial. Since data from Phase I do not contribute to the activity assessment and similarly Phase II data do not contribute to dose-identification, however, it will not lead to improved identification of the OBD, but might shorten the development timelines compared to running two separate studies. Below we describe them on the basis of three examples and discuss their merits and demerits. Seamless Phase I/II trial with an ad-hoc cohort expansion conducted a dose-escalation study with an expansion cohort in advanced hepatocellular carcinoma. The objective of the trial was to identify the MTD among three dose levels (0.30mg/kg, 0.60mg/kg, 0.75mg/kg) and study its activity in a larger group of patients. The target toxicity level, i.e. the maximum acceptable risk of toxicity, was however not explicitly specified. The 3+3 design was used for the dose-escalation part and no DLTs were observed at any of the dosing levels (zero out of 12 patients in Phase I) − . The highest dose, 0.75 mg/kg, was consequently declared as the MTD. A further 31 subjects were planned to be enrolled at the declared MTD. However, after the development of grade 4 thrombocytopenia in two subjects during the expansion phase, the MTD was revised to a reduced dose of 0.6 mg/kg. At the time when the safety signal was identified, 16 patients had their treatment already started. The dose for these patients was subsequently de-escalated for later treatment cycles . These were not used in the formal safety analysis. Such a large expansion based on limited data resulted in little learning about the newly declared MTD − only four additional patients were recruited on the declared MTD, 0.60mg/kg during the expansion phase. Data at the end of the expansion phase might suggest that the MTD could have been reconsidered given 1/9 DLTs on 0.60mg/kg (declared MTD) and 0/16 DLTs in the patients who switched to a lower dose. This would depend on the target toxicity level − for example, for a target toxicity level of 30%, one could have concluded that 0.60 mg/kg is underdosing. Finally, as the rule-based 3+3 design was used, the trial missed the opportunity to increase the efficiency through borrowing information across doses (e.g. the more drug the more toxicity risk) and to account for the data in patients with “switched treatment”. A further potential criticism of this design is that the size of the expansion cohort was not based on formal assessment of activity considerations, so that it is unclear if the study had sufficient power to show activity if present. Seamless Phase I/II trial with 3+3 design followed by Simon’s design The Phase I/II study of dasatinib in relapsed or refractory Non-Hodgkin lymphoma (NCT00550615) reported in utilized a seamless design that combined a traditional 3+3 dose-escalation design with a non-randomized Simon’s design in order to evaluate the safety and efficacy of dasatinib. Three dose levels, 100, 150 and 200 mg of dasatinib were investigated and the main objective of the study was to determine the MTD and to assess the objective response rate. The dose found to be the most tolerable and also has the best activity outcomes during the dose-escalation part, was to be evaluated in the second part of the study. For the latter, initially 10 patients were to be recruited and if two or more responded 19 more patients were to be enrolled further. During the dose-escalation phase, the study sequentially allocated 14 patients: three patients received 100 mg, three patients received 150 mg, and eight patients received 200 mg of dasatinib daily. The MTD was determined to be 200 mg as no DLTs were observed. During the initial stage of the Phase II evaluation an increased rate of grade 3 pleural effusions was noted so that the dose in the phase II portion was reduced to 150 mg for the remaining patients in the study. A total of 19 patients were recruited during this part of the study, with 10 patients receiving 200 mg and 9 receiving 150 mg. As a consequence of the dose reduction during the phase II evaluation, further activity appeared to no longer follow the pre-planned Simon’s design and instead overall response assessment across all patients have been reported. Amongst all patients in whom response assessment was possible (24 of 33 patients enrolled both during the dose-escalation and expansion phase) the objective response rate was 29% (7/24) with a 95% confidence interval (CI) of (13%, 51%). The clinical benefit rate was 71% (17/24) with a 95% CI of (49%, 87%). As in the previous example, the use of the 3+3 design meant that there was no formal opportunity to continuously learn from accumulating safety data in the study, a feature that might have been helpful given the safety signals observed during the Phase II evaluation. Seamless Phase I/II with simultaneous evaluation of toxicity and efficacy The Matchpoint trial (ISRCTN98986889, ) was a seamless phase I/II dose finding study, aiming to estimate a tolerable dose that can yield the greatest efficacy of ponatinib, in combination with conventional chemotherapy (FLAG-IDA regimen), for treating patients in the blastic transformation phase of chronic myeloid leukaemia (CML). Four dose levels of ponatinib, 7.5, 15, 30, 45 mg/day, were available for exploration in the study. The study used toxicity during the first cycle of treatment, together with haematologic or cytogenetic response at the end of the first cycle as a measure of activity, to inform dose-escalation decisions. In particular the so-called EffTox design was used to define the trade-off between safety information and activity information. A total of 17 patients were enrolled in successive cohorts, each contributing both safety and activity information towards the decision about subsequent cohorts dose and the final dose recommendation. The dose of 30 mg/day, which is also the starting dose, was shown to provide the best safety/activity trade-off. Hence this dose was recommended at the end of the trial as a tolerable combination in blast phase CML, with promising activity. reviewed several practical challenges of the EffTox design, when implemented in the Matchpoint trial, along with their solutions. Notably, the Matchpoint trial was one of the few published examples of seamless phase I/II studies that had utilised both toxicity and activity information for interim dose selections based on a Bayesian probabilistic model. A strength of the design is that all data contribute to the safety and activity assessment in a formal way. As part of the design the current dose is chosen to hone in on the optimal dose, meaning that exploration of uninteresting doses (e.g. doses that are more toxic but provide limited added activity) is reduced. This strength, however, comes at the cost of more complex statistical modelling and significant amount of additional planning time required to ensure that the design and trade-off between safety and activity is fit for purpose. Reflections on seamless trial examples While the approaches above are commonly referred to in the literature as Phase I/II trials, they are different in both the decision-making process and the information provided in the end. In the Phase I trial followed by a cohort expansion, formal decisions during the doseescalation part of the trial are made on the toxicity/tolerability information only. Looking at the safety data only can facilitate more rapid decision-making as the safety endpoints typically have shorter evaluation windows than activity/efficacy endpoints, and is more straightforward as only one source of information is used. Moreover, both Phase I and Phase II parts of the trial are conducted using the same infrastructure (most of which would be set up for a single trial) and often the same facilities. This can provide further logistical gains to reaching a conclusion faster. This approach can be efficient in terms of collecting more toxicity and activity data on the recommended MTD, as the whole expansion proceeds at a single dose level and the end data are less sparse (and are indeed about the dose of most interest). This can provide good evidence to go into further investigations of this dose. However, this efficiency is subject to expanding at the truly MTD, which is often unlikely, due to the limited data made for the expansion decision. For example, in the advanced hepatocellular carcinoma trial , the expansion was made on a dose that was later declared as overly toxic. A large expansion on a single dose based on limited data may result in quite restricted learning about the new declared MTD. Using a smaller cohort size for the expansion phase, or continuing the dose-escalation trial for an equivalent total sample size, is likely to facilitate a more ethical and statistically efficient learning about the toxicity. It is more ethical because fewer patients are assigned possibly more toxic doses (and hence potentially avoiding some DLTs). It is meanwhile more statistically efficient for assigning more patients close to the finally recommended MTD that supports a better efficient learning about its toxicity. Another crucial aspect of this type of trial is the size of the expansion and the decisionmaking at its end. In such an expansion, there is typically no formal criteria for both. This leads to further inefficiencies in such trials. Without a formal sample size calculation for the expansion cohort(s), it is unclear what effects on the activity can be claimed as clinically meaningful. Moreover, in the absence of a clear decision-making framework, the probability of making incorrect conclusions is also unclear. Both can jeopardise the value of the trial and its acceptability by wider groups. The design as given in the example in with Simon’s two-stage design in the expansion phase, can resolve the latter concerns about the sample size and decision-making. It still facilitates more rapid decision-making in the drug development process through setting up a single infrastructure for both phases. In this sense, such a setting would represent two separately planned trials but run seamlessly for the logistical gains. However, such operationally seamless trials could still suffer from expanding to a large group of patients at once based on the limited data. Simon’s two-stage design, for example, accounts for the efficacy information only in the decision-making yet does not formally include the toxicity assessment. Were the recommended Phase I dose to be considered overly toxic, and a new MTD to be declared during (or after) the Phase II part, there would be limited learning about it due to exhausting all the resources for the toxic dose. A Phase I/II design which analyses both toxicity and efficacy endpoints simultaneously, tackles this deficiency of the cohort expansion designs while maintaining the operational gains of conducting initial safety and efficacy assessments within one established infrastructure. By allowing dose change throughout the whole trial, such Phase I/II trials mitigate the risk of assigning many patients to overly toxic doses, and hence can be considered as a more ethical approach to first-in-patient Phase I studies. Furthermore, by assessing both safety and efficacy, a typical recommendation of such a trial will be the optimal biological dose (OBD), a dose that balances the risks and benefits. While such a trial can be expected to result in fewer patients at the recommended OBD (compared to the expansion), more patients would be treated in the neighbourhood of the OBD that, again, contributes toward focusing on a single possibly incorrect dose. Additionally, by allowing a dose change throughout the trial and assessing both endpoints on them, such trials can provide richer data to establish the dose-toxicity and dose-efficacy relationships of the studied drug. This can facilitate a more efficient recommendation of a dose (or doses) for further evaluation rather than recommending (or not recommending) the one at which the expansion was made. This Phase I/II seamless design approach, however, can be associated with several challenges which are, primarily, logistical. Identification of the OBD requires tracking of two (or more) outcomes throughout the trial that can take more time than when assessing toxicity only. In a multi-site trial, it should also be ensured that all sites have resources to evaluate both outcomes. Furthermore, there might be different evaluation windows for both endpoints with the efficacy window, typically, being longer. While waiting for the complete evaluation can significantly increase the trial duration, there are approaches allowing only partial evaluation of the response but still leading to efficient decision-making. Finally, such a design typically has no formal hypothesis testing as it focuses on the selection of the OBD.
conducted a dose-escalation study with an expansion cohort in advanced hepatocellular carcinoma. The objective of the trial was to identify the MTD among three dose levels (0.30mg/kg, 0.60mg/kg, 0.75mg/kg) and study its activity in a larger group of patients. The target toxicity level, i.e. the maximum acceptable risk of toxicity, was however not explicitly specified. The 3+3 design was used for the dose-escalation part and no DLTs were observed at any of the dosing levels (zero out of 12 patients in Phase I) − . The highest dose, 0.75 mg/kg, was consequently declared as the MTD. A further 31 subjects were planned to be enrolled at the declared MTD. However, after the development of grade 4 thrombocytopenia in two subjects during the expansion phase, the MTD was revised to a reduced dose of 0.6 mg/kg. At the time when the safety signal was identified, 16 patients had their treatment already started. The dose for these patients was subsequently de-escalated for later treatment cycles . These were not used in the formal safety analysis. Such a large expansion based on limited data resulted in little learning about the newly declared MTD − only four additional patients were recruited on the declared MTD, 0.60mg/kg during the expansion phase. Data at the end of the expansion phase might suggest that the MTD could have been reconsidered given 1/9 DLTs on 0.60mg/kg (declared MTD) and 0/16 DLTs in the patients who switched to a lower dose. This would depend on the target toxicity level − for example, for a target toxicity level of 30%, one could have concluded that 0.60 mg/kg is underdosing. Finally, as the rule-based 3+3 design was used, the trial missed the opportunity to increase the efficiency through borrowing information across doses (e.g. the more drug the more toxicity risk) and to account for the data in patients with “switched treatment”. A further potential criticism of this design is that the size of the expansion cohort was not based on formal assessment of activity considerations, so that it is unclear if the study had sufficient power to show activity if present.
The Phase I/II study of dasatinib in relapsed or refractory Non-Hodgkin lymphoma (NCT00550615) reported in utilized a seamless design that combined a traditional 3+3 dose-escalation design with a non-randomized Simon’s design in order to evaluate the safety and efficacy of dasatinib. Three dose levels, 100, 150 and 200 mg of dasatinib were investigated and the main objective of the study was to determine the MTD and to assess the objective response rate. The dose found to be the most tolerable and also has the best activity outcomes during the dose-escalation part, was to be evaluated in the second part of the study. For the latter, initially 10 patients were to be recruited and if two or more responded 19 more patients were to be enrolled further. During the dose-escalation phase, the study sequentially allocated 14 patients: three patients received 100 mg, three patients received 150 mg, and eight patients received 200 mg of dasatinib daily. The MTD was determined to be 200 mg as no DLTs were observed. During the initial stage of the Phase II evaluation an increased rate of grade 3 pleural effusions was noted so that the dose in the phase II portion was reduced to 150 mg for the remaining patients in the study. A total of 19 patients were recruited during this part of the study, with 10 patients receiving 200 mg and 9 receiving 150 mg. As a consequence of the dose reduction during the phase II evaluation, further activity appeared to no longer follow the pre-planned Simon’s design and instead overall response assessment across all patients have been reported. Amongst all patients in whom response assessment was possible (24 of 33 patients enrolled both during the dose-escalation and expansion phase) the objective response rate was 29% (7/24) with a 95% confidence interval (CI) of (13%, 51%). The clinical benefit rate was 71% (17/24) with a 95% CI of (49%, 87%). As in the previous example, the use of the 3+3 design meant that there was no formal opportunity to continuously learn from accumulating safety data in the study, a feature that might have been helpful given the safety signals observed during the Phase II evaluation.
The Matchpoint trial (ISRCTN98986889, ) was a seamless phase I/II dose finding study, aiming to estimate a tolerable dose that can yield the greatest efficacy of ponatinib, in combination with conventional chemotherapy (FLAG-IDA regimen), for treating patients in the blastic transformation phase of chronic myeloid leukaemia (CML). Four dose levels of ponatinib, 7.5, 15, 30, 45 mg/day, were available for exploration in the study. The study used toxicity during the first cycle of treatment, together with haematologic or cytogenetic response at the end of the first cycle as a measure of activity, to inform dose-escalation decisions. In particular the so-called EffTox design was used to define the trade-off between safety information and activity information. A total of 17 patients were enrolled in successive cohorts, each contributing both safety and activity information towards the decision about subsequent cohorts dose and the final dose recommendation. The dose of 30 mg/day, which is also the starting dose, was shown to provide the best safety/activity trade-off. Hence this dose was recommended at the end of the trial as a tolerable combination in blast phase CML, with promising activity. reviewed several practical challenges of the EffTox design, when implemented in the Matchpoint trial, along with their solutions. Notably, the Matchpoint trial was one of the few published examples of seamless phase I/II studies that had utilised both toxicity and activity information for interim dose selections based on a Bayesian probabilistic model. A strength of the design is that all data contribute to the safety and activity assessment in a formal way. As part of the design the current dose is chosen to hone in on the optimal dose, meaning that exploration of uninteresting doses (e.g. doses that are more toxic but provide limited added activity) is reduced. This strength, however, comes at the cost of more complex statistical modelling and significant amount of additional planning time required to ensure that the design and trade-off between safety and activity is fit for purpose.
While the approaches above are commonly referred to in the literature as Phase I/II trials, they are different in both the decision-making process and the information provided in the end. In the Phase I trial followed by a cohort expansion, formal decisions during the doseescalation part of the trial are made on the toxicity/tolerability information only. Looking at the safety data only can facilitate more rapid decision-making as the safety endpoints typically have shorter evaluation windows than activity/efficacy endpoints, and is more straightforward as only one source of information is used. Moreover, both Phase I and Phase II parts of the trial are conducted using the same infrastructure (most of which would be set up for a single trial) and often the same facilities. This can provide further logistical gains to reaching a conclusion faster. This approach can be efficient in terms of collecting more toxicity and activity data on the recommended MTD, as the whole expansion proceeds at a single dose level and the end data are less sparse (and are indeed about the dose of most interest). This can provide good evidence to go into further investigations of this dose. However, this efficiency is subject to expanding at the truly MTD, which is often unlikely, due to the limited data made for the expansion decision. For example, in the advanced hepatocellular carcinoma trial , the expansion was made on a dose that was later declared as overly toxic. A large expansion on a single dose based on limited data may result in quite restricted learning about the new declared MTD. Using a smaller cohort size for the expansion phase, or continuing the dose-escalation trial for an equivalent total sample size, is likely to facilitate a more ethical and statistically efficient learning about the toxicity. It is more ethical because fewer patients are assigned possibly more toxic doses (and hence potentially avoiding some DLTs). It is meanwhile more statistically efficient for assigning more patients close to the finally recommended MTD that supports a better efficient learning about its toxicity. Another crucial aspect of this type of trial is the size of the expansion and the decisionmaking at its end. In such an expansion, there is typically no formal criteria for both. This leads to further inefficiencies in such trials. Without a formal sample size calculation for the expansion cohort(s), it is unclear what effects on the activity can be claimed as clinically meaningful. Moreover, in the absence of a clear decision-making framework, the probability of making incorrect conclusions is also unclear. Both can jeopardise the value of the trial and its acceptability by wider groups. The design as given in the example in with Simon’s two-stage design in the expansion phase, can resolve the latter concerns about the sample size and decision-making. It still facilitates more rapid decision-making in the drug development process through setting up a single infrastructure for both phases. In this sense, such a setting would represent two separately planned trials but run seamlessly for the logistical gains. However, such operationally seamless trials could still suffer from expanding to a large group of patients at once based on the limited data. Simon’s two-stage design, for example, accounts for the efficacy information only in the decision-making yet does not formally include the toxicity assessment. Were the recommended Phase I dose to be considered overly toxic, and a new MTD to be declared during (or after) the Phase II part, there would be limited learning about it due to exhausting all the resources for the toxic dose. A Phase I/II design which analyses both toxicity and efficacy endpoints simultaneously, tackles this deficiency of the cohort expansion designs while maintaining the operational gains of conducting initial safety and efficacy assessments within one established infrastructure. By allowing dose change throughout the whole trial, such Phase I/II trials mitigate the risk of assigning many patients to overly toxic doses, and hence can be considered as a more ethical approach to first-in-patient Phase I studies. Furthermore, by assessing both safety and efficacy, a typical recommendation of such a trial will be the optimal biological dose (OBD), a dose that balances the risks and benefits. While such a trial can be expected to result in fewer patients at the recommended OBD (compared to the expansion), more patients would be treated in the neighbourhood of the OBD that, again, contributes toward focusing on a single possibly incorrect dose. Additionally, by allowing a dose change throughout the trial and assessing both endpoints on them, such trials can provide richer data to establish the dose-toxicity and dose-efficacy relationships of the studied drug. This can facilitate a more efficient recommendation of a dose (or doses) for further evaluation rather than recommending (or not recommending) the one at which the expansion was made. This Phase I/II seamless design approach, however, can be associated with several challenges which are, primarily, logistical. Identification of the OBD requires tracking of two (or more) outcomes throughout the trial that can take more time than when assessing toxicity only. In a multi-site trial, it should also be ensured that all sites have resources to evaluate both outcomes. Furthermore, there might be different evaluation windows for both endpoints with the efficacy window, typically, being longer. While waiting for the complete evaluation can significantly increase the trial duration, there are approaches allowing only partial evaluation of the response but still leading to efficient decision-making. Finally, such a design typically has no formal hypothesis testing as it focuses on the selection of the OBD.
Challenges and opportunities for seamless trials In this section, we reflect on the different approaches to seamless Phase I/II designs and specifically consider good practice, the often missed and future opportunities to improve early phase oncology trials. Good practice One principal feature of seamless clinical trial designs is the efficient use of all available data for decision making. As a first step, it means that model-based and model-assisted designs should be routinely used instead of algorithmic approaches such as the 3+3 design. In phase I/II dose-finding trials, it is strongly advisable to collect data on both toxicity and efficacy outcomes from each patient throughout the trial and incorporate them in the decision making throughout. Conducting the trial first with a dose-escalation procedure before an embedded proof-of-concept study may sometimes be desirable. In those circumstances, however, it is essential to continue monitoring toxicity data at the individual level until trial completion. When the dose selection decisions in the phase I component are based on toxicity data only, collecting efficacy data from this stage can still be advantageous. Because through combining these data with the newly accrued efficacy data from the phase II component, higher statistical power to test the treatment effect and a better characterisation of the dose-efficacy relationship is possible. Full specification of the decision criteria, including the dose (de-)escalation rules, when and how to expand on certain selected dose(s), etc., should be precisely outlined in the protocol. Any unplanned decision making may induce bias and possibly yield a suboptimal dose for further evaluation in a confirmatory setting. Moreover, the size of the expansion cohort(s) should be based on statistical considerations to ensure that any findings can be meaningfully interpreted. This also means that the method for determining the sample size of the expansion cohort(s) should be provided in the protocol. As noted earlier, expansion on several doses rather than a single dose can be more ethical and statistically efficient, since the MTD may be incorrectly selected from the dose-escalation procedure based on limited information - a problem observed in two of our examples. Frequently missed opportunities Above we have made the argument that collecting both safety and efficacy data throughout the study and incorporating activity data from the Phase I dose-finding component in the Phase II evaluation is efficient, even when only considering a single selected dose to explore further. Similar to the use of dose-toxicity models that govern the dose-escalation, dose-activity models can be used to obtain a better understanding of the activity profile and hence help to identify the dose with the best trade-off between safety and activity. This would be particularly relevant, when multiple doses are expanded. Similarly, using a model to describe the dose-toxicity relationship means that the model can continue to be updated once the Phase II part has been initiated. This has the advantage that learning about the MTD continues throughout the study so that necessary changes to the MTD can be identified based on the same coherent framework. In addition, it would often be more quickly than when relying on ad-hoc adjustments and replaces the need for safety run-ins at Phase II trials. In this case it is helpful to have pre-determined rules to guide the continued safety monitoring in a similar fashion as is routinely done during the Phase I portion of the study. For example, one can use the same rule to determine the MTD as during the Phase I portion (e.g., the dose whose probability of having a toxicity risk between 25 and 35 % is largest), but only make a dose change after at least 5 patients have been recruited in the Phase II portion of the study to avoid changing the dose to quickly. Alternatively, a minimum difference of 5% in the estimated toxicity risk between the dose used in the Phase II portion and the MTD according to the model could be used. Another area of often missed opportunity relates to the utility of pharmacokinetic (PK) information. It is sometimes argued that the dose of a treatment is meaningless on its own right, as what reaches the relevant site is important. PK information often plays a secondary role in dose-escalation decisions. This is often due to the late availability of PK information on the current cohort. More precisely, due to a lack of resources and/or the time required to analyse the samples, PK information would likely be collected and interpreted only after the next dose-escalation decision has been made. When this hurdle can be overcome, however, PK information can be incorporated directly in the dose selection decision in real time (e.g. ) or alternatively exposure-toxicity/exposure-activity (e.g. ) can be considered directly instead of dose-toxicity/dose-activity. Seamless Phase I/II trial with 3+3 design followed by Simon’s design revisited To illustrate the potential benefit of simultaneously assessing safety and efficacy we revisit the Phase I/II study of dasatinib (NCT00550615) reported in . In this illustration we use the EffTox design by instead of the original design and that a toxicity risk of at most 15% is deemed acceptable since observing 2 grade 3 pleural effusions among 10 patients triggered a dose reduction. shows the allocation of patients for the first 24 patients and assumes that safety is revisited in the Phase II portion after every 5 patients. For this sequence of observations we can see that a dose reduction is already recommended by the model after the first 5 patients in the Phase II portion. Note that we did not have the information on which patient had a DLT or response and hence randomly generated when toxicities and activity was observed in line with the frequencies reported for the trial. Different sequences of observations might, however, result in different recommendations. Future opportunities One of the most exciting future opportunities for seamless Phase I/II trials in oncology is the use of novel biomarkers such as circulating tumour DNA (ctDNA) and novel imaging techniques which are promising tools to identify if a patient reacts favourably (or insufficiently) to a treatment much faster than traditional metrics such as progression-free survival. Such biomarker data can then be used to either determine sufficient activity (e.g. favourable ctDNA decrease) or lack thereof (e.g. no decrease in ctDNA levels after initial treatment period) quickly enabling these activity data to be readily available for dose selection decisions during the escalation phase. A challenge to embrace these future opportunities is the limited availability of methods to exploit these biomarkers at present and hence future work in this area is required. Similarly, use of PK-PD models to guide dose-escalation decisions has great promise. Due to the challenges in obtaining PK data sufficiently quickly (as outlined above) and the time required to build rather complex models, this promise is yet to be realised. One possible way forward could be to utilise PK-PD models that have been established for similar compounds (e.g. with the same mechanism of action) in the first instance and consecutively refine these models as data on the novel compound are observed. In this case one could well imagine the use of PK-PD models linked to multiple PD makers of interest. Finally, an often overlooked aspect in Phase I/II dose-escalation trials is patient heterogeneity. Some statistical methods to account for subgroups have been developed; see for example. These often rely on subgroups defined through a single biomarker-defined subgroup. The hope is that novel methods that allow personalization of the optimal dose during the Phase I/II trial can be developed. This is not only to ensure appropriate dosing would be used in the confirmatory trials and beyond, but also that participation in the study maximises the patients benefit.
One principal feature of seamless clinical trial designs is the efficient use of all available data for decision making. As a first step, it means that model-based and model-assisted designs should be routinely used instead of algorithmic approaches such as the 3+3 design. In phase I/II dose-finding trials, it is strongly advisable to collect data on both toxicity and efficacy outcomes from each patient throughout the trial and incorporate them in the decision making throughout. Conducting the trial first with a dose-escalation procedure before an embedded proof-of-concept study may sometimes be desirable. In those circumstances, however, it is essential to continue monitoring toxicity data at the individual level until trial completion. When the dose selection decisions in the phase I component are based on toxicity data only, collecting efficacy data from this stage can still be advantageous. Because through combining these data with the newly accrued efficacy data from the phase II component, higher statistical power to test the treatment effect and a better characterisation of the dose-efficacy relationship is possible. Full specification of the decision criteria, including the dose (de-)escalation rules, when and how to expand on certain selected dose(s), etc., should be precisely outlined in the protocol. Any unplanned decision making may induce bias and possibly yield a suboptimal dose for further evaluation in a confirmatory setting. Moreover, the size of the expansion cohort(s) should be based on statistical considerations to ensure that any findings can be meaningfully interpreted. This also means that the method for determining the sample size of the expansion cohort(s) should be provided in the protocol. As noted earlier, expansion on several doses rather than a single dose can be more ethical and statistically efficient, since the MTD may be incorrectly selected from the dose-escalation procedure based on limited information - a problem observed in two of our examples.
Above we have made the argument that collecting both safety and efficacy data throughout the study and incorporating activity data from the Phase I dose-finding component in the Phase II evaluation is efficient, even when only considering a single selected dose to explore further. Similar to the use of dose-toxicity models that govern the dose-escalation, dose-activity models can be used to obtain a better understanding of the activity profile and hence help to identify the dose with the best trade-off between safety and activity. This would be particularly relevant, when multiple doses are expanded. Similarly, using a model to describe the dose-toxicity relationship means that the model can continue to be updated once the Phase II part has been initiated. This has the advantage that learning about the MTD continues throughout the study so that necessary changes to the MTD can be identified based on the same coherent framework. In addition, it would often be more quickly than when relying on ad-hoc adjustments and replaces the need for safety run-ins at Phase II trials. In this case it is helpful to have pre-determined rules to guide the continued safety monitoring in a similar fashion as is routinely done during the Phase I portion of the study. For example, one can use the same rule to determine the MTD as during the Phase I portion (e.g., the dose whose probability of having a toxicity risk between 25 and 35 % is largest), but only make a dose change after at least 5 patients have been recruited in the Phase II portion of the study to avoid changing the dose to quickly. Alternatively, a minimum difference of 5% in the estimated toxicity risk between the dose used in the Phase II portion and the MTD according to the model could be used. Another area of often missed opportunity relates to the utility of pharmacokinetic (PK) information. It is sometimes argued that the dose of a treatment is meaningless on its own right, as what reaches the relevant site is important. PK information often plays a secondary role in dose-escalation decisions. This is often due to the late availability of PK information on the current cohort. More precisely, due to a lack of resources and/or the time required to analyse the samples, PK information would likely be collected and interpreted only after the next dose-escalation decision has been made. When this hurdle can be overcome, however, PK information can be incorporated directly in the dose selection decision in real time (e.g. ) or alternatively exposure-toxicity/exposure-activity (e.g. ) can be considered directly instead of dose-toxicity/dose-activity.
To illustrate the potential benefit of simultaneously assessing safety and efficacy we revisit the Phase I/II study of dasatinib (NCT00550615) reported in . In this illustration we use the EffTox design by instead of the original design and that a toxicity risk of at most 15% is deemed acceptable since observing 2 grade 3 pleural effusions among 10 patients triggered a dose reduction. shows the allocation of patients for the first 24 patients and assumes that safety is revisited in the Phase II portion after every 5 patients. For this sequence of observations we can see that a dose reduction is already recommended by the model after the first 5 patients in the Phase II portion. Note that we did not have the information on which patient had a DLT or response and hence randomly generated when toxicities and activity was observed in line with the frequencies reported for the trial. Different sequences of observations might, however, result in different recommendations.
One of the most exciting future opportunities for seamless Phase I/II trials in oncology is the use of novel biomarkers such as circulating tumour DNA (ctDNA) and novel imaging techniques which are promising tools to identify if a patient reacts favourably (or insufficiently) to a treatment much faster than traditional metrics such as progression-free survival. Such biomarker data can then be used to either determine sufficient activity (e.g. favourable ctDNA decrease) or lack thereof (e.g. no decrease in ctDNA levels after initial treatment period) quickly enabling these activity data to be readily available for dose selection decisions during the escalation phase. A challenge to embrace these future opportunities is the limited availability of methods to exploit these biomarkers at present and hence future work in this area is required. Similarly, use of PK-PD models to guide dose-escalation decisions has great promise. Due to the challenges in obtaining PK data sufficiently quickly (as outlined above) and the time required to build rather complex models, this promise is yet to be realised. One possible way forward could be to utilise PK-PD models that have been established for similar compounds (e.g. with the same mechanism of action) in the first instance and consecutively refine these models as data on the novel compound are observed. In this case one could well imagine the use of PK-PD models linked to multiple PD makers of interest. Finally, an often overlooked aspect in Phase I/II dose-escalation trials is patient heterogeneity. Some statistical methods to account for subgroups have been developed; see for example. These often rely on subgroups defined through a single biomarker-defined subgroup. The hope is that novel methods that allow personalization of the optimal dose during the Phase I/II trial can be developed. This is not only to ensure appropriate dosing would be used in the confirmatory trials and beyond, but also that participation in the study maximises the patients benefit.
|
Variability in cadmium tolerance of closely related | 17893321-68aa-4125-a9aa-a784f81f1bfb | 11784300 | Microbiology[mh] | Listeria monocytogenes is a Gram-positive, facultative anaerobe with one of the highest mortality rates among the foodborne pathogens. Its ubiquity in natural and agricultural environments, as well as its psychrotrophic abilities, make this pathogen a particular challenge for the food industry. L. monocytogenes has shown an ability to persist in dairy processing environments, including cheese processing facilities, sometimes for years . Typically, persistence is defined as the repeated isolation of identical subtypes of L. monocytogenes over an extended period of time . Subtyping classification of strains differs by study and may range from non-sequence-based methods, such as pulsed-field gel electrophoresis and serotyping, to sequence-based methods, such as single nucleotide polymorphisms (SNPs) . While reasons behind the persistence of L. monocytogenes are typically complex and multifaceted, cadmium tolerance (minimum inhibitory concentration [MIC] ≥256 ppm) has been shown to be more common in resident strains in comparison to sporadic or transient strains from raw milk and non-dairy foods . This is hypothesized to be due to environmental cadmium contamination from anthropogenic sources, making it advantageous for L. monocytogenes to adopt these mechanisms . While cadmium can be naturally found alongside zinc ore in the earth’s crust, anthropogenic sources are responsible for much of the metal found in the environment . These sources range from cadmium byproducts in inorganic phosphate fertilizer to mining, to burning coal . Worldwide, there is between 0.01 and 1 mg/kg of cadmium in the soil, with an average of 0.36 mg/kg . Cadmium enters the cells through divalent cation channels (Mg 2+ , Mn 2+ , Zn 2+ ) . It is toxic to bacteria, as it binds to sulfhydryl groups of essential proteins (e.g., respiratory proteins) and generates free oxidative radicals . Bacteria, including L. monocytogenes, have two main strategies to manage intracellular heavy metal concentration: efflux pumps that actively remove heavy metal ions and small cysteine-rich proteins that sequester heavy metal ions . Cadmium efflux pump genes in L. monocytogenes are well documented, with the first cadmium efflux pump, cadA1 , found on a plasmid-associated transposon Tn 5422 nearly 30 years ago . The second cadmium efflux pump, CadA2, was discovered on plasmid pLM80, and has similar lengths and aa homology to CadA1 (711 aa for CadA1 and 705 aa for CadA2; 69% identity) . Additional cadmium efflux pumps, CadA3 and CadA4, have been identified on the chromosome of some L. monocytogenes strains , with CadA3 being quite rare . Notably, CadA4 has higher divergence compared to the first three cadmium efflux pumps and has been associated with lower cadmium resistance . Other efflux pump genes that act on heavy metals have been documented in L. monocytogenes , including mdrL ; however, it has not been associated with cadmium tolerance . All cadmium-specific efflux pump operons in L. monocytogenes have a similar structure, with a DNA-binding transcriptional regulator, encoded by cadC , followed by the efflux pump gene, cadA . All cadA genes in L. monocytogenes encode for efflux pumps belonging to ATP-binding cassette (ABC) family . The ABC family of efflux pumps relies on ATP hydrolysis to transport cadmium up its concentration gradient and out of the cell . Phenotypic assessment of cadmium resistance spurred the initial characterization of L. monocytogenes plasmids ; however, there are few recent phenotypic studies and little consistency in the assay format, cadmium salts, or cadmium concentrations used. A common method used in the cadmium studies involves spotting cultures on agar plates supplemented with either a single concentration of cadmium or a range of different concentrations and assessing the bacterial growth after incubation . Another method is similar to the disk diffusion assay commonly used for antibiotic testing, where filters impregnated with cadmium salt solutions are placed onto Petri dishes seeded with a high level of bacteria and a zone of inhibition is measured after incubation . Both of these methods provide semi-quantitative results. A more quantitative method is similar to an antibiotic broth dilution assay that evaluates the growth kinetics of bacteria inoculated in liquid media containing a range of cadmium salt concentrations . The broth diffusion assay, used in combination with a spectrophotometer, facilitates the collection of data points to create growth curves based on changes in optical density, which can then be modeled to determine lag phase duration (LPD), maximum growth rate, and/or maximum cell density. Measuring these parameters supports quantitative and statistical comparisons of cadmium response, including the impacts of cation concentration on individual strains or differences between strains. Here, we examined the effect of cadmium salts on the growth behavior of closely (<16 SNPs) and distantly related L. monocytogenes . We primarily tested a set of isolates ( n = 88) from five different dairy facilities collected over the span of a decade, from 2007 to 2017. A majority of the strains within this culture set were isolated from a single facility over a span of 7 years. Prior genomic analysis found that 63 of these strains had very limited genetic variation (<16 SNPs; core genome) and were considered to represent a persistent clonal group . Because extensive phenotypic assessment of cadmium resistance is rare in the literature, these data contribute to better understanding of the interplay between cadmium tolerance and the persistence of L. monocytogenes . Bacterial strains and inoculum preparation Listeria monocytogenes isolates previously identified as possessing variants of the cadAC operon were used to assess the ability of strains to grow in the presence of different cadmium salts and different concentrations . A larger set of Listeria monocytogenes strains ( n = 88; Table S1) previously isolated from dairy processing facilities in British Columbia, Canada, between 2010 and 2017 were used to evaluate strain-level variability in response to cadmium exposure, particularly between strains possessing cadA1C1 . Isolates were resuscitated from frozen stock culture (−80°C) in tryptic soy broth (TSB, Neogen, Lansing, MI) with incubation at 37°C for 24 h. Cultures were then streaked on tryptic soy agar (TSA, Neogen) and incubated at 37°C for 24 h. A single isolated colony was transferred to TSB and incubated at 37°C for 24 h. The overnight culture was diluted in Mueller-Hinton broth (MHB, BD Difco, Franklin Lakes, NJ) to achieve a cell density of approximately 1 × 10 7 CFU/mL (colony forming units/mL). This diluted culture served as the inoculum for MIC and growth kinetic studies. Inoculum was used immediately after preparation (within 1 h). MIC and kinetic growth parameters The impact of cadmium on the growth kinetics of L. monocytogenes strains was determined using a 96-well plate assay. Cadmium salts, CdCl 2 (MilliporeSigma, St. Louis, Missouri) and CdSO 4 (MilliporeSigma, St. Louis, Missouri), were dissolved in MHB to create stock solutions (750 mM) and filter sterilized using a 0.45 mm filter (VWR, Radnor, Pennsylvania). The stock solutions were diluted to targeted test concentrations (40 µM–120 µM) using MHB. MHB + Cd 2+ solutions were aliquoted (195 µL) into a 96-well plate (Corning Falcon, Corning, NY). MHB without the addition of Cd 2+ served as the control medium. Wells were inoculated with L. monocytogenes inoculum (5 µL) to achieve an approximate initial cell density of 1 × 10 5 CFU/mL. The 96-well plate was placed in a 96-well spectrophotometer (FilterMax F5, Molecular Devices, San Jose, CA) for controlled incubation at 37°C for 24 h. These conditions are representative of optimum growth conditions for L. monocytogenes and were selected to support comparisons with prior studies . Optical densities (OD 595nm ) were recorded every 10 min after a 3-s plate agitation. Growth curve determinations for cadAC variants were performed in triplicate for each treatment combination, with experiments repeated on 2 separate days (biological replication). MICs were assigned for each strain as the minimum Cd 2+ concentration that prevented growth (no significant increase in OD 595nm compared to baseline) throughout the 24-h incubation period. L. monocytogenes isolates from dairy facilities ( n = 88) were screened for phenotypic cadmium tolerance based on their ability to grow in the presence of 8 ppm (43.8 µM) CdCl 2 . Screening experiments were performed in triplicate within the same 96-well plate (technical replicates). The majority of growth curves (93.4%) were as expected; however, aberrant growth patterns were observed in 6.6% of wells and classified into four categories: baseline drift, baseline shift, late control, and erratic early growth. Wells with a baseline drift were deemed irremediable and removed ( n = 2). Wells with a baseline shift were defined as having growth that was noticeably shifted higher and were corrected by normalizing with another well ( n = 15). Late control wells were noticed where one control well started growing noticeably later than the other two wells and were ultimately removed ( n = 2). Erratic early growth was deemed to be associated with a measurement error by the spectrophotometer. When the erratic nature was resolved by 5.8 h (early in lag phase), these early OD values were deleted, and the growth curve reconstructed ( n = 9). When the erratic nature had not resolved itself by 5.8 h, the well was removed ( n = 7). Using the adjusted growth curves data ( n = 517), growth curve parameters (LPD [h], and maximum cell density [OD]) were calculated using the nonlinear fit curve (mechanistic growth model with inverse prediction; OD cutoff = 0.11, inverse prediction OD = 0.09) in JMP Pro 16.0.0 (SAS Institute, Cary, NC). OD values were selected by preliminary visualization of the entire data set to ensure all samples with exponential growth were included in the analysis. Due to variability in starting cell densities between strains and replicates, LPDs of control wells were subtracted from the LPD of MHB + Cd wells for each individual strain. These values were reported as increase in LPD and they were used for all statistical comparisons. A multiple linear regression model was generated to assess differences between the two tested cadmium salts, CdCl 2 and CdSO 4 , while also controlling for the effects of different concentrations and cadmium genes. The linear model was used where the cadmium salt serves as an indicator variable. μ ( L P D | C o n c e n t r a t i o n , c a d G e n e c a d A 1 , c a d A 2 , c a d A 3 , c a d A 4 , C d S a l t ) = β 0 + β 1 C o n c e n t r a t i o n + β 2 c a d G e n e c a d A 1 , c a d A 2 , c a d A 3 , c a d A 4 + β 3 C d S a l t Interaction terms were not included as their addition unnecessarily increased the complexity of the model. All statistical analyses and graphs were created using R (4.1.3) and RStudio (2023.06.0+421, Posit, Boston, MA) with the tidyverse (1.3.2) package. Long-read sequencing and genome resolution of selected L. monocytogenes strains All the L. monocytogenes strains used in this study had previously been whole genome sequenced using short-read sequencing technology by our laboratory or by other investigators . Six L. monocytogenes strains displaying relative differences in cadmium response (slow growth – WRLP95, LPD increase >5 h; intermediate growth – WRLP17, WRLP46, and WRLP77, LPD increase 2–3 h; fast growth – WRLP76 and WRLP81, LPD increase <2 h) were selected for genome resolution using long-read sequencing. Strains were resuscitated from frozen stock in TSB with 0.6% yeast extract (TSBYE; Neogen, Lansing, MI), incubated at 37°C for 24 h. Cultures were then streaked for isolation on TSA with 0.6% yeast extract (Neogen, Lansing, MI), incubated at 37°C for 24 h. A single colony was transferred to TSBYE and incubated at 37°C for 24 h to achieve high cell density. High molecular weight DNA was extracted using the Qiagen Blood & Cell Culture DNA kit (Qiagen, Hilden, Germany) with Qiagen Genomic-tip 100/G following the manufacturer’s protocol. DNA extracts were submitted to Oregon State University’s Center for Quantitative Life Sciences (Corvallis, OR) for genomic DNA quality determination (concentration and size distribution) using the Agilent TapeStation 4200 (Santa Clara, CA). DNA extracts were prepared for long-read sequencing using the Native Barcoding Kit (Oxford Nanopore Technologies, Oxford, UK) with the Rapid Sequencing Barcoding protocol. Sequencing was performed using the MinION flow cell with R10.4.1 chemistry type. Genome assembly and annotation were performed using Bacterial and Viral Bioinformatics Resource Center (BV-BRC) Comprehensive Genome Analysis with 50,000 of the raw reads from the long-read sequencing (described above) and previously reported short-read sequencing of these strains (BioProject PRJNA998448 ) . Sequences were assembled using Unicycler v.0.4.8 with the following parameters: trimming reads before assembly; two racon iterations; two pilon iterations; a minimum contig length of 300 bp; and minimum contig coverage of five . Unicycler was chosen as it has been shown to increase the likelihood of plasmid recovery . Genomes were annotated using RAST tool kit v.1.3.0 . Genomes and plasmids were visualized and compared using BLAST Ring Image Generator (BRIG; v.0.95) . Multiple sequence alignments were performed using MAFFT v.7.4.5 on BV-BRC . Variation analysis on the clonal group was performed using BV-BRC using BWA-mem v.0.7.17 as the aligner and FreeBayes v.1.3.6 for SNP calling . Listeria monocytogenes isolates previously identified as possessing variants of the cadAC operon were used to assess the ability of strains to grow in the presence of different cadmium salts and different concentrations . A larger set of Listeria monocytogenes strains ( n = 88; Table S1) previously isolated from dairy processing facilities in British Columbia, Canada, between 2010 and 2017 were used to evaluate strain-level variability in response to cadmium exposure, particularly between strains possessing cadA1C1 . Isolates were resuscitated from frozen stock culture (−80°C) in tryptic soy broth (TSB, Neogen, Lansing, MI) with incubation at 37°C for 24 h. Cultures were then streaked on tryptic soy agar (TSA, Neogen) and incubated at 37°C for 24 h. A single isolated colony was transferred to TSB and incubated at 37°C for 24 h. The overnight culture was diluted in Mueller-Hinton broth (MHB, BD Difco, Franklin Lakes, NJ) to achieve a cell density of approximately 1 × 10 7 CFU/mL (colony forming units/mL). This diluted culture served as the inoculum for MIC and growth kinetic studies. Inoculum was used immediately after preparation (within 1 h). The impact of cadmium on the growth kinetics of L. monocytogenes strains was determined using a 96-well plate assay. Cadmium salts, CdCl 2 (MilliporeSigma, St. Louis, Missouri) and CdSO 4 (MilliporeSigma, St. Louis, Missouri), were dissolved in MHB to create stock solutions (750 mM) and filter sterilized using a 0.45 mm filter (VWR, Radnor, Pennsylvania). The stock solutions were diluted to targeted test concentrations (40 µM–120 µM) using MHB. MHB + Cd 2+ solutions were aliquoted (195 µL) into a 96-well plate (Corning Falcon, Corning, NY). MHB without the addition of Cd 2+ served as the control medium. Wells were inoculated with L. monocytogenes inoculum (5 µL) to achieve an approximate initial cell density of 1 × 10 5 CFU/mL. The 96-well plate was placed in a 96-well spectrophotometer (FilterMax F5, Molecular Devices, San Jose, CA) for controlled incubation at 37°C for 24 h. These conditions are representative of optimum growth conditions for L. monocytogenes and were selected to support comparisons with prior studies . Optical densities (OD 595nm ) were recorded every 10 min after a 3-s plate agitation. Growth curve determinations for cadAC variants were performed in triplicate for each treatment combination, with experiments repeated on 2 separate days (biological replication). MICs were assigned for each strain as the minimum Cd 2+ concentration that prevented growth (no significant increase in OD 595nm compared to baseline) throughout the 24-h incubation period. L. monocytogenes isolates from dairy facilities ( n = 88) were screened for phenotypic cadmium tolerance based on their ability to grow in the presence of 8 ppm (43.8 µM) CdCl 2 . Screening experiments were performed in triplicate within the same 96-well plate (technical replicates). The majority of growth curves (93.4%) were as expected; however, aberrant growth patterns were observed in 6.6% of wells and classified into four categories: baseline drift, baseline shift, late control, and erratic early growth. Wells with a baseline drift were deemed irremediable and removed ( n = 2). Wells with a baseline shift were defined as having growth that was noticeably shifted higher and were corrected by normalizing with another well ( n = 15). Late control wells were noticed where one control well started growing noticeably later than the other two wells and were ultimately removed ( n = 2). Erratic early growth was deemed to be associated with a measurement error by the spectrophotometer. When the erratic nature was resolved by 5.8 h (early in lag phase), these early OD values were deleted, and the growth curve reconstructed ( n = 9). When the erratic nature had not resolved itself by 5.8 h, the well was removed ( n = 7). Using the adjusted growth curves data ( n = 517), growth curve parameters (LPD [h], and maximum cell density [OD]) were calculated using the nonlinear fit curve (mechanistic growth model with inverse prediction; OD cutoff = 0.11, inverse prediction OD = 0.09) in JMP Pro 16.0.0 (SAS Institute, Cary, NC). OD values were selected by preliminary visualization of the entire data set to ensure all samples with exponential growth were included in the analysis. Due to variability in starting cell densities between strains and replicates, LPDs of control wells were subtracted from the LPD of MHB + Cd wells for each individual strain. These values were reported as increase in LPD and they were used for all statistical comparisons. A multiple linear regression model was generated to assess differences between the two tested cadmium salts, CdCl 2 and CdSO 4 , while also controlling for the effects of different concentrations and cadmium genes. The linear model was used where the cadmium salt serves as an indicator variable. μ ( L P D | C o n c e n t r a t i o n , c a d G e n e c a d A 1 , c a d A 2 , c a d A 3 , c a d A 4 , C d S a l t ) = β 0 + β 1 C o n c e n t r a t i o n + β 2 c a d G e n e c a d A 1 , c a d A 2 , c a d A 3 , c a d A 4 + β 3 C d S a l t Interaction terms were not included as their addition unnecessarily increased the complexity of the model. All statistical analyses and graphs were created using R (4.1.3) and RStudio (2023.06.0+421, Posit, Boston, MA) with the tidyverse (1.3.2) package. L. monocytogenes strains All the L. monocytogenes strains used in this study had previously been whole genome sequenced using short-read sequencing technology by our laboratory or by other investigators . Six L. monocytogenes strains displaying relative differences in cadmium response (slow growth – WRLP95, LPD increase >5 h; intermediate growth – WRLP17, WRLP46, and WRLP77, LPD increase 2–3 h; fast growth – WRLP76 and WRLP81, LPD increase <2 h) were selected for genome resolution using long-read sequencing. Strains were resuscitated from frozen stock in TSB with 0.6% yeast extract (TSBYE; Neogen, Lansing, MI), incubated at 37°C for 24 h. Cultures were then streaked for isolation on TSA with 0.6% yeast extract (Neogen, Lansing, MI), incubated at 37°C for 24 h. A single colony was transferred to TSBYE and incubated at 37°C for 24 h to achieve high cell density. High molecular weight DNA was extracted using the Qiagen Blood & Cell Culture DNA kit (Qiagen, Hilden, Germany) with Qiagen Genomic-tip 100/G following the manufacturer’s protocol. DNA extracts were submitted to Oregon State University’s Center for Quantitative Life Sciences (Corvallis, OR) for genomic DNA quality determination (concentration and size distribution) using the Agilent TapeStation 4200 (Santa Clara, CA). DNA extracts were prepared for long-read sequencing using the Native Barcoding Kit (Oxford Nanopore Technologies, Oxford, UK) with the Rapid Sequencing Barcoding protocol. Sequencing was performed using the MinION flow cell with R10.4.1 chemistry type. Genome assembly and annotation were performed using Bacterial and Viral Bioinformatics Resource Center (BV-BRC) Comprehensive Genome Analysis with 50,000 of the raw reads from the long-read sequencing (described above) and previously reported short-read sequencing of these strains (BioProject PRJNA998448 ) . Sequences were assembled using Unicycler v.0.4.8 with the following parameters: trimming reads before assembly; two racon iterations; two pilon iterations; a minimum contig length of 300 bp; and minimum contig coverage of five . Unicycler was chosen as it has been shown to increase the likelihood of plasmid recovery . Genomes were annotated using RAST tool kit v.1.3.0 . Genomes and plasmids were visualized and compared using BLAST Ring Image Generator (BRIG; v.0.95) . Multiple sequence alignments were performed using MAFFT v.7.4.5 on BV-BRC . Variation analysis on the clonal group was performed using BV-BRC using BWA-mem v.0.7.17 as the aligner and FreeBayes v.1.3.6 for SNP calling . Growth behavior of L. monocytogenes strains with unique cadAC variants and their response to different cadmium salts All four cadA ( cadA1–cadA4 ) variants tested managed CdCl 2 more effectively than CdSO 4 . This was unexpected as the two salts dissolve readily in water although their overall solubility is different, at 133 mg/mL and 75 mg/mL, respectively . Representative CdCl 2 dose response curves for the growth of the cadA - strains (WRLP85) compared to the cadA4 + Scott A strain are shown in . The cadA - isolate, WRLP85, was able to tolerate a small amount of CdCl 2 (MIC 21.9 µM), whereas Scott A ( cadA4 ) possessed highest tolerance, at MIC of 175 µM. WRLP10, possessing cadA1 , had a CdCl 2 MIC of 95.4 µM. Similar levels of cadmium tolerance in L. monocytogenes , when tested in broth, were reported in the literature . In contrast, the MIC values seen in the present study were far below the values reported in the literature when using agar-based methods. Previous studies have used 382 µM (70 ppm) CdCl 2 as the MIC cutoff determination, where isolates that were able to grow at or above this concentration were deemed tolerant to cadmium . In the original literature, these isolates were called “resistant”; however, considering the concentrations tested and controversy regarding concentrations that would deem isolates resistant for the consistency purpose here, we refer to them as “tolerant.” This is in agreement with L. Jiang , who reported that microbroth dilution assays generally yield lower MIC results compared to agar dilution methods. It may be that the diffusion of cadmium in broth is higher than on agar due to Brownian motion, which also increases the chances of impact with the bacterial membrane as well as its fluidity. The ability of the cadA - isolate WRLP85 to grow in 10.9 µM of cadmium implies a baseline tolerance of L. monocytogenes to cadmium. A previous study had similar findings; when cadA was deleted, the MIC decreased from 100 µM in the wild-type strain to 10 µM in the mutant strain . Increasing the CdCl 2 dose caused corresponding increases in LPD, with growth observed at concentrations up to 87.5 µM for Scott A . For the same strain, no growth was seen at 175 µM CdCl 2 (MIC), indicating that this concentration of cadmium overwhelmed the ability of cadmium tolerance mechanisms to overcome this stress. At concentrations of at least 43.8 µM, Scott A also exhibited a decrease in the final OD, suggesting potential ATP expenditure to support efflux pump activity at the expense of biomass production. Similar observations have been made in Pseudomonas aeruginosa during antibiotic resistance development, where the over-expression of an efflux pump is associated with the increased expression of the nitrate respiratory chain to make up for the increased energy demand . The most common cadmium salts used in the studies are CdCl 2 and CdSO 4 ; however, there has been little investigation into the effect of two compounds on L. monocytogenes phenotype. When we exposed L. monocytogenes strains possessing different cadA variants to CdCl 2 and CdSO 4 , at concentrations between 40 and 120 µM, there was a significant difference in LPD of strains at 120 µM for the tested cadmium salts . Notably, the effect was the most pronounced for the cadA4 + strain (Scott A; the strain with the highest CdCl 2 MIC) with an average LPD increase of 2.26 h for CdCl 2 compared to 5.83 h for CdSO 4 . The individual regression graphs demonstrated that Scott A ( cadA4 + ) struggled to adapt to increasingly higher concentrations of CdSO 4 when compared to the strains carrying the other cadmium resistance genes . WRLP81, which possesses cadA2, had the smallest differences in LPD between the two cadmium salts as shown by the slope of the regression lines (CdCl 2 : 0.03 h/µM vs CdSO 4 : 0.05 h/µM). Using a multiple linear regression model, on average CdSO 4 increased strains’ LPD over CdCl 2 by 2.25 h (± 0.34, 95% CI), when controlling for the effect of the different cadmium genes and holding the concentration constant ( P ≤ 0.05). A more complex model considering the various slopes of the different cadmium genes incurred a loss of precision and interpretability due to redundant terms. Caution should be used if applying this model for future predictions, especially for isolates such as Scott A, that are particularly sensitive to CdSO 4 concentration. Prior studies on cadmium tolerance most commonly used CdCl 2 or CdSO 4 ; however, the two salts have only been compared against L. monocytogenes by R. Pombinho et al. . Using a disk diffusion assay, they found no difference in response between the two salts; however, this method is unlikely to provide the sensitivity to discern differences. Since cadmium salts impact the growth kinetics of L. monocytogenes, methods that evaluate growth at an endpoint, such as disk diffusion assays and agar dilutions, would be minimally influenced by cadmium salt selection, assuming incubation time was sufficient. Cadmium salt hydrates, including CdSO 4 •8H 2 O and CdCl 2 •H 2 O, have also been used in prior studies, albeit far less commonly . Most cadmium studies use a single salt and report concentration as a function of mass (ppm), making conversion to molarity essential prior to comparing data across studies. Screening L. monocytogenes isolates recovered from Canadian dairy facilities from 2007 to 2017 for cadmium tolerance Based on cadmium growth kinetics for control strains with various cadA variants, 43.8 µM (8 ppm) of CdCl 2 was selected as the concentration for screening L. monocytogenes isolates from dairy processing facilities ( n = 88) for cadmium tolerance. Representative growth curves for cadA1 + (WRLP 46) and cadA - (WRLP 94) strains are shown in . At this concentration, none of the cadA - isolates were able to grow while all cadA + ( n = 67) isolates grew, albeit with a significant delay compared to MHB without added cadmium ( P < 0.05, two sample t -test). As cadA - isolates failed to grow, they were omitted from subsequent data analysis. The average LPD for control wells was 8.54 ± 0.79 h compared to 11.26 ± 1.08 h average LPD for cadA + isolates grown in the presence of cadmium. cadA + strains also exhibited a significantly lower final OD when grown in the presence of 43.8 µM CdCl 2 compared to no cadmium ( P < 0.05, two sample t -test). The average final OD reading for strains grown in the presence of cadmium was 0.128 ± 0.03 compared to 0.172 ± 0.01 for no cadmium controls. Increase in LPD due to the presence of 43.8 µM of CdCl 2 of cadA + dairy isolates is shown in . Within the tested isolate set, 63/67 cadA + isolates were part of a persistent clonal group (<16 SNPs, core genome) , isolated from a single facility. Two additional strains (WRLP76, WRLP79) were isolated from the same facility, years apart, and very closely related (<33 SNPs from clonal group). All of these strains carried cadA1C1 on an 87 kb plasmid . They also carried a cadmium resistance protein (133 aa; ) adjacent to the 627 aa soft metal ATPase on the same plasmid. Compared to other Listeria sequences in the NCBI database using BLASTp, the cadmium resistance protein found in these isolates is truncated, with the first 53 aa missing. The average LPD increase in the presence of cadmium for strains in this persistent cluster was 2.69 ± 0.7 h. For statistical analysis purposes, WRLP46 was selected as a strain representing an average increase in LPD (2.55 ± 0.2 h) when grown in the presence of CdCl 2 . The LDP of WRLP22 (0.99 ± 0.14 h) and WRLP51 (1.35 ± 0.2 h) was significantly less impacted by cadmium when compared to WRLP46 ( P < 0.001). The presence of cadmium led to significantly higher LPD for WRLP62 (3.99 ± 0.26 h), WRLP13 (4.45 ± 0.18 h), and WRLP14 (4.56 ± 0.49 h) compared to WRLP46. Plasmid mapping and variation analysis of these isolates demonstrated that the plasmids have identical sequences (0 SNPs; ). Further evaluation of the plasmids in the entire persistent clonal group revealed that WRLP77 was the only isolate that possessed a plasmid with a variation in the sequence; a ~2,100 bp deletion was seen immediately downstream of the Tn 5422 cluster containing cadA1 . SNP variation analysis between WRLP46 and isolates that exhibited either highest (WRLP13, WRLP14) or lowest LPD increase (WRLP22, WRLP51) in the presence of cadmium was also performed. WRLP13 and WRLP14 had two conserved non-synonymous SNPs in an NtrC family transcriptional regulator (I746T) and a sodium/hydrogen exchanger family protein (E54D). Two of the fastest growing isolates that belong to the clonal group, WRLP22 and WRLP51, only had conserved synonymous SNPs in phage components. There were additional SNPs identified (WRLP51, 18 SNPs; WRLP14, 46 SNPs); however, no patterns based on LPD differences were observed ( ; Table S2). Similarly, the comparative systems analysis identifying the presence/absence of genes within this strain set provided limited information. Only five coding regions differed across the set, with two being associated with phage structural components, one putative integrase, and two hypothetical proteins. There was no pattern based on LPD differences. The two remaining cadA + dairy isolates, WRLP81 and WRLP95, were isolated from other dairy facilities and possessed cadA1 and cadA2 on 69 kb and 64 kb plasmids, respectively . The growth of WRLP95 was particularly impacted by cadmium, with an increase in LPD of 6.44 ± 0.08 h (Table S2) despite having identical cadA1 sequences to all other strains in the set. This discrepancy among cadA + isolates has been seen before, with up to a 860 µg/mL (4691 µM, converted) difference in MICs between two cadA1 isolates . Additionally, variation in gene expression has been seen in closely related strains of the same serotype . The cadA1 gene in WRLP95 was conserved in association with Tn 5422 but in an opposite orientation compared to WRLP46 . These plasmids share limited homology (33%). This is predominantly in the regions associated with Tn 5422 as well as the region containing the 627 aa soft metal ATPase, Pli0046, betaine ABC transporter, and the NADH dehydrogenase, likely components of a mobile element. WRLP95 was the only isolate within the set to also carry a bcrABC cassette, associated with an efflux pump that confers tolerance to quaternary ammonium compounds (QACs). Research has shown a correlation between cadA genes and the QAC tolerance gene bcrABC , and a potential link between these genes and increased likelihood of isolate persistence in food production environments . In L. monocytogenes isolates with bcrABC , it is highly likely the isolates will also possess a cadA gene; however, the reverse is not as consistent . This is in line with our data, where isolates within the same clonal group all possessed cadA1 , but only one had bcrABC . The co-selection of these two genes was also demonstrated previously, when the transfer of both bcrABC and cadA to L. monocytogenes was consistent enough that acquiring cadA was an indicator that an isolate had also received bcrABC . WRLP81, the only cadA2 + isolate in the strain set, showed high tolerance to cadmium with LPD of 1.32 ± 0.17 h. It was grouped with more cadmium-tolerant cadA1 isolates. This is not surprising, considering cadA1 and cadA2 share 70% aa identity (Fig. S1) and have been shown to have similar MICs . Genomic properties of control and cadA - strains The chromosomes of the control strains (WRLP85, WRLP46, WRLP81, and EGD-e) were mapped to L. monocytogenes Scott A ( cadA4 + ), and genes associated with soft metal tolerance are shown in . L. monocytogenes EGD-e, Scott A, and WRLP85 do not carry plasmids. Plasmid maps for WRLP81 ( cadA2 + ) and WRLP95 ( cadA1 +) are displayed in , respectively. The plasmid map of WRLP46 ( cadA1 + ) and additional representative isolates from clonal group isolated from Facility 71 (e.g., increased cadmium tolerance: WRLP22, WRLP51; reduced cadmium tolerance: WRLP13, WRLP14) as well as two closely related isolates from Facility 71 (WRLP76 and WRLP79) are shown in . Additional details related to soft metal-associated genes for the control strains are highlighted on the chromosome and plasmid maps and presented in . Other genetic features of interest (i.e., transposases, repeat regions, and the sanitizer efflux pump bcrABC ) are also highlighted on the plasmid maps. The most well-studied of the soft metal-associated genes in L. monocytogenes are the cadmium-specific efflux pumps ( cadA ) and their nearby regulators ( cadC ). These cassettes share reasonably similar homology (Fig. S1), but they differ in their mobility and localization within the genome. Scott A possesses the cadA4C4 variant located on the Listeria genomic island 2, which also carries several genes associated with arsenic tolerance in the 2.41 Mb region . EGD-e carries the cadA3, ispB, and cadC3 cassette within an integrative and conjugative element in a distant location on the chromosome (~1.13 Mb) . WRLP81 carries cadA2C2 on a putative Tn 552 within a ~69 kb plasmid . WRLP95 possesses cadA1C1 on a ~64 kb plasmid . WRLP46 and 65 of the other dairy isolates from Facility 71 also carry cadA1C1 , located on a ~87 kb plasmid . Both plasmids carrying cadA1C1 are associated with Tn 5422 ; however, the two plasmids share only 33% homology. Many of the L. monocytogenes strains, including WRLP85 and 21 of the other dairy isolates, do not harbor any plasmids. These strains do not possess a cadAC cassette within their genomes. All six examined L. monocytogenes strains carry three small cobalt, zinc, and cadmium resistance proteins (286, 291, and 303 aa) on the chromosome, encoded by czc genes. L. monocytogenes serotype 1/2a isolates (e.g., EGD-e, WRLP46, WRLP85, and WRLP95) have identical predicted amino acid sequences for all three czc genes, whereas sequences for Scott A (serotype 4b) and WRLP81 (serotype 1/2b) differ from the serotype 1/2a isolates but are identical to one another. Predicted amino acid sequences for the czc genes differ between the two groups by 4 (Q138R, F163L, R242K, E262A), 1 (T222A), and 1 (S256T) amino acid for the 286, 291, and 303 aa proteins, respectively. These czc genes are dispersed throughout the chromosome but are highly conserved within their own groups with the czc genes associated with 286 aa and 291 aa proteins existing within conserved regions of >45,000 bp and >25,000 bp, respectively. The czc gene ( czcD ) with 303 aa protein length also sits in a highly conserved region of >30,000 bp; however, there are two options for genes immediately upstream of the czcD and downstream of a Cof-like hydrolase: (i) a putative peptidoglycan bound protein (LPTXG motif) (EGD-e, WRLP46, WRLP85, WRLP95) or (ii) a mobile element protein (Scott A, WRLP81). czcD has been shown to have a significant increase in expression directly after cadmium exposure that soon levels off . It was hypothesized that czcD was therefore a regulatory gene that was upregulated upon exposure to activate the relevant genes . Given the size of CzcD, it is unlikely that it is an efflux pump; however, it is somewhat similar in size to the CadA regulatory protein, CadC (protein length 119 aa for CadC1). Due to the high conservation of czc genes, these genes likely play a role in low-level cadmium tolerance of L. monocytogenes . All six strains also carry two soft metal efflux ATPases on the chromosome (626 aa and 737 aa long). The 626 aa protein is in a highly conserved region (20,000 bp) and the 737 aa protein is in a conserved region of >65,000 bp. Both have the conserved DKTGTLT ATP phosphorylation site for P-type ATPases; however, only the 737 aa protein has the CXXC metal binding motif . Considering the longer 737 aa ATPase has both requisite ATP phosphorylation site and the heavy metal binding site, it is a possible reason for the inherent cadmium tolerance observed in the cadA - isolate, WRLP85. The shorter of the two proteins, 626 aa, lacks the CXXC metal binding motif, making it less likely to contribute to cadmium tolerance . To our knowledge, no one has tested either of these efflux pumps phenotypically in L. monocytogenes . Again, homology of these proteins was conserved within the serogrouping. WRLP46 plasmid, and by extension of the rest of the plasmid-harboring isolates from Facility 71, carried a (third) soft metal efflux ATPase (627 aa) adjacent to a short protein (133 aa) that was annotated as a “cadmium resistance protein.” This short “cadmium resistance protein” shares a low homology (24.3% aa identity) with CadX, a negative transcriptional regulator in Streptococcus salivarius . WRLP95 and WRLP81 plasmids also carry the 627 aa soft metal ATPase and both plasmids carry a second (fourth), but different, soft metal ATPase (WRLP95, 681 aa; WRLP81, 653 aa). Alignments of these soft metal ATPases are provided in Fig. S1. None of the three of proteins (627 aa, 681 aa, 653 aa) have the metal binding motif and only the 627 aa and 681 aa proteins have the ATP phosphorylation site. It is unclear if any of these other annotated ATPases play a role in conferring cadmium tolerance in isolates that possess them; however, the only metal ATPase in isolates WRLP95 and WRLP81 other than cadA to have both the necessary CXXC metal binding motif and the ATP phosphorylation cite is the 737 aa protein that is shared among all of the isolates listed in . Conclusion Phenotypic responses of the L. monocytogenes isolates examined in the present study varied in the presence of cadmium. When isolates were exposed to two commonly studied cadmium salts, CdCl 2 and CdSO 4 , the companion anion significantly affected growth kinetics of L. monocytogenes strains. Isolates possessing different cadmium resistance genes, namely cadA1, cadA2, cadA3, or cadA4 , were able to grow faster in CdCl 2 than CdSO 4 . A low level of inherent cadmium tolerance (~10 µM) was also observed in L. monocytogenes isolates that do not carry any plasmids associated with cadA and other soft metal resistance genes. Despite genetic similarity, within a clonal group with over 60 strains possessing identical plasmids with cadA genes, there were phenotypic differences observed in the presence of cadmium. These findings enhance our understanding of L. monocytogenes cadmium tolerance; however, further research is needed to discover the underlying genetic and physiological factors involved in cadmium tolerance. To the best of our knowledge, no studies have examined the difference in genetic expression among isolates with small SNP differences. Further transcriptomics testing would be useful to establish differences in gene expression among closely related isolates in relation to their tolerance of cadmium and other compounds encountered in the food production chain. L. monocytogenes strains with unique cadAC variants and their response to different cadmium salts All four cadA ( cadA1–cadA4 ) variants tested managed CdCl 2 more effectively than CdSO 4 . This was unexpected as the two salts dissolve readily in water although their overall solubility is different, at 133 mg/mL and 75 mg/mL, respectively . Representative CdCl 2 dose response curves for the growth of the cadA - strains (WRLP85) compared to the cadA4 + Scott A strain are shown in . The cadA - isolate, WRLP85, was able to tolerate a small amount of CdCl 2 (MIC 21.9 µM), whereas Scott A ( cadA4 ) possessed highest tolerance, at MIC of 175 µM. WRLP10, possessing cadA1 , had a CdCl 2 MIC of 95.4 µM. Similar levels of cadmium tolerance in L. monocytogenes , when tested in broth, were reported in the literature . In contrast, the MIC values seen in the present study were far below the values reported in the literature when using agar-based methods. Previous studies have used 382 µM (70 ppm) CdCl 2 as the MIC cutoff determination, where isolates that were able to grow at or above this concentration were deemed tolerant to cadmium . In the original literature, these isolates were called “resistant”; however, considering the concentrations tested and controversy regarding concentrations that would deem isolates resistant for the consistency purpose here, we refer to them as “tolerant.” This is in agreement with L. Jiang , who reported that microbroth dilution assays generally yield lower MIC results compared to agar dilution methods. It may be that the diffusion of cadmium in broth is higher than on agar due to Brownian motion, which also increases the chances of impact with the bacterial membrane as well as its fluidity. The ability of the cadA - isolate WRLP85 to grow in 10.9 µM of cadmium implies a baseline tolerance of L. monocytogenes to cadmium. A previous study had similar findings; when cadA was deleted, the MIC decreased from 100 µM in the wild-type strain to 10 µM in the mutant strain . Increasing the CdCl 2 dose caused corresponding increases in LPD, with growth observed at concentrations up to 87.5 µM for Scott A . For the same strain, no growth was seen at 175 µM CdCl 2 (MIC), indicating that this concentration of cadmium overwhelmed the ability of cadmium tolerance mechanisms to overcome this stress. At concentrations of at least 43.8 µM, Scott A also exhibited a decrease in the final OD, suggesting potential ATP expenditure to support efflux pump activity at the expense of biomass production. Similar observations have been made in Pseudomonas aeruginosa during antibiotic resistance development, where the over-expression of an efflux pump is associated with the increased expression of the nitrate respiratory chain to make up for the increased energy demand . The most common cadmium salts used in the studies are CdCl 2 and CdSO 4 ; however, there has been little investigation into the effect of two compounds on L. monocytogenes phenotype. When we exposed L. monocytogenes strains possessing different cadA variants to CdCl 2 and CdSO 4 , at concentrations between 40 and 120 µM, there was a significant difference in LPD of strains at 120 µM for the tested cadmium salts . Notably, the effect was the most pronounced for the cadA4 + strain (Scott A; the strain with the highest CdCl 2 MIC) with an average LPD increase of 2.26 h for CdCl 2 compared to 5.83 h for CdSO 4 . The individual regression graphs demonstrated that Scott A ( cadA4 + ) struggled to adapt to increasingly higher concentrations of CdSO 4 when compared to the strains carrying the other cadmium resistance genes . WRLP81, which possesses cadA2, had the smallest differences in LPD between the two cadmium salts as shown by the slope of the regression lines (CdCl 2 : 0.03 h/µM vs CdSO 4 : 0.05 h/µM). Using a multiple linear regression model, on average CdSO 4 increased strains’ LPD over CdCl 2 by 2.25 h (± 0.34, 95% CI), when controlling for the effect of the different cadmium genes and holding the concentration constant ( P ≤ 0.05). A more complex model considering the various slopes of the different cadmium genes incurred a loss of precision and interpretability due to redundant terms. Caution should be used if applying this model for future predictions, especially for isolates such as Scott A, that are particularly sensitive to CdSO 4 concentration. Prior studies on cadmium tolerance most commonly used CdCl 2 or CdSO 4 ; however, the two salts have only been compared against L. monocytogenes by R. Pombinho et al. . Using a disk diffusion assay, they found no difference in response between the two salts; however, this method is unlikely to provide the sensitivity to discern differences. Since cadmium salts impact the growth kinetics of L. monocytogenes, methods that evaluate growth at an endpoint, such as disk diffusion assays and agar dilutions, would be minimally influenced by cadmium salt selection, assuming incubation time was sufficient. Cadmium salt hydrates, including CdSO 4 •8H 2 O and CdCl 2 •H 2 O, have also been used in prior studies, albeit far less commonly . Most cadmium studies use a single salt and report concentration as a function of mass (ppm), making conversion to molarity essential prior to comparing data across studies. L. monocytogenes isolates recovered from Canadian dairy facilities from 2007 to 2017 for cadmium tolerance Based on cadmium growth kinetics for control strains with various cadA variants, 43.8 µM (8 ppm) of CdCl 2 was selected as the concentration for screening L. monocytogenes isolates from dairy processing facilities ( n = 88) for cadmium tolerance. Representative growth curves for cadA1 + (WRLP 46) and cadA - (WRLP 94) strains are shown in . At this concentration, none of the cadA - isolates were able to grow while all cadA + ( n = 67) isolates grew, albeit with a significant delay compared to MHB without added cadmium ( P < 0.05, two sample t -test). As cadA - isolates failed to grow, they were omitted from subsequent data analysis. The average LPD for control wells was 8.54 ± 0.79 h compared to 11.26 ± 1.08 h average LPD for cadA + isolates grown in the presence of cadmium. cadA + strains also exhibited a significantly lower final OD when grown in the presence of 43.8 µM CdCl 2 compared to no cadmium ( P < 0.05, two sample t -test). The average final OD reading for strains grown in the presence of cadmium was 0.128 ± 0.03 compared to 0.172 ± 0.01 for no cadmium controls. Increase in LPD due to the presence of 43.8 µM of CdCl 2 of cadA + dairy isolates is shown in . Within the tested isolate set, 63/67 cadA + isolates were part of a persistent clonal group (<16 SNPs, core genome) , isolated from a single facility. Two additional strains (WRLP76, WRLP79) were isolated from the same facility, years apart, and very closely related (<33 SNPs from clonal group). All of these strains carried cadA1C1 on an 87 kb plasmid . They also carried a cadmium resistance protein (133 aa; ) adjacent to the 627 aa soft metal ATPase on the same plasmid. Compared to other Listeria sequences in the NCBI database using BLASTp, the cadmium resistance protein found in these isolates is truncated, with the first 53 aa missing. The average LPD increase in the presence of cadmium for strains in this persistent cluster was 2.69 ± 0.7 h. For statistical analysis purposes, WRLP46 was selected as a strain representing an average increase in LPD (2.55 ± 0.2 h) when grown in the presence of CdCl 2 . The LDP of WRLP22 (0.99 ± 0.14 h) and WRLP51 (1.35 ± 0.2 h) was significantly less impacted by cadmium when compared to WRLP46 ( P < 0.001). The presence of cadmium led to significantly higher LPD for WRLP62 (3.99 ± 0.26 h), WRLP13 (4.45 ± 0.18 h), and WRLP14 (4.56 ± 0.49 h) compared to WRLP46. Plasmid mapping and variation analysis of these isolates demonstrated that the plasmids have identical sequences (0 SNPs; ). Further evaluation of the plasmids in the entire persistent clonal group revealed that WRLP77 was the only isolate that possessed a plasmid with a variation in the sequence; a ~2,100 bp deletion was seen immediately downstream of the Tn 5422 cluster containing cadA1 . SNP variation analysis between WRLP46 and isolates that exhibited either highest (WRLP13, WRLP14) or lowest LPD increase (WRLP22, WRLP51) in the presence of cadmium was also performed. WRLP13 and WRLP14 had two conserved non-synonymous SNPs in an NtrC family transcriptional regulator (I746T) and a sodium/hydrogen exchanger family protein (E54D). Two of the fastest growing isolates that belong to the clonal group, WRLP22 and WRLP51, only had conserved synonymous SNPs in phage components. There were additional SNPs identified (WRLP51, 18 SNPs; WRLP14, 46 SNPs); however, no patterns based on LPD differences were observed ( ; Table S2). Similarly, the comparative systems analysis identifying the presence/absence of genes within this strain set provided limited information. Only five coding regions differed across the set, with two being associated with phage structural components, one putative integrase, and two hypothetical proteins. There was no pattern based on LPD differences. The two remaining cadA + dairy isolates, WRLP81 and WRLP95, were isolated from other dairy facilities and possessed cadA1 and cadA2 on 69 kb and 64 kb plasmids, respectively . The growth of WRLP95 was particularly impacted by cadmium, with an increase in LPD of 6.44 ± 0.08 h (Table S2) despite having identical cadA1 sequences to all other strains in the set. This discrepancy among cadA + isolates has been seen before, with up to a 860 µg/mL (4691 µM, converted) difference in MICs between two cadA1 isolates . Additionally, variation in gene expression has been seen in closely related strains of the same serotype . The cadA1 gene in WRLP95 was conserved in association with Tn 5422 but in an opposite orientation compared to WRLP46 . These plasmids share limited homology (33%). This is predominantly in the regions associated with Tn 5422 as well as the region containing the 627 aa soft metal ATPase, Pli0046, betaine ABC transporter, and the NADH dehydrogenase, likely components of a mobile element. WRLP95 was the only isolate within the set to also carry a bcrABC cassette, associated with an efflux pump that confers tolerance to quaternary ammonium compounds (QACs). Research has shown a correlation between cadA genes and the QAC tolerance gene bcrABC , and a potential link between these genes and increased likelihood of isolate persistence in food production environments . In L. monocytogenes isolates with bcrABC , it is highly likely the isolates will also possess a cadA gene; however, the reverse is not as consistent . This is in line with our data, where isolates within the same clonal group all possessed cadA1 , but only one had bcrABC . The co-selection of these two genes was also demonstrated previously, when the transfer of both bcrABC and cadA to L. monocytogenes was consistent enough that acquiring cadA was an indicator that an isolate had also received bcrABC . WRLP81, the only cadA2 + isolate in the strain set, showed high tolerance to cadmium with LPD of 1.32 ± 0.17 h. It was grouped with more cadmium-tolerant cadA1 isolates. This is not surprising, considering cadA1 and cadA2 share 70% aa identity (Fig. S1) and have been shown to have similar MICs . cadA - strains The chromosomes of the control strains (WRLP85, WRLP46, WRLP81, and EGD-e) were mapped to L. monocytogenes Scott A ( cadA4 + ), and genes associated with soft metal tolerance are shown in . L. monocytogenes EGD-e, Scott A, and WRLP85 do not carry plasmids. Plasmid maps for WRLP81 ( cadA2 + ) and WRLP95 ( cadA1 +) are displayed in , respectively. The plasmid map of WRLP46 ( cadA1 + ) and additional representative isolates from clonal group isolated from Facility 71 (e.g., increased cadmium tolerance: WRLP22, WRLP51; reduced cadmium tolerance: WRLP13, WRLP14) as well as two closely related isolates from Facility 71 (WRLP76 and WRLP79) are shown in . Additional details related to soft metal-associated genes for the control strains are highlighted on the chromosome and plasmid maps and presented in . Other genetic features of interest (i.e., transposases, repeat regions, and the sanitizer efflux pump bcrABC ) are also highlighted on the plasmid maps. The most well-studied of the soft metal-associated genes in L. monocytogenes are the cadmium-specific efflux pumps ( cadA ) and their nearby regulators ( cadC ). These cassettes share reasonably similar homology (Fig. S1), but they differ in their mobility and localization within the genome. Scott A possesses the cadA4C4 variant located on the Listeria genomic island 2, which also carries several genes associated with arsenic tolerance in the 2.41 Mb region . EGD-e carries the cadA3, ispB, and cadC3 cassette within an integrative and conjugative element in a distant location on the chromosome (~1.13 Mb) . WRLP81 carries cadA2C2 on a putative Tn 552 within a ~69 kb plasmid . WRLP95 possesses cadA1C1 on a ~64 kb plasmid . WRLP46 and 65 of the other dairy isolates from Facility 71 also carry cadA1C1 , located on a ~87 kb plasmid . Both plasmids carrying cadA1C1 are associated with Tn 5422 ; however, the two plasmids share only 33% homology. Many of the L. monocytogenes strains, including WRLP85 and 21 of the other dairy isolates, do not harbor any plasmids. These strains do not possess a cadAC cassette within their genomes. All six examined L. monocytogenes strains carry three small cobalt, zinc, and cadmium resistance proteins (286, 291, and 303 aa) on the chromosome, encoded by czc genes. L. monocytogenes serotype 1/2a isolates (e.g., EGD-e, WRLP46, WRLP85, and WRLP95) have identical predicted amino acid sequences for all three czc genes, whereas sequences for Scott A (serotype 4b) and WRLP81 (serotype 1/2b) differ from the serotype 1/2a isolates but are identical to one another. Predicted amino acid sequences for the czc genes differ between the two groups by 4 (Q138R, F163L, R242K, E262A), 1 (T222A), and 1 (S256T) amino acid for the 286, 291, and 303 aa proteins, respectively. These czc genes are dispersed throughout the chromosome but are highly conserved within their own groups with the czc genes associated with 286 aa and 291 aa proteins existing within conserved regions of >45,000 bp and >25,000 bp, respectively. The czc gene ( czcD ) with 303 aa protein length also sits in a highly conserved region of >30,000 bp; however, there are two options for genes immediately upstream of the czcD and downstream of a Cof-like hydrolase: (i) a putative peptidoglycan bound protein (LPTXG motif) (EGD-e, WRLP46, WRLP85, WRLP95) or (ii) a mobile element protein (Scott A, WRLP81). czcD has been shown to have a significant increase in expression directly after cadmium exposure that soon levels off . It was hypothesized that czcD was therefore a regulatory gene that was upregulated upon exposure to activate the relevant genes . Given the size of CzcD, it is unlikely that it is an efflux pump; however, it is somewhat similar in size to the CadA regulatory protein, CadC (protein length 119 aa for CadC1). Due to the high conservation of czc genes, these genes likely play a role in low-level cadmium tolerance of L. monocytogenes . All six strains also carry two soft metal efflux ATPases on the chromosome (626 aa and 737 aa long). The 626 aa protein is in a highly conserved region (20,000 bp) and the 737 aa protein is in a conserved region of >65,000 bp. Both have the conserved DKTGTLT ATP phosphorylation site for P-type ATPases; however, only the 737 aa protein has the CXXC metal binding motif . Considering the longer 737 aa ATPase has both requisite ATP phosphorylation site and the heavy metal binding site, it is a possible reason for the inherent cadmium tolerance observed in the cadA - isolate, WRLP85. The shorter of the two proteins, 626 aa, lacks the CXXC metal binding motif, making it less likely to contribute to cadmium tolerance . To our knowledge, no one has tested either of these efflux pumps phenotypically in L. monocytogenes . Again, homology of these proteins was conserved within the serogrouping. WRLP46 plasmid, and by extension of the rest of the plasmid-harboring isolates from Facility 71, carried a (third) soft metal efflux ATPase (627 aa) adjacent to a short protein (133 aa) that was annotated as a “cadmium resistance protein.” This short “cadmium resistance protein” shares a low homology (24.3% aa identity) with CadX, a negative transcriptional regulator in Streptococcus salivarius . WRLP95 and WRLP81 plasmids also carry the 627 aa soft metal ATPase and both plasmids carry a second (fourth), but different, soft metal ATPase (WRLP95, 681 aa; WRLP81, 653 aa). Alignments of these soft metal ATPases are provided in Fig. S1. None of the three of proteins (627 aa, 681 aa, 653 aa) have the metal binding motif and only the 627 aa and 681 aa proteins have the ATP phosphorylation site. It is unclear if any of these other annotated ATPases play a role in conferring cadmium tolerance in isolates that possess them; however, the only metal ATPase in isolates WRLP95 and WRLP81 other than cadA to have both the necessary CXXC metal binding motif and the ATP phosphorylation cite is the 737 aa protein that is shared among all of the isolates listed in . Phenotypic responses of the L. monocytogenes isolates examined in the present study varied in the presence of cadmium. When isolates were exposed to two commonly studied cadmium salts, CdCl 2 and CdSO 4 , the companion anion significantly affected growth kinetics of L. monocytogenes strains. Isolates possessing different cadmium resistance genes, namely cadA1, cadA2, cadA3, or cadA4 , were able to grow faster in CdCl 2 than CdSO 4 . A low level of inherent cadmium tolerance (~10 µM) was also observed in L. monocytogenes isolates that do not carry any plasmids associated with cadA and other soft metal resistance genes. Despite genetic similarity, within a clonal group with over 60 strains possessing identical plasmids with cadA genes, there were phenotypic differences observed in the presence of cadmium. These findings enhance our understanding of L. monocytogenes cadmium tolerance; however, further research is needed to discover the underlying genetic and physiological factors involved in cadmium tolerance. To the best of our knowledge, no studies have examined the difference in genetic expression among isolates with small SNP differences. Further transcriptomics testing would be useful to establish differences in gene expression among closely related isolates in relation to their tolerance of cadmium and other compounds encountered in the food production chain. |
Applying the estimand framework to clinical pharmacology trials with a case study in bioequivalence | 158ea745-3f59-4822-bae9-cc9ec6438789 | 11773105 | Pharmacology[mh] | INTRODUCTION In 2019, the International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (ICH) published the ICH E9(R1) Addendum on Estimands and Sensitivity Analysis in Clinical Trials to the Guideline on Statistical Principles in Clinical Trials. This document provides a framework for defining clear trial objectives, stating that “without such clarity, there is a concern that the reported treatment effect will be misunderstood”. The framework advocates greater transparency and alignment of trial planning, conduct, analysis and interpretation. It encourages deeper thinking and closer collaboration between the trial team and other stakeholders prior to trial start‐up. It also recommends considering whether collecting data after certain events occur is relevant, and it proposes that decisions about data handling be made while planning the trial protocol rather than while drafting or implementing the statistical analysis plan. ICH E9(R1) outlines principles that are “relevant whenever a treatment effect is estimated, or a hypothesis related to a treatment effect is tested” but acknowledges that “regulatory interest […] will be greater for confirmatory clinical trials”. Accordingly, the addendum and its accompanying training slides focus on Phase 3 efficacy trials. How and whether these principles should be implemented in clinical pharmacology trials remain unclear. We believe that the estimand framework may add value if applied to clinical pharmacology trials, but because the setting is very different from Phase 3 efficacy trials, the resulting considerations and estimands also differ. Most clinical pharmacology trials are exploratory and conducted early in the drug development process. However, some clinical pharmacology trials are pivotal to the licensing and labelling of a drug. This includes trials on bioequivalence, food effect and drug–drug interaction, as well as those in specific populations such as hepatic or renally impaired participants. For instance, bioequivalence trials are necessary if the manufacturing process changes. If bioequivalence can be demonstrated, safety and efficacy results are usually extrapolated from an approved formulation to the new formulation. This requires that the new and approved formulations have equivalent rates of absorption and availability of the active ingredient at the site of drug action. In many cases, the availability of the active ingredient is determined using parameters derived from plasma concentration–time curves, in particular, the area under the plasma concentration curve (AUC) and maximum plasma concentration ( C max ), which are often the co‐primary endpoints, along with the corresponding time to C max ( T max ), often a key secondary endpoint. Generally, the primary objective of bioequivalence trials is to confirm that the 90% confidence intervals for the ratio of geometric means of test to reference of the AUC and C max fall within predefined equivalence margins, typically 80% to 125%. Current regulatory guidelines , , , , , , give detailed guidance on the design, conduct and evaluation of bioequivalence. A standard two‐period crossover design is often used, with criteria for excluding participants from the statistical analysis specified a priori. For instance, participants with vomiting or diarrhoea have incomplete drug absorption of an oral product rendering their plasma concentration–time profiles unreliable in the affected period. The data from both unaffected and affected periods of such participants are often excluded from the analysis. However, most of the guidance has been written before the adoption of the ICH E9(R1), and consequently, these guidelines do not use the estimand framework to define the underlying scientific question that needs to be addressed to establish bioequivalence. Clinical pharmacology trials are not explicitly mentioned in ICH E9(R1), and the ICH does not take a position on the role of estimands in confirmatory clinical pharmacology trials. To date, the only recommendation on the use of estimands is from the US Food and Drug Administration (FDA) in their 2022 draft guidance on Statistical Approaches to Establishing Bioequivalence, which requests that sponsors specify estimands in bioequivalence trial protocols. Little has been published on the implementation of the estimand framework in a clinical pharmacology setting. Shortly after the publication of ICH E9(R1), Ring and Wolfsegger discussed the applicability of the estimand framework to clinical pharmacology trials. However, rather than using this framework and considering the scientific question a priori, they inferred estimands using the analysis methods in guidelines from the European Medicines Agency (EMA) and FDA. , Consequently, their estimands incorporated estimation details that should have been specified only in the statistical methods. Akacha et al. described estimands and their importance in clinical pharmacology but did not advise on how the framework can be implemented and the relevant intercurrent events (ICEs) can be handled. In pharmacokinetics (PK), ICEs include events that are expected to impact absorption, distribution, metabolism or excretion and therefore would affect the interpretation of drug or metabolite endpoints. One could argue that clinical pharmacology trials (e.g., bioequivalence) are straightforward, well‐established trials, with clear objectives and designs that are well documented across many regulatory guidelines. , , , , , , So, is there really any need for estimands in a clinical pharmacology setting? In assessment reports for medicinal products submitted to the EMA between 2013 and 2018, 120 out of 551 (23%) Day 120 assessments/products included major objections raised in the clinical pharmacology section of the assessment report. The major objections involved several topics such as pharmacokinetic characterization in the target population, PK/PD relationships, bioavailability and bioequivalence. We anticipate that the application of the estimand framework potentially reduces clinical pharmacology‐related major deficiencies and increases the rate of regulatory success, consistent with the aims of the existing clinical pharmacology guidelines. The Estimand Implementation Working Group (EIWG), co‐sponsored by the European Federation of Statisticians in the Pharmaceutical Industry and the European Federation of Pharmaceutical Industries and Associations, was formed in 2019 to support the implementation of the ICH E9(R1) addendum by sharing experiences across industry stakeholders. , We developed this paper because of the limited literature and regulatory guidance on using estimands in clinical pharmacology trials, as well as the considerable differences between these trials and Phase 3 efficacy trials. The paper, the result of a collaboration between clinical pharmacologists and statisticians from the EIWG, illustrates how the estimand framework , can be applied to clinical pharmacology trials as best practice with an anticipated increase in quality. In it, we first introduce the steps of the estimand thinking process needed to define and implement estimands in trial protocols. This process can help to create a clear link between the trial objective, estimand, trial design and statistical analysis. We illustrate this both with general considerations for clinical pharmacology trials and through a case study of a bioequivalence trial. Additionally, we review current regulatory guidance for bioequivalence trials and discuss the broader applicability of defining estimands in clinical pharmacology trials. GENERAL CONSIDERATIONS ON APPLICATION OF THE ESTIMAND THINKING PROCESS 2.1 What is the estimand framework and thinking process? The estimand framework, as outlined in the ICH E9(R1) addendum, distinguishes between what we want to estimate (the estimand) and how we estimate it (the estimator). Defining the estimand is not a statistical exercise but a multidisciplinary discussion that describes the trial objective in depth and can affect the trial design and conduct as well as the analysis and interpretation of the results. Unfortunately, the conduct of clinical trials rarely goes to plan; the estimand framework ensures that events occurring after treatment initiation that affect the interpretation or existence of the outcome are discussed and documented at the protocol stage, and that strategies are put in place to handle such ICEs. Figure presents an adaptation of the framework and the thinking process. Although this is depicted as a linear process, revisiting previous steps may be warranted if, for example, the planned ICE handling strategies (Step 3) are not feasible due to design considerations, difficulties in collecting the required data or lack of a robust estimator (Step 5). The estimand is the description of the treatment effect we want to estimate, and it should align with the trial objective and the underlying scientific question. It includes treatment conditions, target population, endpoint, population‐level summary and ICEs along with their associated handling strategies. Four of these attributes are typically described (at least partly) in different sections of the trial protocol, but ICEs and their associated handling strategies may be less familiar to the reader. As such, these will be discussed in Steps 2 and 3 below. The considerations described in the steps that follow are intended to be applicable to a variety of clinical pharmacology trials and we will focus on a bioequivalence case study later. 2.1.1 Step 1: Therapeutic setting and intent of treatment determining a trial objective The main questions of interest should be clearly identified. This requires a full understanding of the therapeutic setting and includes considerations about the disease or condition and the clinical context. Questions include: Will the drug be used for diagnosis, prophylaxis or treatment? Which patients would benefit (e.g., paediatric, elderly, renally impaired or requiring concomitant medications)? What is known or expected from the drug metabolism and tolerability? Does the trial aim to find the optimal dosing regimen of a new entity or to understand the relative bioavailability for the new formulation of an already approved product? Does it investigate interactions with other medicines, effects of administration with food or use in special patient populations? For entities not studied before, questions may also concern tolerability and PK: What are the preferred route and dosing regimen? What are the overall exposure and the C max ? How is the drug eliminated and how long does that take? For new formulations, are they designed to improve convenience (for instance, a single high‐dose tablet), tolerability or stability/shelf‐life, or can they be considered a generic? Are they expected to have a slower absorption (for instance, to attain therapeutic concentrations for longer than the comparator formulation)? Discussions around these points help establish the trial objective. 2.1.2 Step 2: Identify intercurrent events ICH E9(R1) defines ICEs as “events occurring after treatment initiation that affect either the interpretation or the existence of the measurements associated with the clinical question of interest”. Identifying ICEs requires discussions between at least clinical pharmacologists, clinicians and statisticians. In clinical pharmacology, ICEs affecting absorption, distribution, metabolism and excretion affect the interpretation of an outcome, whereas terminal events such as death, transplant or amputation affect the outcome's existence. However, the terminal events are unlikely to be relevant in clinical pharmacology unless patients are severely ill or are dosed over an extended period. Note, ICEs should be discussed both in relation to the profile measurements and the endpoint. It may be that a few profile measurements are affected, but still the endpoint can be derived reliably. In other cases, it may also be that the endpoint will be impacted, and the event is therefore to be defined as an ICE. Absorption may be affected by the route of administration. For example, individuals not experienced with inhalers may use them incorrectly, patches may come off, or products intended to be administered subcutaneously may erroneously be given intramuscularly or intravenously. Deviations from dosing instructions may also affect absorption, for example, if the product is taken in the fed state instead of the fasted state, or with a hot drink instead of water, where the coating of a capsule may dissolve faster than intended. Tolerability issues like vomiting and diarrhoea will also affect absorption of an oral drug. By contrast, other ICEs, such as interacting drugs that induce or inhibit metabolism, may affect elimination. To decide if absorption or clearance are affected by ICEs, the timing of the latter needs to be considered. For example, if interacting substances are used well after the C max is expected to be attained, the event is not considered an ICE for C max . Once undesirable ICEs are identified, they can be minimized with good trial conduct and, potentially, training (e.g., on how to use inhalers). This is feasible in the controlled setting typical of clinical pharmacology trials. Table includes some examples of ICEs in clinical pharmacology trials to highlight how they differ from protocol deviations, whilst also acknowledging some overlap. ICEs should not be confused with issues related to missing data, such as missing blood samples or laboratory failings. Missing data issues are not considered ICEs because the underlying level of exposure in the body is unaffected. ICH E9(R1) defines missing data as “data that would be meaningful for the analysis of a given estimand but were not collected” and they should be minimized during trial design and conduct. 2.1.3 Step 3: Discuss strategies to address intercurrent events Table describes five potential strategies to handle ICEs per ICH E9(R1) and how they can be applied in clinical pharmacology. A strategy needs to be selected for each type of ICE, but strategies can vary across ICEs within the same estimand. The choice of strategy should be driven by the intent of the treatment and the trial objectives rather than the trial design and statistical methods. In clinical practice, a generic and originator product will be considered to be interchangeable and thus the focus for demonstrating bioequivalence may be the principal stratum of individuals able to tolerate both products together with demonstrating that the tolerability does not differ between the treatments. In the case of a biosimilar or generic, both would generally be expected to be equally well tolerated, whereas a new formulation or different route of administration may be designed to be better tolerated and the principal stratum of those able to tolerate either formulation may also be of interest together with a demonstration of better tolerability. Note, it may not always be possible to identify the principal stratum of interest. A different strategy, specifically a hypothetical strategy, could be proposed to handle the use of interacting substances to remove their impact on the measurements and to ensure sensitivity to detect any difference between formulations. If a new formulation was designed to improve tolerability, the objective might be to demonstrate not only bioequivalence with the reference formulation in the principal stratum of tolerators but also more consistent exposure in all patients after multiple dosing measured irrespective of any use of other medications, dose interruptions or compliance issues. This interest in a treatment effect that includes the effects of the ICEs is referred to as the treatment policy strategy. It tends to be a common choice in later phase trials, in line with the intent‐to‐treat principle, which promotes the complete follow‐up of participants and use of their data in analysis regardless of whether the ICEs occur. Notably, a higher steady‐state exposure results from different PK properties, such as increased absorption, lower volume of distribution or decreased clearance. A higher exposure may also be associated with poor tolerability, for instance, severe adverse reactions. Therefore, PK could differ between individuals that tolerate a drug well and those who do not. The while on treatment strategy focuses interest on the response prior to the occurrence of ICEs but without implicit or explicit imputation beyond that point. This strategy requires careful consideration of the choice of endpoint. For example, we might consider the proportion of time inhibition of a biomarker is above a threshold or plasma concentrations are within a therapeutic window, from the first dose of a trial drug to the time of treatment discontinuation. The final strategy to consider is the composite variable strategy where certain undesirable ICEs such as tolerability issues would be considered a failure in the endpoint. Although composite binary responder endpoints are being used to assess treatment efficacy in confirmatory trials, we are not aware of much use of composite binary or continuous variables for PK. In the above example evaluating trough exposure in patients after multiple dosing, it might be an option to define a composite trough endpoint as zero (or any other unfavourable value) for patients who discontinue treatment due to a tolerability issue, if that is addressing a relevant question. 2.1.4 Step 4: Precisely define the estimand(s) The trial team should define the treatment effects (estimands) according to the treatment conditions, endpoints, population‐level summary, target population and strategies for ICEs. Depending on the ICEs and the strategy chosen, the precise specification of the treatment conditions, the endpoint or the target population may be impacted. For example, ICEs handled by the principal stratum strategy will be part of the target population, such as “biosimilarity in healthy individuals who do not develop anti‐drug antibodies”. See Table for some other examples. In a PK context, the treatment conditions would often be the two interventions to be compared, that is, the test and reference formulations and would include the route of administration, titration scheme and whether single or multiple dosing is of interest. In bioequivalence trials, PK concentration measurements are used to derive their co‐primary endpoints, typically the C max and the AUC, measured in a dosing interval for multiple dosing or extrapolated to infinity (AUC 0‐ꝏ ) for single dosing. In these trials, the population‐level summary needs to aggregate the data in a way that allows comparison of treatment conditions and to demonstrate equivalence within a specified bioequivalence limit. This is often done using the geometric mean ratio between the test and reference formulations. However, the estimand attributes are interrelated; for example, endpoints that can have zero values (see the composite continuous endpoint described in Table ) will affect the choice of appropriate population‐level summary. In this case, a median difference would be more suitable than a geometric mean ratio. In general, the population‐level summary should be chosen under careful evaluation of the context. Although patients are the ultimate target population, in alignment with guideline recommendations, , , , many clinical pharmacology trials and most bioequivalence trials are conducted in healthy participants. For instance, EMA guidance states: the subject population for bioequivalence studies should be selected with the aim of permitting detection of differences between pharmaceutical products. In order to reduce variability not related to differences between products, the studies should normally be performed in healthy volunteers unless the drug carries safety concerns that make this unethical. This model, in vivo healthy volunteers, is regarded as adequate in most instances to detect formulation differences and to allow extrapolation of the results to populations for which the reference medicinal product is approved (the elderly, children, patients with renal or liver impairment, etc.). In these cases, we recommend that the population attribute of the estimand should be specified as “healthy individuals” or a suitable subset such as “healthy adult males”, but the protocol should explicitly discuss whether it is reasonable to conclude that the two drugs are clinically equivalent in the ultimate patient population if bioequivalence in healthy participants is demonstrated (Figure ). In contrast, if the clinical pharmacology trial is conducted in patients, the target population should be the patients targeted by the scientific question of interest posed by the trial. We would like to emphasize that the terms “participant”, “subject” and “volunteer” are not appropriate when describing the target population of the estimand, as they infer participation in the trial rather than who we seek to study. The entire “population of healthy individuals” will not participate in the trial, so this terminology is distinct from the terminology of “participant”, “subject” or “volunteer”, which are reserved for describing the people enrolled in the trial and the participant sets (Step 5). Instead, the description of the target population should include “patients” or “healthy individuals”. Consequently, referring to “participants meeting eligibility criteria” and “all randomised participants” or other participant sets will not be appropriate for describing the target population. 2.1.5 Step 5: Align choices on trial design, data collection and method of estimation Considerations about trial design, trial conduct, data to be collected, and estimation methods should be aligned with the choice of estimand(s). A crossover design is used in many clinical pharmacology trials, especially if the time required for washout is not prohibitive. Crossovers provide an efficient design in terms of the required sample size and may also facilitate the identification of a principal stratum of interest, if applicable. Such a trial will be most sensitive if certain ICEs can be minimized, for example, if the intake of food and other substances can be controlled. In clinical pharmacology trials, mitigation plans can be developed to minimize the impact of certain ICEs and thereby improve their sensitivity. By contrast, Phase 3 trials may seek to mirror the occurrence of ICEs in real‐world clinical practice, especially when the aim is to apply the treatment policy strategy to all ICEs; in this context, mitigation plans may not be advisable. Consideration should be given to whether the data after an ICE occurs would be relevant in the current and subsequent periods, which would be dependent on the strategy used for the ICE. This may guide decisions during trial conduct such as stopping blood sampling for PK after one participant vomits, but continuing data collection for safety analyses, and then whether the participant should or should not continue into the subsequent dosing period. In some trial designs, such as parallel group trials, participants only receive one formulation and so it is unknown if the ICE would occur for both formulations. Thus, there is uncertainty in the membership of the principal stratum , for example, those who would tolerate both formulations. Furthermore, it is difficult to estimate the effect in the principal stratum without additional assumptions and use of causal inference. Traditional per‐protocol analysis should not be seen as implementing this principal stratum strategy and ICH E9(R1) is critical of the biased nature of per‐protocol analyses. Using the terminology from the TransCelerate common protocol template, the analysis data set consists of two components: (1) the participant set : the participants relevant to the estimation; and (2) the data points set : indicates the data points to be used at the observation level in relation to the timing of intercurrent events and their handling strategies. In clinical pharmacology trials, the data points set should be considered first at the concentration‐level and secondly at the PK endpoint‐level since the endpoints such as AUC 0‐ꝏ and C max need to be derived from the selected concentrations, e.g., by non‐compartmental modelling, before the estimand can be estimated. The two analysis data set components (1) and (2) should be aligned to the estimand description and specified explicitly in the statistical considerations section of the trial protocol. Statistical methods (the estimator) should be selected to reliably estimate the estimand under plausible assumptions. 2.1.6 Step 6: Identify assumptions for the main analysis and suitable sensitivity analyses The underlying assumptions for the main analysis should be specified in the clinical trial protocol, and sensitivity analyses should be proposed to evaluate the robustness of the results towards these assumptions. Such sensitivity analyses should be aligned with the same estimand as the main statistical approach. 2.1.7 Step 7: Document the estimand Considerations for documenting estimands in clinical trial protocols are discussed in Lynggaard et al., and protocol templates incorporating estimands have been developed by TransCelerate CPT and in the ICH M11 draft guidance. The chosen estimand(s) should be documented and justified in the protocol so that stakeholders can review and confirm agreement early on. Leaving documentation of the estimands to the statistical analysis plan is not appropriate because they are closely related to the trial objectives, and they impact trial design, case report form design and trial conduct. The estimands should also be documented in subsequent reporting together with an overview of the frequency and timing of ICEs under the investigated treatments to facilitate interpretation of estimated results. It should be noted that any difference in frequency or pattern in the occurrence of relevant ICEs might be informative about differences between treatments and should be carefully investigated. This is of particular importance in cases where it cannot be determined with a high degree of confidence whether the ICEs are related or unrelated to treatment. In addition, a precise description of the targeted treatment effect is encouraged for press releases and product labels. What is the estimand framework and thinking process? The estimand framework, as outlined in the ICH E9(R1) addendum, distinguishes between what we want to estimate (the estimand) and how we estimate it (the estimator). Defining the estimand is not a statistical exercise but a multidisciplinary discussion that describes the trial objective in depth and can affect the trial design and conduct as well as the analysis and interpretation of the results. Unfortunately, the conduct of clinical trials rarely goes to plan; the estimand framework ensures that events occurring after treatment initiation that affect the interpretation or existence of the outcome are discussed and documented at the protocol stage, and that strategies are put in place to handle such ICEs. Figure presents an adaptation of the framework and the thinking process. Although this is depicted as a linear process, revisiting previous steps may be warranted if, for example, the planned ICE handling strategies (Step 3) are not feasible due to design considerations, difficulties in collecting the required data or lack of a robust estimator (Step 5). The estimand is the description of the treatment effect we want to estimate, and it should align with the trial objective and the underlying scientific question. It includes treatment conditions, target population, endpoint, population‐level summary and ICEs along with their associated handling strategies. Four of these attributes are typically described (at least partly) in different sections of the trial protocol, but ICEs and their associated handling strategies may be less familiar to the reader. As such, these will be discussed in Steps 2 and 3 below. The considerations described in the steps that follow are intended to be applicable to a variety of clinical pharmacology trials and we will focus on a bioequivalence case study later. 2.1.1 Step 1: Therapeutic setting and intent of treatment determining a trial objective The main questions of interest should be clearly identified. This requires a full understanding of the therapeutic setting and includes considerations about the disease or condition and the clinical context. Questions include: Will the drug be used for diagnosis, prophylaxis or treatment? Which patients would benefit (e.g., paediatric, elderly, renally impaired or requiring concomitant medications)? What is known or expected from the drug metabolism and tolerability? Does the trial aim to find the optimal dosing regimen of a new entity or to understand the relative bioavailability for the new formulation of an already approved product? Does it investigate interactions with other medicines, effects of administration with food or use in special patient populations? For entities not studied before, questions may also concern tolerability and PK: What are the preferred route and dosing regimen? What are the overall exposure and the C max ? How is the drug eliminated and how long does that take? For new formulations, are they designed to improve convenience (for instance, a single high‐dose tablet), tolerability or stability/shelf‐life, or can they be considered a generic? Are they expected to have a slower absorption (for instance, to attain therapeutic concentrations for longer than the comparator formulation)? Discussions around these points help establish the trial objective. 2.1.2 Step 2: Identify intercurrent events ICH E9(R1) defines ICEs as “events occurring after treatment initiation that affect either the interpretation or the existence of the measurements associated with the clinical question of interest”. Identifying ICEs requires discussions between at least clinical pharmacologists, clinicians and statisticians. In clinical pharmacology, ICEs affecting absorption, distribution, metabolism and excretion affect the interpretation of an outcome, whereas terminal events such as death, transplant or amputation affect the outcome's existence. However, the terminal events are unlikely to be relevant in clinical pharmacology unless patients are severely ill or are dosed over an extended period. Note, ICEs should be discussed both in relation to the profile measurements and the endpoint. It may be that a few profile measurements are affected, but still the endpoint can be derived reliably. In other cases, it may also be that the endpoint will be impacted, and the event is therefore to be defined as an ICE. Absorption may be affected by the route of administration. For example, individuals not experienced with inhalers may use them incorrectly, patches may come off, or products intended to be administered subcutaneously may erroneously be given intramuscularly or intravenously. Deviations from dosing instructions may also affect absorption, for example, if the product is taken in the fed state instead of the fasted state, or with a hot drink instead of water, where the coating of a capsule may dissolve faster than intended. Tolerability issues like vomiting and diarrhoea will also affect absorption of an oral drug. By contrast, other ICEs, such as interacting drugs that induce or inhibit metabolism, may affect elimination. To decide if absorption or clearance are affected by ICEs, the timing of the latter needs to be considered. For example, if interacting substances are used well after the C max is expected to be attained, the event is not considered an ICE for C max . Once undesirable ICEs are identified, they can be minimized with good trial conduct and, potentially, training (e.g., on how to use inhalers). This is feasible in the controlled setting typical of clinical pharmacology trials. Table includes some examples of ICEs in clinical pharmacology trials to highlight how they differ from protocol deviations, whilst also acknowledging some overlap. ICEs should not be confused with issues related to missing data, such as missing blood samples or laboratory failings. Missing data issues are not considered ICEs because the underlying level of exposure in the body is unaffected. ICH E9(R1) defines missing data as “data that would be meaningful for the analysis of a given estimand but were not collected” and they should be minimized during trial design and conduct. 2.1.3 Step 3: Discuss strategies to address intercurrent events Table describes five potential strategies to handle ICEs per ICH E9(R1) and how they can be applied in clinical pharmacology. A strategy needs to be selected for each type of ICE, but strategies can vary across ICEs within the same estimand. The choice of strategy should be driven by the intent of the treatment and the trial objectives rather than the trial design and statistical methods. In clinical practice, a generic and originator product will be considered to be interchangeable and thus the focus for demonstrating bioequivalence may be the principal stratum of individuals able to tolerate both products together with demonstrating that the tolerability does not differ between the treatments. In the case of a biosimilar or generic, both would generally be expected to be equally well tolerated, whereas a new formulation or different route of administration may be designed to be better tolerated and the principal stratum of those able to tolerate either formulation may also be of interest together with a demonstration of better tolerability. Note, it may not always be possible to identify the principal stratum of interest. A different strategy, specifically a hypothetical strategy, could be proposed to handle the use of interacting substances to remove their impact on the measurements and to ensure sensitivity to detect any difference between formulations. If a new formulation was designed to improve tolerability, the objective might be to demonstrate not only bioequivalence with the reference formulation in the principal stratum of tolerators but also more consistent exposure in all patients after multiple dosing measured irrespective of any use of other medications, dose interruptions or compliance issues. This interest in a treatment effect that includes the effects of the ICEs is referred to as the treatment policy strategy. It tends to be a common choice in later phase trials, in line with the intent‐to‐treat principle, which promotes the complete follow‐up of participants and use of their data in analysis regardless of whether the ICEs occur. Notably, a higher steady‐state exposure results from different PK properties, such as increased absorption, lower volume of distribution or decreased clearance. A higher exposure may also be associated with poor tolerability, for instance, severe adverse reactions. Therefore, PK could differ between individuals that tolerate a drug well and those who do not. The while on treatment strategy focuses interest on the response prior to the occurrence of ICEs but without implicit or explicit imputation beyond that point. This strategy requires careful consideration of the choice of endpoint. For example, we might consider the proportion of time inhibition of a biomarker is above a threshold or plasma concentrations are within a therapeutic window, from the first dose of a trial drug to the time of treatment discontinuation. The final strategy to consider is the composite variable strategy where certain undesirable ICEs such as tolerability issues would be considered a failure in the endpoint. Although composite binary responder endpoints are being used to assess treatment efficacy in confirmatory trials, we are not aware of much use of composite binary or continuous variables for PK. In the above example evaluating trough exposure in patients after multiple dosing, it might be an option to define a composite trough endpoint as zero (or any other unfavourable value) for patients who discontinue treatment due to a tolerability issue, if that is addressing a relevant question. 2.1.4 Step 4: Precisely define the estimand(s) The trial team should define the treatment effects (estimands) according to the treatment conditions, endpoints, population‐level summary, target population and strategies for ICEs. Depending on the ICEs and the strategy chosen, the precise specification of the treatment conditions, the endpoint or the target population may be impacted. For example, ICEs handled by the principal stratum strategy will be part of the target population, such as “biosimilarity in healthy individuals who do not develop anti‐drug antibodies”. See Table for some other examples. In a PK context, the treatment conditions would often be the two interventions to be compared, that is, the test and reference formulations and would include the route of administration, titration scheme and whether single or multiple dosing is of interest. In bioequivalence trials, PK concentration measurements are used to derive their co‐primary endpoints, typically the C max and the AUC, measured in a dosing interval for multiple dosing or extrapolated to infinity (AUC 0‐ꝏ ) for single dosing. In these trials, the population‐level summary needs to aggregate the data in a way that allows comparison of treatment conditions and to demonstrate equivalence within a specified bioequivalence limit. This is often done using the geometric mean ratio between the test and reference formulations. However, the estimand attributes are interrelated; for example, endpoints that can have zero values (see the composite continuous endpoint described in Table ) will affect the choice of appropriate population‐level summary. In this case, a median difference would be more suitable than a geometric mean ratio. In general, the population‐level summary should be chosen under careful evaluation of the context. Although patients are the ultimate target population, in alignment with guideline recommendations, , , , many clinical pharmacology trials and most bioequivalence trials are conducted in healthy participants. For instance, EMA guidance states: the subject population for bioequivalence studies should be selected with the aim of permitting detection of differences between pharmaceutical products. In order to reduce variability not related to differences between products, the studies should normally be performed in healthy volunteers unless the drug carries safety concerns that make this unethical. This model, in vivo healthy volunteers, is regarded as adequate in most instances to detect formulation differences and to allow extrapolation of the results to populations for which the reference medicinal product is approved (the elderly, children, patients with renal or liver impairment, etc.). In these cases, we recommend that the population attribute of the estimand should be specified as “healthy individuals” or a suitable subset such as “healthy adult males”, but the protocol should explicitly discuss whether it is reasonable to conclude that the two drugs are clinically equivalent in the ultimate patient population if bioequivalence in healthy participants is demonstrated (Figure ). In contrast, if the clinical pharmacology trial is conducted in patients, the target population should be the patients targeted by the scientific question of interest posed by the trial. We would like to emphasize that the terms “participant”, “subject” and “volunteer” are not appropriate when describing the target population of the estimand, as they infer participation in the trial rather than who we seek to study. The entire “population of healthy individuals” will not participate in the trial, so this terminology is distinct from the terminology of “participant”, “subject” or “volunteer”, which are reserved for describing the people enrolled in the trial and the participant sets (Step 5). Instead, the description of the target population should include “patients” or “healthy individuals”. Consequently, referring to “participants meeting eligibility criteria” and “all randomised participants” or other participant sets will not be appropriate for describing the target population. 2.1.5 Step 5: Align choices on trial design, data collection and method of estimation Considerations about trial design, trial conduct, data to be collected, and estimation methods should be aligned with the choice of estimand(s). A crossover design is used in many clinical pharmacology trials, especially if the time required for washout is not prohibitive. Crossovers provide an efficient design in terms of the required sample size and may also facilitate the identification of a principal stratum of interest, if applicable. Such a trial will be most sensitive if certain ICEs can be minimized, for example, if the intake of food and other substances can be controlled. In clinical pharmacology trials, mitigation plans can be developed to minimize the impact of certain ICEs and thereby improve their sensitivity. By contrast, Phase 3 trials may seek to mirror the occurrence of ICEs in real‐world clinical practice, especially when the aim is to apply the treatment policy strategy to all ICEs; in this context, mitigation plans may not be advisable. Consideration should be given to whether the data after an ICE occurs would be relevant in the current and subsequent periods, which would be dependent on the strategy used for the ICE. This may guide decisions during trial conduct such as stopping blood sampling for PK after one participant vomits, but continuing data collection for safety analyses, and then whether the participant should or should not continue into the subsequent dosing period. In some trial designs, such as parallel group trials, participants only receive one formulation and so it is unknown if the ICE would occur for both formulations. Thus, there is uncertainty in the membership of the principal stratum , for example, those who would tolerate both formulations. Furthermore, it is difficult to estimate the effect in the principal stratum without additional assumptions and use of causal inference. Traditional per‐protocol analysis should not be seen as implementing this principal stratum strategy and ICH E9(R1) is critical of the biased nature of per‐protocol analyses. Using the terminology from the TransCelerate common protocol template, the analysis data set consists of two components: (1) the participant set : the participants relevant to the estimation; and (2) the data points set : indicates the data points to be used at the observation level in relation to the timing of intercurrent events and their handling strategies. In clinical pharmacology trials, the data points set should be considered first at the concentration‐level and secondly at the PK endpoint‐level since the endpoints such as AUC 0‐ꝏ and C max need to be derived from the selected concentrations, e.g., by non‐compartmental modelling, before the estimand can be estimated. The two analysis data set components (1) and (2) should be aligned to the estimand description and specified explicitly in the statistical considerations section of the trial protocol. Statistical methods (the estimator) should be selected to reliably estimate the estimand under plausible assumptions. 2.1.6 Step 6: Identify assumptions for the main analysis and suitable sensitivity analyses The underlying assumptions for the main analysis should be specified in the clinical trial protocol, and sensitivity analyses should be proposed to evaluate the robustness of the results towards these assumptions. Such sensitivity analyses should be aligned with the same estimand as the main statistical approach. 2.1.7 Step 7: Document the estimand Considerations for documenting estimands in clinical trial protocols are discussed in Lynggaard et al., and protocol templates incorporating estimands have been developed by TransCelerate CPT and in the ICH M11 draft guidance. The chosen estimand(s) should be documented and justified in the protocol so that stakeholders can review and confirm agreement early on. Leaving documentation of the estimands to the statistical analysis plan is not appropriate because they are closely related to the trial objectives, and they impact trial design, case report form design and trial conduct. The estimands should also be documented in subsequent reporting together with an overview of the frequency and timing of ICEs under the investigated treatments to facilitate interpretation of estimated results. It should be noted that any difference in frequency or pattern in the occurrence of relevant ICEs might be informative about differences between treatments and should be carefully investigated. This is of particular importance in cases where it cannot be determined with a high degree of confidence whether the ICEs are related or unrelated to treatment. In addition, a precise description of the targeted treatment effect is encouraged for press releases and product labels. Step 1: Therapeutic setting and intent of treatment determining a trial objective The main questions of interest should be clearly identified. This requires a full understanding of the therapeutic setting and includes considerations about the disease or condition and the clinical context. Questions include: Will the drug be used for diagnosis, prophylaxis or treatment? Which patients would benefit (e.g., paediatric, elderly, renally impaired or requiring concomitant medications)? What is known or expected from the drug metabolism and tolerability? Does the trial aim to find the optimal dosing regimen of a new entity or to understand the relative bioavailability for the new formulation of an already approved product? Does it investigate interactions with other medicines, effects of administration with food or use in special patient populations? For entities not studied before, questions may also concern tolerability and PK: What are the preferred route and dosing regimen? What are the overall exposure and the C max ? How is the drug eliminated and how long does that take? For new formulations, are they designed to improve convenience (for instance, a single high‐dose tablet), tolerability or stability/shelf‐life, or can they be considered a generic? Are they expected to have a slower absorption (for instance, to attain therapeutic concentrations for longer than the comparator formulation)? Discussions around these points help establish the trial objective. Step 2: Identify intercurrent events ICH E9(R1) defines ICEs as “events occurring after treatment initiation that affect either the interpretation or the existence of the measurements associated with the clinical question of interest”. Identifying ICEs requires discussions between at least clinical pharmacologists, clinicians and statisticians. In clinical pharmacology, ICEs affecting absorption, distribution, metabolism and excretion affect the interpretation of an outcome, whereas terminal events such as death, transplant or amputation affect the outcome's existence. However, the terminal events are unlikely to be relevant in clinical pharmacology unless patients are severely ill or are dosed over an extended period. Note, ICEs should be discussed both in relation to the profile measurements and the endpoint. It may be that a few profile measurements are affected, but still the endpoint can be derived reliably. In other cases, it may also be that the endpoint will be impacted, and the event is therefore to be defined as an ICE. Absorption may be affected by the route of administration. For example, individuals not experienced with inhalers may use them incorrectly, patches may come off, or products intended to be administered subcutaneously may erroneously be given intramuscularly or intravenously. Deviations from dosing instructions may also affect absorption, for example, if the product is taken in the fed state instead of the fasted state, or with a hot drink instead of water, where the coating of a capsule may dissolve faster than intended. Tolerability issues like vomiting and diarrhoea will also affect absorption of an oral drug. By contrast, other ICEs, such as interacting drugs that induce or inhibit metabolism, may affect elimination. To decide if absorption or clearance are affected by ICEs, the timing of the latter needs to be considered. For example, if interacting substances are used well after the C max is expected to be attained, the event is not considered an ICE for C max . Once undesirable ICEs are identified, they can be minimized with good trial conduct and, potentially, training (e.g., on how to use inhalers). This is feasible in the controlled setting typical of clinical pharmacology trials. Table includes some examples of ICEs in clinical pharmacology trials to highlight how they differ from protocol deviations, whilst also acknowledging some overlap. ICEs should not be confused with issues related to missing data, such as missing blood samples or laboratory failings. Missing data issues are not considered ICEs because the underlying level of exposure in the body is unaffected. ICH E9(R1) defines missing data as “data that would be meaningful for the analysis of a given estimand but were not collected” and they should be minimized during trial design and conduct. Step 3: Discuss strategies to address intercurrent events Table describes five potential strategies to handle ICEs per ICH E9(R1) and how they can be applied in clinical pharmacology. A strategy needs to be selected for each type of ICE, but strategies can vary across ICEs within the same estimand. The choice of strategy should be driven by the intent of the treatment and the trial objectives rather than the trial design and statistical methods. In clinical practice, a generic and originator product will be considered to be interchangeable and thus the focus for demonstrating bioequivalence may be the principal stratum of individuals able to tolerate both products together with demonstrating that the tolerability does not differ between the treatments. In the case of a biosimilar or generic, both would generally be expected to be equally well tolerated, whereas a new formulation or different route of administration may be designed to be better tolerated and the principal stratum of those able to tolerate either formulation may also be of interest together with a demonstration of better tolerability. Note, it may not always be possible to identify the principal stratum of interest. A different strategy, specifically a hypothetical strategy, could be proposed to handle the use of interacting substances to remove their impact on the measurements and to ensure sensitivity to detect any difference between formulations. If a new formulation was designed to improve tolerability, the objective might be to demonstrate not only bioequivalence with the reference formulation in the principal stratum of tolerators but also more consistent exposure in all patients after multiple dosing measured irrespective of any use of other medications, dose interruptions or compliance issues. This interest in a treatment effect that includes the effects of the ICEs is referred to as the treatment policy strategy. It tends to be a common choice in later phase trials, in line with the intent‐to‐treat principle, which promotes the complete follow‐up of participants and use of their data in analysis regardless of whether the ICEs occur. Notably, a higher steady‐state exposure results from different PK properties, such as increased absorption, lower volume of distribution or decreased clearance. A higher exposure may also be associated with poor tolerability, for instance, severe adverse reactions. Therefore, PK could differ between individuals that tolerate a drug well and those who do not. The while on treatment strategy focuses interest on the response prior to the occurrence of ICEs but without implicit or explicit imputation beyond that point. This strategy requires careful consideration of the choice of endpoint. For example, we might consider the proportion of time inhibition of a biomarker is above a threshold or plasma concentrations are within a therapeutic window, from the first dose of a trial drug to the time of treatment discontinuation. The final strategy to consider is the composite variable strategy where certain undesirable ICEs such as tolerability issues would be considered a failure in the endpoint. Although composite binary responder endpoints are being used to assess treatment efficacy in confirmatory trials, we are not aware of much use of composite binary or continuous variables for PK. In the above example evaluating trough exposure in patients after multiple dosing, it might be an option to define a composite trough endpoint as zero (or any other unfavourable value) for patients who discontinue treatment due to a tolerability issue, if that is addressing a relevant question. Step 4: Precisely define the estimand(s) The trial team should define the treatment effects (estimands) according to the treatment conditions, endpoints, population‐level summary, target population and strategies for ICEs. Depending on the ICEs and the strategy chosen, the precise specification of the treatment conditions, the endpoint or the target population may be impacted. For example, ICEs handled by the principal stratum strategy will be part of the target population, such as “biosimilarity in healthy individuals who do not develop anti‐drug antibodies”. See Table for some other examples. In a PK context, the treatment conditions would often be the two interventions to be compared, that is, the test and reference formulations and would include the route of administration, titration scheme and whether single or multiple dosing is of interest. In bioequivalence trials, PK concentration measurements are used to derive their co‐primary endpoints, typically the C max and the AUC, measured in a dosing interval for multiple dosing or extrapolated to infinity (AUC 0‐ꝏ ) for single dosing. In these trials, the population‐level summary needs to aggregate the data in a way that allows comparison of treatment conditions and to demonstrate equivalence within a specified bioequivalence limit. This is often done using the geometric mean ratio between the test and reference formulations. However, the estimand attributes are interrelated; for example, endpoints that can have zero values (see the composite continuous endpoint described in Table ) will affect the choice of appropriate population‐level summary. In this case, a median difference would be more suitable than a geometric mean ratio. In general, the population‐level summary should be chosen under careful evaluation of the context. Although patients are the ultimate target population, in alignment with guideline recommendations, , , , many clinical pharmacology trials and most bioequivalence trials are conducted in healthy participants. For instance, EMA guidance states: the subject population for bioequivalence studies should be selected with the aim of permitting detection of differences between pharmaceutical products. In order to reduce variability not related to differences between products, the studies should normally be performed in healthy volunteers unless the drug carries safety concerns that make this unethical. This model, in vivo healthy volunteers, is regarded as adequate in most instances to detect formulation differences and to allow extrapolation of the results to populations for which the reference medicinal product is approved (the elderly, children, patients with renal or liver impairment, etc.). In these cases, we recommend that the population attribute of the estimand should be specified as “healthy individuals” or a suitable subset such as “healthy adult males”, but the protocol should explicitly discuss whether it is reasonable to conclude that the two drugs are clinically equivalent in the ultimate patient population if bioequivalence in healthy participants is demonstrated (Figure ). In contrast, if the clinical pharmacology trial is conducted in patients, the target population should be the patients targeted by the scientific question of interest posed by the trial. We would like to emphasize that the terms “participant”, “subject” and “volunteer” are not appropriate when describing the target population of the estimand, as they infer participation in the trial rather than who we seek to study. The entire “population of healthy individuals” will not participate in the trial, so this terminology is distinct from the terminology of “participant”, “subject” or “volunteer”, which are reserved for describing the people enrolled in the trial and the participant sets (Step 5). Instead, the description of the target population should include “patients” or “healthy individuals”. Consequently, referring to “participants meeting eligibility criteria” and “all randomised participants” or other participant sets will not be appropriate for describing the target population. Step 5: Align choices on trial design, data collection and method of estimation Considerations about trial design, trial conduct, data to be collected, and estimation methods should be aligned with the choice of estimand(s). A crossover design is used in many clinical pharmacology trials, especially if the time required for washout is not prohibitive. Crossovers provide an efficient design in terms of the required sample size and may also facilitate the identification of a principal stratum of interest, if applicable. Such a trial will be most sensitive if certain ICEs can be minimized, for example, if the intake of food and other substances can be controlled. In clinical pharmacology trials, mitigation plans can be developed to minimize the impact of certain ICEs and thereby improve their sensitivity. By contrast, Phase 3 trials may seek to mirror the occurrence of ICEs in real‐world clinical practice, especially when the aim is to apply the treatment policy strategy to all ICEs; in this context, mitigation plans may not be advisable. Consideration should be given to whether the data after an ICE occurs would be relevant in the current and subsequent periods, which would be dependent on the strategy used for the ICE. This may guide decisions during trial conduct such as stopping blood sampling for PK after one participant vomits, but continuing data collection for safety analyses, and then whether the participant should or should not continue into the subsequent dosing period. In some trial designs, such as parallel group trials, participants only receive one formulation and so it is unknown if the ICE would occur for both formulations. Thus, there is uncertainty in the membership of the principal stratum , for example, those who would tolerate both formulations. Furthermore, it is difficult to estimate the effect in the principal stratum without additional assumptions and use of causal inference. Traditional per‐protocol analysis should not be seen as implementing this principal stratum strategy and ICH E9(R1) is critical of the biased nature of per‐protocol analyses. Using the terminology from the TransCelerate common protocol template, the analysis data set consists of two components: (1) the participant set : the participants relevant to the estimation; and (2) the data points set : indicates the data points to be used at the observation level in relation to the timing of intercurrent events and their handling strategies. In clinical pharmacology trials, the data points set should be considered first at the concentration‐level and secondly at the PK endpoint‐level since the endpoints such as AUC 0‐ꝏ and C max need to be derived from the selected concentrations, e.g., by non‐compartmental modelling, before the estimand can be estimated. The two analysis data set components (1) and (2) should be aligned to the estimand description and specified explicitly in the statistical considerations section of the trial protocol. Statistical methods (the estimator) should be selected to reliably estimate the estimand under plausible assumptions. Step 6: Identify assumptions for the main analysis and suitable sensitivity analyses The underlying assumptions for the main analysis should be specified in the clinical trial protocol, and sensitivity analyses should be proposed to evaluate the robustness of the results towards these assumptions. Such sensitivity analyses should be aligned with the same estimand as the main statistical approach. Step 7: Document the estimand Considerations for documenting estimands in clinical trial protocols are discussed in Lynggaard et al., and protocol templates incorporating estimands have been developed by TransCelerate CPT and in the ICH M11 draft guidance. The chosen estimand(s) should be documented and justified in the protocol so that stakeholders can review and confirm agreement early on. Leaving documentation of the estimands to the statistical analysis plan is not appropriate because they are closely related to the trial objectives, and they impact trial design, case report form design and trial conduct. The estimands should also be documented in subsequent reporting together with an overview of the frequency and timing of ICEs under the investigated treatments to facilitate interpretation of estimated results. It should be noted that any difference in frequency or pattern in the occurrence of relevant ICEs might be informative about differences between treatments and should be carefully investigated. This is of particular importance in cases where it cannot be determined with a high degree of confidence whether the ICEs are related or unrelated to treatment. In addition, a precise description of the targeted treatment effect is encouraged for press releases and product labels. APPLYING THE ESTIMAND THINKING PROCESS TO A BIOEQUIVALENCE CASE STUDY In our case study, we will discuss how the estimand framework and steps of the thinking process might be applied in a specific bioequivalence setting where there are tolerability issues to consider. There are many other settings where the thinking process could add value, but we have selected a bioequivalence setting since that is familiar to most clinical pharmacologists. Pan et al. evaluated the bioequivalence of a new high‐dose formulation of pirfenidone in the fed and fasted states using a 4 × 4 crossover design. Although Pan et al. did not use the estimand framework and thinking process, we will retrospectively implement them in this case study. We only had access to the information published in Pan et al. because none of the authors of this paper were involved in the original trial. Therefore, we may have overlooked important considerations that influenced some of the choices made in the design of the trial. The choices made in defining the estimand for this case study should not be seen as guidance for other pharmacology trials but rather as a way of illustrating the thinking process. Accordingly, the estimand presented is not necessarily the only relevant one that could be defined. 3.1 Case study step 1: Therapeutic setting and intent of treatment determining a trial objective Pirfenidone is an oral antifibrotic agent used to treat the life‐threatening condition of idiopathic pulmonary fibrosis. In clinical practice, the dose of pirfenidone is titrated to up to three 267 mg capsules three times daily with each meal, but only about 60% of patients tolerate the highest (801 mg) dose level. The drug label advises dose administration with food to reduce the risk of gastrointestinal side effects. A new formulation of a high‐dose (801 mg) tablet has been developed for improved convenience to patients rather than with any anticipated improvement in tolerability. The new formulation is expected to have a similar PK profile as three 267 mg capsules. We propose that the objective of the trial is to establish bioequivalence in the rate and extent of absorption of pirfenidone in healthy adults after a single administration with food of a new film‐coated 801 mg tablet compared to the reference formulation of three 267 mg capsules, as summarized by the co‐primary endpoints, C max and AUC 0‐ꝏ . Pan et al. evaluated the bioequivalence in both fasted and fed states in the original trial and the effect of food on the C max and AUC 0‐ꝏ of pirfenidone after administering the new formulation. However, dosing without food leads to a higher C max and exacerbates skin‐related adverse events, nausea and vomiting, and opposes the advice of the drug label. Therefore, here, we focus on the objective of demonstrating bioequivalence in the fed state only and we ascertain there is good reason to not study administration of pirfenidone in the fasted state, which carries an increased risk of vomiting. The rationale for this needs to be provided in the protocol, and advice may be needed if the regulatory acceptability of the approach is uncertain. In this context we note that current FDA guidance recommends two separate studies in healthy adults to demonstrate bioequivalence for pirfenidone (one fasting and one fed study) , whereas current EMA guidance accepts only one trial in healthy adults either fasting or fed: “The SmPC recommends intake in fed state to minimise the risk of risk of nausea and dizziness. A fed study is, therefore, acceptable. However, a fasted study is also acceptable.” Finally, we will assume transportability of results so that the bioequivalence in healthy adults would be the same as in patients. 3.2 Case study step 2: Identify intercurrent events While additional ICEs may be relevant, we focus on the ICEs listed in Table , which for brevity, we will reference as ICE1a, ICE1b, ICE2a, ICE2b, ICE3 and ICE4 throughout the subsequent steps of this case study. The wording of ICE2 (“Does not take both formulations”) has been refined after iterating through the thinking process as discussed in the section on general considerations. It reflects the situation where we want to evaluate bioequivalence in those who can tolerate both formulations, and we know that a useful design that is aligned to such an estimand is a crossover design. In such a design, a dose may not be received in the second period either because of (a) a tolerability issue in the previous period or (b) due to other reasons (for example, logistical issues in attending the subsequent period). Importantly, using different estimand strategies according to whether ICEs are related (ICE1a and ICE2a) or unrelated (ICE1b and ICE2b) to treatment requires that they can be sufficiently distinguished. This needs to be considered before the trial is started and may affect the case report form design and trial conduct. If ICE1 and ICE2 cannot be reliably sub‐categorized, then use of different estimand strategies to handle (a) related and (b) unrelated reasons would be precluded (Step 3 below). As highlighted in the general Step 7, any difference in frequency or pattern in the occurrence of relevant ICEs might be informative about differences in the investigated treatments and should be carefully investigated. Specifically, this comparison can be used to investigate the assumption that the tolerability does not differ between formulations. 3.3 Case study step 3: Discuss strategies to address intercurrent events The selected handling strategy for each ICE defined in Step 2 is provided in Table . The rationale for selecting these strategies is described below. In clinical practice, patients with idiopathic pulmonary fibrosis who can tolerate an 801 mg total dose of pirfenidone with food are expected to use either three 267 mg capsules or one 801 mg tablet interchangeably. Those taking capsules are likely to transition to the tablet for convenience when this becomes their stable dose. We focus on establishing bioequivalence in the principal stratum of healthy adults able to tolerate both formulations, which corresponds to considering the subset of healthy adults for whom ICE1a and ICE2a would not occur. As this is a bioequivalence trial comparing different formulations, sensitivity to changes in absorption is paramount. In line with good clinical practice requirements, PK trials of healthy participants are conducted in a way that minimizes dosing errors and the need for interacting drugs or substances. Issues might emerge, such as viral or bacterial gastroenteritis (not treatment‐related nausea) or a requirement for a prohibited medicine to treat a new condition, such as taking a painkiller that contains caffeine to treat a severe headache from caffeine withdrawal. These events introduce noise in the data collected afterwards and, depending on their frequency and pattern, may jeopardize interpretation of the trial data and highlight a quality issue concerning trial conduct. In any case, the trial protocol should prespecify how to handle such events, although data collected before them may still be relevant. It might not be sensible to define a principal stratum of people who would never take caffeine or any other interacting substances, have gastroenteritis or be dosed in error. Instead, we wish to estimate the treatment effect purely due to the difference in formulations in the absence of these issues, which corresponds to the approach described by ICH E9(R1) as a hypothetical strategy. In other words, we are interested in the PK profile that would have been observed if the individual had not been unwell with viral or bacterial gastroenteritis (ICE1b), as though they had taken the second formulation rather than discontinuing due to reasons unrelated to treatment (ICE2b), without a dosing error (ICE3) or taking any interacting drugs or substances that might affect absorption or clearance (ICE4). Note, rather than focus on a principal stratum of those able to tolerate both formulations, we may have chosen to apply a hypothetical strategy to assess bioequivalence in healthy adults as though they would not have vomiting issues or discontinuation from subsequent administration for any reason. See Step 3 of the general description of the thinking process and Table for more details. 3.4 Case study step 4: Precisely define the estimands Idiopathic pulmonary fibrosis is strongly associated with pre‐existing gastroesophageal reflux disease, where stomach acid flows back into the oesophagus, potentially delaying drug absorption and introducing variability in PK profiles. Such patients may be receiving multiple drugs, further altering metabolism of the formulations under investigation. For these patients, multiple blood draws for PK sampling may be an unnecessary burden. Although these issues may alter PK for both formulations and increase variability, it is reasonable to believe that the estimated ratio of exposure can be generalized from healthy trial participants to healthy adults and finally transported to patients (see Figure ). Therefore, to meet the trial objective, we recommend specifying “healthy adults” rather than “patients with idiopathic pulmonary fibrosis” as the target population attribute of the estimand. The transportability of bioequivalence from healthy individuals to the target patient population should be justified in the trial protocol. All five attributes of our co‐primary estimands with the endpoints AUC 0‐ꝏ and C max are detailed in Table . In a sentence, estimand 1a is described as: The geometric mean ratio (test/reference) of pirfenidone AUC 0‐ꝏ after a single oral dose, comparing the test formulation of one 801 mg film‐coated tablet with the reference of 3 × 267 mg capsules in healthy adults able to tolerate both formulations administered with food as though dosed correctly with a standard meal, without intake of interacting drugs or substances and without vomiting or diarrhoea due to gastroenteritis (unrelated to tolerability). For C max , we propose a similar estimand 1b where AUC 0‐ꝏ is replaced by C max . Further details on the individual attributes are described in Table . 3.5 Case study step 5: Align choices on trial design, data collection and method of estimation ICH E9(R1) advocates early discussions on the choice of estimand before making design choices (Step 5) rather than vice versa. Although the trial described by Pan et al. investigated bioequivalence under fed and fasted states in a 4 × 4 crossover design, here we focus only on the fed state. Due to the therapeutic setting and the intent of treatment (Step 1), we propose a 2 × 2 crossover design with a sufficient washout in healthy individuals under controlled conditions; this design is selected for efficiency and to facilitate identification of the principal stratum of those who can tolerate both formulations. Identifying ICEs (Step 2) may help adapt how the trial is conducted to minimize the extent of ICEs unrelated to tolerability. For instance, the trial can be adapted to anticipate potential interacting substances. In this crossover trial, participants who do not tolerate pirfenidone in Period 1 would not be exposed in Period 2, which would allow them to withdraw from the trial after sufficient safety follow‐up. It would not be necessary to collect blood samples within a PK profile after the occurrence of ICE1b, ICE2b, ICE3 and ICE4 events in the affected period. The first step in the analysis is to derive the PK endpoints AUC 0‐ꝏ and C max , typically by use of non‐compartmental methods on the relevant data dependent on the ICE handling strategies. Figure illustrates the interplay between the ICE handling strategies and the definition of participant and data points sets. The participant set would include Participants 2–5 as being those in the principal stratum who did not have any tolerability issues after administration of either formulation. Participants 1 and 6 would be completely excluded from the participant set due to treatment‐related vomiting and intolerable rash, respectively. Participants who experienced dosing deviations, did not take both formulations for treatment‐unrelated reasons, used interacting drugs/substances or experienced treatment‐unrelated vomiting or diarrhoea would be included in the participant set of interest, and PK endpoints from the period not affected by ICEs would be used. The concentration‐level data points set would include the relevant concentration data for those in the participant set. Data in the affected period after the occurrence of ICE1b, ICE2b, ICE3 and ICE4 (Table ) would be considered irrelevant given the selected hypothetical strategy. This may result in some PK endpoints being non‐evaluable and set as missing. If an ICE clearly occurred after the expected T max , an observed C max could be used. In addition, AUC 0‐ꝏ could be extrapolated using the concentration values prior to the ICE, as if the ICE had not occurred. However, clear rules for handling such cases would need to be predefined. In Figure , the following ICEs affect the selection of relevant concentration data and subsequent PK endpoint set: ○ Participant 2 does not take all three tablets in Period 1 but complies in Period 2, so only data from Period 2 are included. ○ Participant 3 presents a dosing deviation in Period 2, so only data from Period 1 are included. ○ Participant 4 uses interacting substances in Period 2 after C max occurred, so data from Period 1 and parts of Period 2 can be used. In this case, the C max derived from the PK measurements would be used in the analysis. It may also be possible to extrapolate AUC 0‐ꝏ from the last quantifiable concentration before interacting substances were used in Period 2. ○ Participant 5 has no ICEs so all data from both periods can be used. Admittedly, this is an extreme example to illustrate the various impacts of ICEs, as it results in having to exclude some data from five of the six participants. In a well‐conducted bioequivalence trial of a well‐tolerated drug, most participants would not experience ICEs and, thus, all of their data could be used. A high number of participants with partly excluded data points would raise concerns about the validity of the trial or integrity of the data and would require a review of trial conduct and mitigation plans in line with good clinical practice. The second step is to estimate the two co‐primary estimands of interest (Table ). Different statistical approaches could be used. For this case study, we use a linear mixed model on log‐transformed PK endpoints AUC 0‐ꝏ and C max with sequence, treatment and period as fixed effects and participant as random effect. The linear mixed model would implicitly predict the endpoint values that could not be derived due to an ICE in the hypothetical scenario making use of PK endpoints from the period not affected by ICEs where available. Truly missing data, e.g., laboratory failure, would be predicted in the same way. Finally, if the 90% confidence interval for the estimated ratio of geometric means (test/reference) lies within standard bioequivalence limits of 0.80–1.25 (i.e., 80%–125%), , , we would declare bioequivalence. 3.6 Case study step 6: Identify assumptions for the main analysis and suitable sensitivity analyses The aim of the main analysis is to demonstrate bioequivalence in participants who can tolerate both the test and reference formulations. The 2 × 2 crossover design allows us to determine if a participant can tolerate both formulations. In this context, and assuming both formulations have equivalent and minimal carryover in the measured variable between periods, the effect in the principal stratum can be estimated without bias by limiting the analysis to participants who do not experience drug‐related tolerability issues. If a participant can tolerate the formulation in Period 1 and is not exposed in Period 2, we could assume that they would also tolerate the formulation given in Period 2. However, this may be incorrect. In this case, a sensitivity analysis could be performed in which the participant is completely excluded from the participant set and so none of their data would contribute. We assume that missing data, dosing deviations, use of interacting drugs or substances or treatment‐unrelated vomiting or diarrhoea occur at random with both formulations. More specifically, we assume that the probability of these ICEs depends only on the variables included in the mixed model (sequence, treatment, period and participant). When these assumptions are appropriate, a mixed‐model approach would allow for an unbiased assessment by including all available measurements. This approach implicitly models the value the endpoint would have taken in the hypothetical scenario, as defined by introducing a random effect of participant into the model, in other words, by combining between‐ and within‐participant information. In this approach, the robustness of the results against violations to the assumption that missing data and ICEs unrelated to tolerability issues occur at random (conditionally on the variables included in the model) will need to be explored carefully. For example, a multiple imputation approach with penalties to explore the robustness of results could be used. In this case, the imputed PK values are multiplied or divided by a series of penalties of increasing magnitude to inflate or deflate exposure levels for each group separately. For each pair of penalties, the analysis is repeated to investigate if the bioequivalence still holds. The penalties may be applied to the predicted missing data only, or to both predicted missing data and predicted outcomes for participants with intercurrent events handled by a hypothetical strategy. 3.7 Case study step 7: Document the estimand FDA guidance recommends documenting the estimand in the protocol for bioequivalence trials. Although the original trial did not consider estimands, in future trials, the information in Table could be presented alongside the trial objective. This should be supplemented with the rationale for the defined estimand. Case study step 1: Therapeutic setting and intent of treatment determining a trial objective Pirfenidone is an oral antifibrotic agent used to treat the life‐threatening condition of idiopathic pulmonary fibrosis. In clinical practice, the dose of pirfenidone is titrated to up to three 267 mg capsules three times daily with each meal, but only about 60% of patients tolerate the highest (801 mg) dose level. The drug label advises dose administration with food to reduce the risk of gastrointestinal side effects. A new formulation of a high‐dose (801 mg) tablet has been developed for improved convenience to patients rather than with any anticipated improvement in tolerability. The new formulation is expected to have a similar PK profile as three 267 mg capsules. We propose that the objective of the trial is to establish bioequivalence in the rate and extent of absorption of pirfenidone in healthy adults after a single administration with food of a new film‐coated 801 mg tablet compared to the reference formulation of three 267 mg capsules, as summarized by the co‐primary endpoints, C max and AUC 0‐ꝏ . Pan et al. evaluated the bioequivalence in both fasted and fed states in the original trial and the effect of food on the C max and AUC 0‐ꝏ of pirfenidone after administering the new formulation. However, dosing without food leads to a higher C max and exacerbates skin‐related adverse events, nausea and vomiting, and opposes the advice of the drug label. Therefore, here, we focus on the objective of demonstrating bioequivalence in the fed state only and we ascertain there is good reason to not study administration of pirfenidone in the fasted state, which carries an increased risk of vomiting. The rationale for this needs to be provided in the protocol, and advice may be needed if the regulatory acceptability of the approach is uncertain. In this context we note that current FDA guidance recommends two separate studies in healthy adults to demonstrate bioequivalence for pirfenidone (one fasting and one fed study) , whereas current EMA guidance accepts only one trial in healthy adults either fasting or fed: “The SmPC recommends intake in fed state to minimise the risk of risk of nausea and dizziness. A fed study is, therefore, acceptable. However, a fasted study is also acceptable.” Finally, we will assume transportability of results so that the bioequivalence in healthy adults would be the same as in patients. Case study step 2: Identify intercurrent events While additional ICEs may be relevant, we focus on the ICEs listed in Table , which for brevity, we will reference as ICE1a, ICE1b, ICE2a, ICE2b, ICE3 and ICE4 throughout the subsequent steps of this case study. The wording of ICE2 (“Does not take both formulations”) has been refined after iterating through the thinking process as discussed in the section on general considerations. It reflects the situation where we want to evaluate bioequivalence in those who can tolerate both formulations, and we know that a useful design that is aligned to such an estimand is a crossover design. In such a design, a dose may not be received in the second period either because of (a) a tolerability issue in the previous period or (b) due to other reasons (for example, logistical issues in attending the subsequent period). Importantly, using different estimand strategies according to whether ICEs are related (ICE1a and ICE2a) or unrelated (ICE1b and ICE2b) to treatment requires that they can be sufficiently distinguished. This needs to be considered before the trial is started and may affect the case report form design and trial conduct. If ICE1 and ICE2 cannot be reliably sub‐categorized, then use of different estimand strategies to handle (a) related and (b) unrelated reasons would be precluded (Step 3 below). As highlighted in the general Step 7, any difference in frequency or pattern in the occurrence of relevant ICEs might be informative about differences in the investigated treatments and should be carefully investigated. Specifically, this comparison can be used to investigate the assumption that the tolerability does not differ between formulations. Case study step 3: Discuss strategies to address intercurrent events The selected handling strategy for each ICE defined in Step 2 is provided in Table . The rationale for selecting these strategies is described below. In clinical practice, patients with idiopathic pulmonary fibrosis who can tolerate an 801 mg total dose of pirfenidone with food are expected to use either three 267 mg capsules or one 801 mg tablet interchangeably. Those taking capsules are likely to transition to the tablet for convenience when this becomes their stable dose. We focus on establishing bioequivalence in the principal stratum of healthy adults able to tolerate both formulations, which corresponds to considering the subset of healthy adults for whom ICE1a and ICE2a would not occur. As this is a bioequivalence trial comparing different formulations, sensitivity to changes in absorption is paramount. In line with good clinical practice requirements, PK trials of healthy participants are conducted in a way that minimizes dosing errors and the need for interacting drugs or substances. Issues might emerge, such as viral or bacterial gastroenteritis (not treatment‐related nausea) or a requirement for a prohibited medicine to treat a new condition, such as taking a painkiller that contains caffeine to treat a severe headache from caffeine withdrawal. These events introduce noise in the data collected afterwards and, depending on their frequency and pattern, may jeopardize interpretation of the trial data and highlight a quality issue concerning trial conduct. In any case, the trial protocol should prespecify how to handle such events, although data collected before them may still be relevant. It might not be sensible to define a principal stratum of people who would never take caffeine or any other interacting substances, have gastroenteritis or be dosed in error. Instead, we wish to estimate the treatment effect purely due to the difference in formulations in the absence of these issues, which corresponds to the approach described by ICH E9(R1) as a hypothetical strategy. In other words, we are interested in the PK profile that would have been observed if the individual had not been unwell with viral or bacterial gastroenteritis (ICE1b), as though they had taken the second formulation rather than discontinuing due to reasons unrelated to treatment (ICE2b), without a dosing error (ICE3) or taking any interacting drugs or substances that might affect absorption or clearance (ICE4). Note, rather than focus on a principal stratum of those able to tolerate both formulations, we may have chosen to apply a hypothetical strategy to assess bioequivalence in healthy adults as though they would not have vomiting issues or discontinuation from subsequent administration for any reason. See Step 3 of the general description of the thinking process and Table for more details. Case study step 4: Precisely define the estimands Idiopathic pulmonary fibrosis is strongly associated with pre‐existing gastroesophageal reflux disease, where stomach acid flows back into the oesophagus, potentially delaying drug absorption and introducing variability in PK profiles. Such patients may be receiving multiple drugs, further altering metabolism of the formulations under investigation. For these patients, multiple blood draws for PK sampling may be an unnecessary burden. Although these issues may alter PK for both formulations and increase variability, it is reasonable to believe that the estimated ratio of exposure can be generalized from healthy trial participants to healthy adults and finally transported to patients (see Figure ). Therefore, to meet the trial objective, we recommend specifying “healthy adults” rather than “patients with idiopathic pulmonary fibrosis” as the target population attribute of the estimand. The transportability of bioequivalence from healthy individuals to the target patient population should be justified in the trial protocol. All five attributes of our co‐primary estimands with the endpoints AUC 0‐ꝏ and C max are detailed in Table . In a sentence, estimand 1a is described as: The geometric mean ratio (test/reference) of pirfenidone AUC 0‐ꝏ after a single oral dose, comparing the test formulation of one 801 mg film‐coated tablet with the reference of 3 × 267 mg capsules in healthy adults able to tolerate both formulations administered with food as though dosed correctly with a standard meal, without intake of interacting drugs or substances and without vomiting or diarrhoea due to gastroenteritis (unrelated to tolerability). For C max , we propose a similar estimand 1b where AUC 0‐ꝏ is replaced by C max . Further details on the individual attributes are described in Table . Case study step 5: Align choices on trial design, data collection and method of estimation ICH E9(R1) advocates early discussions on the choice of estimand before making design choices (Step 5) rather than vice versa. Although the trial described by Pan et al. investigated bioequivalence under fed and fasted states in a 4 × 4 crossover design, here we focus only on the fed state. Due to the therapeutic setting and the intent of treatment (Step 1), we propose a 2 × 2 crossover design with a sufficient washout in healthy individuals under controlled conditions; this design is selected for efficiency and to facilitate identification of the principal stratum of those who can tolerate both formulations. Identifying ICEs (Step 2) may help adapt how the trial is conducted to minimize the extent of ICEs unrelated to tolerability. For instance, the trial can be adapted to anticipate potential interacting substances. In this crossover trial, participants who do not tolerate pirfenidone in Period 1 would not be exposed in Period 2, which would allow them to withdraw from the trial after sufficient safety follow‐up. It would not be necessary to collect blood samples within a PK profile after the occurrence of ICE1b, ICE2b, ICE3 and ICE4 events in the affected period. The first step in the analysis is to derive the PK endpoints AUC 0‐ꝏ and C max , typically by use of non‐compartmental methods on the relevant data dependent on the ICE handling strategies. Figure illustrates the interplay between the ICE handling strategies and the definition of participant and data points sets. The participant set would include Participants 2–5 as being those in the principal stratum who did not have any tolerability issues after administration of either formulation. Participants 1 and 6 would be completely excluded from the participant set due to treatment‐related vomiting and intolerable rash, respectively. Participants who experienced dosing deviations, did not take both formulations for treatment‐unrelated reasons, used interacting drugs/substances or experienced treatment‐unrelated vomiting or diarrhoea would be included in the participant set of interest, and PK endpoints from the period not affected by ICEs would be used. The concentration‐level data points set would include the relevant concentration data for those in the participant set. Data in the affected period after the occurrence of ICE1b, ICE2b, ICE3 and ICE4 (Table ) would be considered irrelevant given the selected hypothetical strategy. This may result in some PK endpoints being non‐evaluable and set as missing. If an ICE clearly occurred after the expected T max , an observed C max could be used. In addition, AUC 0‐ꝏ could be extrapolated using the concentration values prior to the ICE, as if the ICE had not occurred. However, clear rules for handling such cases would need to be predefined. In Figure , the following ICEs affect the selection of relevant concentration data and subsequent PK endpoint set: ○ Participant 2 does not take all three tablets in Period 1 but complies in Period 2, so only data from Period 2 are included. ○ Participant 3 presents a dosing deviation in Period 2, so only data from Period 1 are included. ○ Participant 4 uses interacting substances in Period 2 after C max occurred, so data from Period 1 and parts of Period 2 can be used. In this case, the C max derived from the PK measurements would be used in the analysis. It may also be possible to extrapolate AUC 0‐ꝏ from the last quantifiable concentration before interacting substances were used in Period 2. ○ Participant 5 has no ICEs so all data from both periods can be used. Admittedly, this is an extreme example to illustrate the various impacts of ICEs, as it results in having to exclude some data from five of the six participants. In a well‐conducted bioequivalence trial of a well‐tolerated drug, most participants would not experience ICEs and, thus, all of their data could be used. A high number of participants with partly excluded data points would raise concerns about the validity of the trial or integrity of the data and would require a review of trial conduct and mitigation plans in line with good clinical practice. The second step is to estimate the two co‐primary estimands of interest (Table ). Different statistical approaches could be used. For this case study, we use a linear mixed model on log‐transformed PK endpoints AUC 0‐ꝏ and C max with sequence, treatment and period as fixed effects and participant as random effect. The linear mixed model would implicitly predict the endpoint values that could not be derived due to an ICE in the hypothetical scenario making use of PK endpoints from the period not affected by ICEs where available. Truly missing data, e.g., laboratory failure, would be predicted in the same way. Finally, if the 90% confidence interval for the estimated ratio of geometric means (test/reference) lies within standard bioequivalence limits of 0.80–1.25 (i.e., 80%–125%), , , we would declare bioequivalence. Case study step 6: Identify assumptions for the main analysis and suitable sensitivity analyses The aim of the main analysis is to demonstrate bioequivalence in participants who can tolerate both the test and reference formulations. The 2 × 2 crossover design allows us to determine if a participant can tolerate both formulations. In this context, and assuming both formulations have equivalent and minimal carryover in the measured variable between periods, the effect in the principal stratum can be estimated without bias by limiting the analysis to participants who do not experience drug‐related tolerability issues. If a participant can tolerate the formulation in Period 1 and is not exposed in Period 2, we could assume that they would also tolerate the formulation given in Period 2. However, this may be incorrect. In this case, a sensitivity analysis could be performed in which the participant is completely excluded from the participant set and so none of their data would contribute. We assume that missing data, dosing deviations, use of interacting drugs or substances or treatment‐unrelated vomiting or diarrhoea occur at random with both formulations. More specifically, we assume that the probability of these ICEs depends only on the variables included in the mixed model (sequence, treatment, period and participant). When these assumptions are appropriate, a mixed‐model approach would allow for an unbiased assessment by including all available measurements. This approach implicitly models the value the endpoint would have taken in the hypothetical scenario, as defined by introducing a random effect of participant into the model, in other words, by combining between‐ and within‐participant information. In this approach, the robustness of the results against violations to the assumption that missing data and ICEs unrelated to tolerability issues occur at random (conditionally on the variables included in the model) will need to be explored carefully. For example, a multiple imputation approach with penalties to explore the robustness of results could be used. In this case, the imputed PK values are multiplied or divided by a series of penalties of increasing magnitude to inflate or deflate exposure levels for each group separately. For each pair of penalties, the analysis is repeated to investigate if the bioequivalence still holds. The penalties may be applied to the predicted missing data only, or to both predicted missing data and predicted outcomes for participants with intercurrent events handled by a hypothetical strategy. Case study step 7: Document the estimand FDA guidance recommends documenting the estimand in the protocol for bioequivalence trials. Although the original trial did not consider estimands, in future trials, the information in Table could be presented alongside the trial objective. This should be supplemented with the rationale for the defined estimand. REGULATORY GUIDANCE Guidelines on bioequivalence from the EMA, Pharmaceuticals and Medical Devices Agency in Japan, , and Health Canada, all published before ICH E9(R1), as well as the draft ICH M13A guidance, do not explicitly incorporate the estimand framework. Nonetheless, they all have relevant touchpoints with this framework. In particular, all guidelines discuss the trial population or target population, although they do not always clearly distinguish between them (see Step 4 in the case study). Additionally, all guidelines discuss treatment conditions, endpoints and relevant population‐level summaries. The ICH E9(R1) addendum not only brings these interrelated attributes together but also introduces ICEs and related strategies. Although, this is not explicitly covered in the bioequivalence guidelines above, implicit reference to some relevant ICEs can be identified in them. For example, violations of the required fasting status, vomiting or diarrhoea are mentioned and can be interpreted as relevant ICEs. Designs and statistical approaches are suggested in some guidance documents, but the underlying scientific questions of interest they intend to address are not explicitly discussed in terms of the estimand framework. The FDA provides relevant guidelines, , , including specific guidance on bioequivalence studies with PK endpoints for drugs submitted under an abbreviated new drug application. As in the other regulatory guidelines, the estimand framework is not explicitly referred to, although they implicitly discuss estimand attributes in sections on the trial population, PK measures of rate and extent of absorption, and violating dosing requirements. To date, the only clinical pharmacology guidance that explicitly advises using the estimand framework is the draft Guidance on Statistical Approaches to Establishing Bioequivalence published by the FDA in December 2022. It recommends that “the trial protocol of a bioequivalence trial should include the following components of an estimand” and lists all five estimand attributes. Although it refers to the estimand framework, the suggested application does not follow the steps of the thinking process starting with the trial setting, objective and estimand as a prerequisite for discussing design and analysis. Instead, it discusses the estimand under a section on data preparation after trial design considerations. Further, it provides an example of equivalence in a clinical response endpoint rather than bioequivalence using a PK endpoint. This draft guidance comments on both missing data and ICEs under data handling without consideration of how ICEs affect what data would be relevant to the scientific question. In addition, in this guidance there is no clear distinction between the estimand's target population and the trial's analysis data set required for the estimation approach. In the estimand thinking process described, ICEs would be discussed in Steps 2 and 3, and a suitable estimator, including approaches for handling missing data, would be derived separately afterwards in Step 5. The estimand framework is not part of the draft general FDA guidance and is discussed only in their statistical guideline. Whether the direction in the statistical guidance for using the estimand framework will be extended to the final version is not clear. Estimands affect core trial features including the clear definition of the objectives, trial conduct, data collection and statistical methodology. Currently, literature on the application of the estimand framework and thinking process to bioequivalence trials is lacking, but this paper paves the way towards implementing them and highlighting some differences from efficacy trials. It will be important to follow discussions between industry, academia, regulators and the ICH on the application of the estimand framework to bioequivalence , , , , , , , and any changes to the general guidance will need to be monitored. DISCUSSION AND CONCLUSIONS FDA guidance advises sponsors to specify estimand(s) in bioequivalence trial protocols. In this paper, we described how the estimand framework and thinking process can be applied generally to clinical pharmacology trials, and then we illustrated how to implement these in a bioequivalence case study. The relevance of a specific estimand is context‐dependent and careful considerations are needed for all individual trials. The proposed estimand in this case study has not been subject to regulatory scrutiny, and some regulatory reviewers may prefer evaluating bioequivalence using another estimand, specifically using different strategies to handle the ICEs. Note, Ring and Wolfsegger worked backwards from the existing FDA and EMA guidance on bioequivalence trials to derive their underlying estimands, but whether these derived estimands are indeed the regulators' preferred ones remains unclear. An appeal of the estimand framework is that it provides transparency and enables an early dialogue at the protocol development stage between functions and stakeholders on the appropriate strategies to fit the setting. The thinking process starts with multidisciplinary discussions to understand the setting and define the objective(s) and scientific question of interest (estimand). It is then important to ensure alignment of trial design, and statistical analysis but also to clarify whether data need to be collected after ICEs occur and if mitigation plans to minimize the extent of ICEs are required. This highlights the need to record detailed information on ICEs in case report forms, including their timing. Lastly, by identifying the underlying assumptions of the main analysis, suitable sensitivity analyses can help determine whether specific assumption violations affect the results. The deep and structured thinking process discussed in this paper should clarify and improve the dialogue and help obtain agreement between disciplines and stakeholders. In other words, the estimand thinking process ensures that the trial will address the questions it intends to answer. All clinical trials can benefit from the application of estimands, but regulatory interest will be greater for trials affecting regulatory decision‐making and drug labelling. For instance, estimands may need to be defined in protocols for the following clinical pharmacology trials that impact drug labelling: bioequivalence, biosimilar and bioavailability trials; drug–drug interaction trials; trials investigating the effect of administration in fed vs . fasted state; trials investigating the relationship between PK concentration and prolongation of the heart‐rate corrected QT interval; trials in pregnant or lactating females and their infants; and renal and hepatic impairment trials. This framework would also benefit and guide the design of exploratory PK and PD trials, and it could ultimately become best practice across all trial types. Applying the estimand framework can help clarify trial objectives and avoid other ambiguities in the trial design. For example, bioequivalence and bioavailability trials are generally similar in their design and conduct, but their objectives differ because the PK of the test and reference drugs are designed to be comparable in bioequivalence trials but to differ in bioavailability trials. This fundamental difference between trial types should be clear from the protocol. The estimand thinking process can help to clarify the intent of treatment and trial objective. For example, a primary objective of most single ascending dose and multiple ascending dose trials is to characterize the PK of the test drug; this is too vague and poorly defined. Which PK parameters will be derived? Will the relationship between dose and exposure (dose‐proportionality) be assessed? Will the attainment of steady state and/or time‐dependency be evaluated? In drug–drug interaction trials, applying the estimand framework would ensure that key trial considerations, such as which drug is the perpetrator and which is the substrate, and which should be dosed to steady state (if any), are documented clearly. Another example is clinical lactation trials, where the estimand thinking process can help to clarify key PK endpoints and target population (breast milk, maternal serum/plasma or infant serum/plasma). In renal and hepatic impairment trials, defining the target populations precisely as part of the estimand thinking process will ensure that the intended target population are patients with impaired function, allowing them to be compared to individuals with normal organ function. In addition, in studies with pharmacodynamic objectives, it is often unclear which pharmacodynamic markers will be evaluated and how these markers will be analysed: will pharmacodynamic parameters be derived (e.g., maximum effect, area under the effect curve, etc.)? Will similar tests be applied to pharmacodynamic parameters that are applied to PK parameters (e.g., dose‐proportionality)? The above considerations tend to be accounted for by experienced clinical pharmacologists and project teams when designing clinical pharmacology trials; however, these rigorous steps and study design decisions are not always clear and well documented in the trial protocol. The estimand framework encourages teams to address these issues in the trial protocol, removes ambiguity from the trial design, and facilitates a collaborative approach to deliver well‐designed and more transparent clinical pharmacology trials. We anticipate that this structured approach may reduce clinical pharmacology‐related major deficiencies and increase the rate of regulatory success. Our paper demonstrates the feasibility of implementing estimands in clinical pharmacology trials. We expect estimands to become part of the usual dialogue across trial teams and between sponsors, regulators and other stakeholders and to lead to more robust science and more rapid regulatory approval. We welcome further discussions across industry, academia and regulatory agencies on the application of the estimand framework and thinking process beyond Phase 3 efficacy clinical trials, and we hope that additional case studies will be presented. These may lead to further updates to regulatory guidance encouraging the use of estimands in clinical pharmacology trials. The EIWG will continue exploring this field and sharing its experience. 5.1 Nomenclature of targets and ligands Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org , the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY, and are permanently archived in the Concise Guide to PHARMACOLOGY 2023/24. Nomenclature of targets and ligands Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org , the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY, and are permanently archived in the Concise Guide to PHARMACOLOGY 2023/24. All authors contributed equally during our many meeting discussions on the paper and provided written comments. Helle Lynggaard and Sue McKendrick initiated the work on this topic, identified the case study and co‐led the writing of the manuscript. Vivian Lanius, Florian Lasch, David Wright and the co‐leads provided input to the statistical aspects, and Mark Baird and Essam Kerwash gave clinical pharmacology inputs. The following indirect, potential conflicts of interest are declared: Helle Lynggaard, Sue McKendrick, Mark Baird, Vivian Lanius and David Wright continue to be employed solely by their affiliated companies (stated above) that are involved in the development of pharmaceuticals including trials evaluating bioequivalence. Helle Lynggaard, Sue McKendrick, Mark Baird and David Wright hold shares in companies involved in running clinical pharmacology trials. All other authors declared no competing interests for this work. The views expressed in this article are the personal views of the authors and may not be understood or quoted as being made on behalf of or reflecting the position of the regulatory agency/agencies or organizations with which the authors are employed/affiliated. |
An open chat with… Johannes Herrmann | c08096b5-bb6f-4218-af7e-457a65e89109 | 11452295 | Anatomy[mh] | I have always been a curiosity‐driven person. As a child, I drove my parents nuts during our holidays, when I woke up at 5:00 am to go bird‐watching and stroll through nature. I still go into the deep forests around our city for long hikes every free weekend I get (luckily my wife enjoys this as much as I do). I just love to discover things I have not seen before. When hiking, I regularly get lost, even on purpose, heading in new directions as I find it extremely rewarding if I suddenly bump into paths that I know. It is just so cool when things fall into place and get connected. My career as a researcher is actually inspired by the same motivation. As a student, I particularly liked cell biology and all the molecular mechanisms in the microcosmos of a cell. I was particularly intrigued by mitochondria because they are like mini‐worlds in our cells. They have their own genome, systems for protein synthesis and folding, a complex metabolism, and a large degree of autonomy from other cellular processes. Still, they are absolutely dependent on the rest of the cell, they import most of their proteins from outside and are tightly integrated into cellular processes. I was very fortunate as a PhD student to join the laboratory of Walter Neupert in Munich, a world‐leading hotspot of mitochondrial research. There, I studied the biogenesis of mitochondria‐encoded and imported proteins using yeast as a model. I continued working with yeast for my postdoc with Randy Schekman at UC Berkeley but changed the focus from mitochondria to the endoplasmic reticulum (ER). Today, in my laboratory, many projects address interactions of mitochondria with the ER, both being fascinating organelles. Thus, just as in our forests, I stroll around and love to explore new territory and find unexpected new connections between these cellular compartments. The Federation of European Biochemical Societies (FEBS) is an organization with over 30 000 members from all over Europe and neighbouring regions. This includes the European Union countries as well as Turkey, Israel, etc., and also many from the East such as Georgia or Armenia. Thus, FEBS offers a wonderful platform to support interactions between biochemists and biologists from very different regions—this is a major motivation for me to be active in FEBS. The Publication Committee discusses, supports and oversees the operation and development of the four journals of FEBS: FEBS Letters , FEBS Journal , FEBS Openbio and Molecular Oncology . The Publication Committee helps journals to attract excellent submissions from a broad authorship, to find well‐respected editors, and to have close cooperation with Wiley as their publisher. With these journals, FEBS can generate money to support the FEBS courses and meetings as well as FEBS fellowships. Thus, I think being active for these journals supports our scientific community, but I also just enjoy the meetings with other active European scientists at the meetings of the Publication Committee. Mitochondria are the hot heart of the cell! Mitochondria are much more than just ATP generators. They are for example important signalling platforms. Did you know that the antiviral signalling of human cells takes place on the mitochondrial surface? Thus, mitochondria are of central relevance for the defence of cells against viruses. This is why for example the SARS CoV‐2 virus induces the production of high amounts of a protein called Orf9b, which blocks a mitochondrial surface receptor to interfere with antiviral signalling. This viral protein camouflages the virus by binding to the mitochondrial Tom70 receptor and thereby also disturbs mitochondrial biogenesis . Another cool aspect is the ability of mitochondria to eject their genome into the cytosol to trigger inflammation and a defence programme against bacteria . Eukaryotic cells can sense the presence of bacterial DNA in their cytosol, which induces a very strong defence programme. However, in the absence of bacteria, this programme can be triggered by mitochondrial DNA which is ejected into the cytosol. Since the mitochondrial genome has similar features to that of bacteria, for example an absence of histones, the release of mitochondrial DNA is a cell‐autonomous mechanism to unleash this powerful defence system. I think the story my laboratory is currently working on is the coolest discovery we ever made! I have published almost 200 articles over the last 25 years. Still, we just found some exciting and unexpected aspects of mitochondrial biology. This is what keeps me awake at night, literally. You also asked what I am proud of. The word pride does not properly describe what I feel about our discoveries—however, I am very proud of (and grateful for) my wonderful team—we have an amazing laboratory family and I enjoy working with them every single day. The coolest stories are those that are not told yet and which are just emerging. For example, we just discovered that mitochondria deliberately produce a highly aggregation‐prone protein that is encoded in their genome. This protein gives rise to intramitochondrial aggregates which can protect cells from misfolded proteins under extreme stress conditions. This protein serves as a nucleation site for aggregates and absorbs and sequesters other proteins. A cool mechanism and again, a completely unexpected new aspect of mitochondrial biology. As I said before, I love to discover things. However, the best aspect of being a researcher is the stimulating interactions with so many interesting people. The daily discussions about experiments and results with the PhD students and postdocs of my team, teaching students in many different courses, and the interactions with colleagues at conferences and meetings—these social interactions with others are what makes the job of a professor the best profession one can think of. What I like the least: bureaucracy and fighting with editors and referees is not an enjoyable aspect of our work. Wow, this is a very general question. I think there are two major developments: 1) data sets are gettting larger and larger and 2) artificial intelligence (AI). When I was a PhD student, most papers showed western blots and simple figures. Now most papers show omics data and complex data sets. The complexity will continue to increase further and soon most studies will show omics data from many single cells or tissues. This is not only extremely challenging and requires computational biologists, but also makes papers very difficult to read as they often appear unfocused and overly generalized. The second important trend is the influence of artificial intelligence and machine learning. This offers wonderful opportunities, for example for the analysis of these large data sets, but it also makes it very difficult to identify manipulated and fraudulent studies. Actually, I think AI will have an immense effect on research. Alphafold is an excellent example of this potential. However, I do not see AI as being so relevant for students and teaching. Research and publications have changed a lot over the last 25 years. However, how students learn the structures of amino acids or the steps of glycolysis did not change. To be honest, my teaching is extremely old‐fashioned and traditional. For example, I like to make big drawings with chalk on a blackboard to explain the steps of mitosis or how proteins are degraded by the proteasome. This teaching style is rather traditional, but perhaps it is the reason why the students awarded me a teaching prize. Twitter ) which you use to highlight your and others' research, and a respectable number of people follow you. What are the biggest benefits you have reaped from its use as a scientist that you would use as arguments to convince peers to jump on this ship? What should fellow scientists be wary of if they decide to engage in sci‐com through social media? I really loved Twitter to hear about novel discoveries from the laboratories and people I followed, and to tell others about our unpublished observations. For me, Twitter was an alerting platform, and the short format of the tweets was perfect for this. Unfortunately, since it changed to X, there are advertisements and content that is not related to science. So, many scientists became less active, myself included. We discovered that mitochondrial precursor proteins can be transiently stored in the cytosol if their levels exceed the mitochondrial import capacity. Small heat shock proteins serve as nucleation sites to absorb these precursors, thereby forming aggregates which we called MitoStores . As soon as the mitochondria ramp up their import capacity, the precursors are released from the MitoStores and imported into mitochondria. We placed a tweet with this story on Twitter ( https://twitter.com/Herrmann_lab/status/1554703848069111808 ) before we submitted the manuscript which went viral. An editor of the EMBO Journal saw this and invited us to submit our study to EMBO. This was an excellent experience and shows that a platform like Twitter can be very good for scientists, as long as it maintains its focus on science. Well, I hope that FEBS can help with the resolution of conflict between European and neighbouring countries. We, as scientists, have the chance to make and keep contacts across borders which are closed for politicians and others. Scientists therefore should use FEBS as a platform, for example to interact with students from countries such as Iran or Russia. I am an editor of several other journals as well. What I like about the four FEBS journals is that the editorial offices are highly professional, for example, in filtering out problematic and fraudulent manuscripts that come from paper mills before they go to peer‐review. All studies are seriously reviewed which is—as we all know—not the case for some of the new journals from other publishers. The FEBS journals also have very attractive web pages that provide a well‐illustrated overview of the content of the issues. Even if the impact factors are not as high as in some other journals, I regard them as very good and well‐respected scientific journals, at least in my community. I think we can't. Evaluating research grants and manuscripts is part of our profession. It should be an honour for every scientist to serve the community. I might be old‐fashioned, but I am convinced that trying to reward them with some Publon points or with dollars just goes in the wrong direction. The reward for active participation in these affairs is the existence of a strong and vital scientific community. This should be more than enough for encouragement. The authors declare no conflict of interest. IT formulated the questions, conducted the interview and edited the manuscript. JH answered the questions during the interview and provided comments on the manuscript. |
Seeking abortion medications online: experiences from a mystery client study in Colombia | adf7918d-cefd-46dd-a9fa-e789c2495456 | 11752010 | Surgical Procedures, Operative[mh] | Over the last two decades, Colombia has benefited from multiple law changes that have increased access to abortion. Medication and procedural (sometimes called surgical) abortions are now legally available within the formal health system at no cost to patients, regardless of insurance coverage or immigration status. As of 2024, Colombia has the most progressive national legal framework for abortion in Latin America and the Caribbean. Colombia’s current abortion policy environment contrasts starkly with that of the 1990s and early 2000s; liberalisation came about gradually after many years of advocacy and feminist mobilisation of non-governmental organisations, lawyers, healthcare providers and allies. Key among these advocates was La Mesa por la Vida y la Salud de las Mujeres (La Mesa ) (The Committee for the Life and Health of Women (The Committee)), a feminist activist collective focused on women’s sexual and reproductive rights, especially the right to abortion, which in 2017 created the Causa Justa (Just Cause) movement (along with other organisations, including el Centro de Derechos Reproductivos (Center for Reproductive Rights), el Grupo Médico por el Derecho a Decidir (Global Doctors for Choice), Católicas por el Derecho a Decidir (Catholics for the Right to Decide) and Women’s Link, to seek freedom and reproductive autonomy for all women over their bodies and lives. Prior to 2006, Colombia had a complete ban on abortion without any exceptions. As a result of advocacy in the 16 intervening years, in February 2022, the Constitutional Court issued Ruling C-055, which fully decriminalised abortion up to 24 weeks (and also made it available at later gestational ages under three conditions established in the previous 2006 ruling: when the life or health of the pregnant person is in danger, when the fetus presents malformations incompatible with life, and when the pregnancy is the result of rape, incest or non-consensual insemination). Health system policy changes to align with shifting legality of abortion Colombia’s Ministry of Health (MOH) adapted the national health policies and service delivery guidelines to comply with legal changes and operationalise abortion access for the population. These guidelines established abortion as an essential component of maternal healthcare. Misoprostol for medication abortion was approved by the Instituto Nacional de Vigilancia de Medicamentos y Alimentos (the National Institute of Medication and Food Surveillance) in 2006 and as of mid-2024, the medication is currently available under the brand name Cytil in 7-pill and 12-pill packs. Mifepristone was approved 11 years later in 2017. Mifepristone can be prescribed with misoprostol by physicians in a facility or via telemedicine (in the first trimester of pregnancy for the latter). The Colombia MOH guidelines detail recommended dosing for the use of mifepristone and misoprostol or misoprostol alone for medication abortion at different gestational ages of pregnancy. The combined mifepristone-misoprostol regimen is slightly more effective, but in settings where mifepristone is not available, using misoprostol-only is safe, effective and recommended. Serious complications and the need for additional interventions are rare with both regimens. For first trimester abortions using misoprostol-only (<12 weeks), the MOH guidelines are an initial dose (800 μg), followed by two to three repeated doses (a recent systematic review and other international guidelines recommend a minimum of three misoprostol doses to complete the abortion process). For 12–24 weeks gestational age, the guidelines are to use repeated doses of 400 μg misoprostol until the products of conception have been expelled. Colombia’s MOH guidelines also recommend that medication abortions beyond 12 weeks gestational age be performed under clinical supervision at a health facility until the process is complete. Regardless of gestational age, accessing abortion care in Colombia often still involves some contact with a health facility; even for medication abortions, MOH regulations are that misoprostol should be provided with a medical prescription. The sale of medicines without a medical prescription and without the corresponding sanitary registration (as in the case of Cytotec brand of misoprostol in Colombia) can lead to administrative sanctions, fines and even imprisonment. Partly in response to the COVID-19 pandemic, some non-profits began introducing the option of accessing medication abortion through telemedicine. Guidelines for provision of quality clinical abortion care from the World Health Organization (WHO) underscore that people seeking abortion at any point in pregnancy should be screened for gestational age as well as any contraindications for the use of mifepristone or misoprostol and be provided information about what to expect during the abortion process, including when to seek follow-up care for rare, but possible complications. As most pregnant people can accurately determine their eligibility for self-managed medication abortion, administer the medications, and assess the success of the abortion, WHO endorses self-managed abortion models of care in the first trimester (<12 weeks). Completion rates have been 88%–94% in recent studies of first trimester self-managed misoprostol-only regimens. Studies of self-managed second trimester medication abortions have found effectiveness rates of 74%–93%. People self-managing their abortions may also be supported by accompaniment groups, which are often activist organisations made up of trained members, which provide evidence-based counselling and support before, during and after the abortion process. Barriers to obtaining abortion care in line with health policy guidelines Although the Colombia MOH issued policy guidelines to guide the provision of abortion care (Resolution 051 of 2023) in response to legal changes, gaps remain in equitable and universal implementation in service delivery. Before the 2022 court ruling, advocates and researchers in Colombia documented the numerous barriers that pregnant people faced trying to obtain abortion care. Healthcare facilities failed to provide abortion services due to lack of knowledge of the legal framework, and legal and civil society systems lacked the necessary support to protect and facilitate people’s receipt of care. Religious beliefs and abortion stigma deterred some pregnant people from seeking out abortion services altogether and these beliefs and attitudes also prompted providers to refuse abortion requests. For rural and indigenous communities, Venezuelan migrants and other vulnerable groups with precarious socioeconomic and immigration statuses, challenges to obtaining care were compounded. Even after the 2022 ruling, similar barriers to care have persisted, including lack of knowledge and restrictive interpretations of the legal framework, discriminatory or unjust treatment by healthcare providers, delays in care, privacy violations, lack of availability of abortion medications and difficulties obtaining abortion using health insurance. Decriminalising abortion in and of itself has not been enough to guarantee care access in Colombia; rather, it is still necessary to verify that service providers are integrating the new regulations into the care pathway for patients, as well as to widely disseminate information to the public about the new regulations and the ways in which they can access abortion services through the health system. Continued training and knowledge updates for health personnel can also reinforce following the protocols established by the MOH. Again, even after the liberalisation of abortion laws, the quality of care in Colombia is variable. Research participants in Colombia have highlighted that healthcare providers can lack empathy, timeliness and confidentiality, leading people to seek abortion care from other sources. In response to disparate and discriminatory experiences, feminist networks of trained activists and volunteers organised across Latin America, including Colombia, to support abortion seekers throughout their pregnancy termination processes. They provide instructions on how to obtain and/or safely take the drugs in addition to step-by-step emotional and practical support by phone or online in accordance with international guidelines; some of them also provide medication abortion pills (mifepristone and misoprostol or misoprostol alone), sometimes free of charge or for a small donation. Beyond these feminist networks, other individuals sell medication abortion pills through channels outside of the formal health system or guidelines. Obtaining medication abortion pills from these sellers is one of the most common ways of terminating a pregnancy outside of the healthcare system, even though sales of registered drugs without a prescription can lead to legal repercussions for the sellers. Previous studies in Mexico have documented the sale of misoprostol over-the-counter at pharmacies (ostensibly for gastric ulcer and not abortion) and two studies conducted in Colombia before the 2022 law change found that misoprostol could also be accessed without a prescription from a variety of sources, including small-chain and independent drug stores, street vendors and online sellers. One-third of qualitative study participants who had acquired misoprostol from these channels had purchased it online. Although the sample was small, those who used online vendors received more detailed instructions and information about what to expect during the abortion compared with those who had purchased the drug elsewhere. These preliminary data suggested that online sellers may be an important source of medication abortion in Colombia. The practice of buying medication abortion pills online that emerged under previous restrictions may persist even in the face of broad decriminalisation given lack of knowledge of the legal framework or varying quality of experiences at health facilities. Because barriers to legal abortion care persist in Colombia and medication abortion pills continue to be available for purchase outside of the formal healthcare system, the objective of this study was to assess if individuals purchasing medication abortion pills from online sellers in Colombia receive the information, instructions and medication dosages according to Colombia’s MOH guidelines for terminating pregnancies at various gestational ages. To make this assessment, we also sought to gather information about the landscape of online sellers of medication abortion pills in Colombia and to document the medications received from these sources.
Colombia’s Ministry of Health (MOH) adapted the national health policies and service delivery guidelines to comply with legal changes and operationalise abortion access for the population. These guidelines established abortion as an essential component of maternal healthcare. Misoprostol for medication abortion was approved by the Instituto Nacional de Vigilancia de Medicamentos y Alimentos (the National Institute of Medication and Food Surveillance) in 2006 and as of mid-2024, the medication is currently available under the brand name Cytil in 7-pill and 12-pill packs. Mifepristone was approved 11 years later in 2017. Mifepristone can be prescribed with misoprostol by physicians in a facility or via telemedicine (in the first trimester of pregnancy for the latter). The Colombia MOH guidelines detail recommended dosing for the use of mifepristone and misoprostol or misoprostol alone for medication abortion at different gestational ages of pregnancy. The combined mifepristone-misoprostol regimen is slightly more effective, but in settings where mifepristone is not available, using misoprostol-only is safe, effective and recommended. Serious complications and the need for additional interventions are rare with both regimens. For first trimester abortions using misoprostol-only (<12 weeks), the MOH guidelines are an initial dose (800 μg), followed by two to three repeated doses (a recent systematic review and other international guidelines recommend a minimum of three misoprostol doses to complete the abortion process). For 12–24 weeks gestational age, the guidelines are to use repeated doses of 400 μg misoprostol until the products of conception have been expelled. Colombia’s MOH guidelines also recommend that medication abortions beyond 12 weeks gestational age be performed under clinical supervision at a health facility until the process is complete. Regardless of gestational age, accessing abortion care in Colombia often still involves some contact with a health facility; even for medication abortions, MOH regulations are that misoprostol should be provided with a medical prescription. The sale of medicines without a medical prescription and without the corresponding sanitary registration (as in the case of Cytotec brand of misoprostol in Colombia) can lead to administrative sanctions, fines and even imprisonment. Partly in response to the COVID-19 pandemic, some non-profits began introducing the option of accessing medication abortion through telemedicine. Guidelines for provision of quality clinical abortion care from the World Health Organization (WHO) underscore that people seeking abortion at any point in pregnancy should be screened for gestational age as well as any contraindications for the use of mifepristone or misoprostol and be provided information about what to expect during the abortion process, including when to seek follow-up care for rare, but possible complications. As most pregnant people can accurately determine their eligibility for self-managed medication abortion, administer the medications, and assess the success of the abortion, WHO endorses self-managed abortion models of care in the first trimester (<12 weeks). Completion rates have been 88%–94% in recent studies of first trimester self-managed misoprostol-only regimens. Studies of self-managed second trimester medication abortions have found effectiveness rates of 74%–93%. People self-managing their abortions may also be supported by accompaniment groups, which are often activist organisations made up of trained members, which provide evidence-based counselling and support before, during and after the abortion process.
Although the Colombia MOH issued policy guidelines to guide the provision of abortion care (Resolution 051 of 2023) in response to legal changes, gaps remain in equitable and universal implementation in service delivery. Before the 2022 court ruling, advocates and researchers in Colombia documented the numerous barriers that pregnant people faced trying to obtain abortion care. Healthcare facilities failed to provide abortion services due to lack of knowledge of the legal framework, and legal and civil society systems lacked the necessary support to protect and facilitate people’s receipt of care. Religious beliefs and abortion stigma deterred some pregnant people from seeking out abortion services altogether and these beliefs and attitudes also prompted providers to refuse abortion requests. For rural and indigenous communities, Venezuelan migrants and other vulnerable groups with precarious socioeconomic and immigration statuses, challenges to obtaining care were compounded. Even after the 2022 ruling, similar barriers to care have persisted, including lack of knowledge and restrictive interpretations of the legal framework, discriminatory or unjust treatment by healthcare providers, delays in care, privacy violations, lack of availability of abortion medications and difficulties obtaining abortion using health insurance. Decriminalising abortion in and of itself has not been enough to guarantee care access in Colombia; rather, it is still necessary to verify that service providers are integrating the new regulations into the care pathway for patients, as well as to widely disseminate information to the public about the new regulations and the ways in which they can access abortion services through the health system. Continued training and knowledge updates for health personnel can also reinforce following the protocols established by the MOH. Again, even after the liberalisation of abortion laws, the quality of care in Colombia is variable. Research participants in Colombia have highlighted that healthcare providers can lack empathy, timeliness and confidentiality, leading people to seek abortion care from other sources. In response to disparate and discriminatory experiences, feminist networks of trained activists and volunteers organised across Latin America, including Colombia, to support abortion seekers throughout their pregnancy termination processes. They provide instructions on how to obtain and/or safely take the drugs in addition to step-by-step emotional and practical support by phone or online in accordance with international guidelines; some of them also provide medication abortion pills (mifepristone and misoprostol or misoprostol alone), sometimes free of charge or for a small donation. Beyond these feminist networks, other individuals sell medication abortion pills through channels outside of the formal health system or guidelines. Obtaining medication abortion pills from these sellers is one of the most common ways of terminating a pregnancy outside of the healthcare system, even though sales of registered drugs without a prescription can lead to legal repercussions for the sellers. Previous studies in Mexico have documented the sale of misoprostol over-the-counter at pharmacies (ostensibly for gastric ulcer and not abortion) and two studies conducted in Colombia before the 2022 law change found that misoprostol could also be accessed without a prescription from a variety of sources, including small-chain and independent drug stores, street vendors and online sellers. One-third of qualitative study participants who had acquired misoprostol from these channels had purchased it online. Although the sample was small, those who used online vendors received more detailed instructions and information about what to expect during the abortion compared with those who had purchased the drug elsewhere. These preliminary data suggested that online sellers may be an important source of medication abortion in Colombia. The practice of buying medication abortion pills online that emerged under previous restrictions may persist even in the face of broad decriminalisation given lack of knowledge of the legal framework or varying quality of experiences at health facilities. Because barriers to legal abortion care persist in Colombia and medication abortion pills continue to be available for purchase outside of the formal healthcare system, the objective of this study was to assess if individuals purchasing medication abortion pills from online sellers in Colombia receive the information, instructions and medication dosages according to Colombia’s MOH guidelines for terminating pregnancies at various gestational ages. To make this assessment, we also sought to gather information about the landscape of online sellers of medication abortion pills in Colombia and to document the medications received from these sources.
This study was a collaboration between Fundación Oriéntame, a private, non-profit organisation in Colombia focused on promoting and protecting sexual and reproductive health, and the Guttmacher Institute, a research and policy organisation focused on advancing sexual and reproductive health and rights worldwide. The methodology for this study consisted of four stages: (1) determination of keywords related to purchasing abortion medications online in Colombia using Google Trends; (2) identification of the universe of sellers from multiple online platforms (websites, Facebook, Instagram, and TikTok) using keyword searches and documentation of website and profile content; (3) deduplication of universe of sellers who had more than one website or profile and (4) collecting information on mystery clients’ contacts with sellers, including purchase and receipt of pills. More detail on each of these stages is described below. A variation of this methodology was implemented previously by the Guttmacher Institute in Indonesia.
We first used Google Trends to determine keywords being used by individuals seeking abortion pills online. To follow the potential logic of a person who does not know much about medication abortion, we began the search using general keywords such as pastillas abortivas (abortion pills) and progressed to more specific ones, using keywords that showed up as “related queries” on Google Trends. The keywords were identified on 4 July 2023. We then added complementary keywords to the search such as precio (price) and comprar (buy) as well as the names of the largest cities in Colombia: “Bogotá”, “Medellín” and “Cali”, with the goal of identifying sellers trying to reach purchasers in specific cities. We also used another set of keywords that included misspellings that sellers may strategically adopt to prevent their ads from being removed from the platforms. We filtered by location, time range and search type, but did not filter by category. In this way, we used “Colombia” as our location, “12 months” to “last day” as the time range and “web search” as our search type. We did several searches with the same keywords and different time ranges to see how the “related queries” varied and confirmed that more recent searches were often about price, while searches that went up to 12 months had additional topics including instructions and physical effects (see for specific keywords used in the searches). It may be the case that the most recent searches were people seeking to buy pills for medication abortion. However, by including a longer time frame in our search, we captured a broader range of relevant keywords, including those related to symptoms and complications of the pills.
To identify the universe of sellers, we used the identified keywords to search on four different platforms: Google’s search engine (for websites), Facebook, Instagram and TikTok. (Some popular marketplaces in the country such as Mercado Libre and OLX were considered, but we did not obtain any relevant findings using the keywords). To identify the websites, we used a Google Chrome browser in Incognito mode, searched within Google using the keywords, and checked all the results found on the first 10 pages of each search, starting with the highest-ranking results (the most relevant according to the Google algorithm). We excluded news and academic articles and pharmacy websites selling misoprostol with a prescription. No Facebook, Instagram and TikTok profile websites shown in Google search results were captured in this stage—those platforms were searched separately. International sellers were included if they sold to people in Colombia. To conduct searches on Facebook, Instagram and TikTok, the data collectors serving as mystery clients created accounts on each platform, logged into those accounts using a Chrome Incognito window and searched each platform. For searches on Facebook, we used its Marketplace feature. On Instagram, we used the Map search, in addition to the regular search function, which we also used on TikTok. For search results offering mifepristone and/or misoprostol pills, we identified the profile of each post included in the list of results, taking care to include them only once, no matter how many different posts had the same profile. For these three social networks, all the search results returned for each of the keywords were reviewed. Searches used all identified keywords on every platform and were conducted between July and August 2023. A website or profile was included if: (a) it stated or suggested selling mifepristone and/or misoprostol and (b) its landing page was not otherwise listed in search results (eg, a subpage of a website that was already captured). Each website and social media profile was given an anonymised seller ID consisting of an abbreviation of the platform on which their page was found and a number. For each website/profile identified, a fieldworker recorded the information listed about medication abortion (including products offered, number of pills for sale and price, and information on contraindications, physical effects and complications described) using a predesigned survey in SurveyCTO.
We identified duplicates by comparing the contact information provided on websites or social media profiles/messages, usually a WhatsApp number. Just over a third (38%) of sellers had more than one site or profile. A little over half of the duplicates were within the same platform (56%), while the remaining 44% were across platforms (eg, website(s) plus social media profile(s)). It was common for sellers to have duplicate websites with different specific names of cities, as well as multiple WhatsApp numbers posted on the same website/profile. For websites which contained the same WhatsApp number, the first website identified chronologically by date of search was retained. For websites and social media pages found to have duplicate WhatsApp numbers, the page with the most detail was retained, usually a website. After deduplication, we identified 65 unique sellers .
The fourth stage of the research comprised four steps: (i) contacting each seller twice using two different mystery client profiles with different gestational ages; (ii) recording the details of each contact attempt and conversation including instructions offered; (iii) purchasing the medication offered and (iv) documenting the medications received by mail. Each step is described in more detail below. Three individuals who identified as cisgender women were trained as mystery clients who would pose as pregnant people attempting to purchase pills online for medication abortion. Using the contact information from the deduplicated list of sellers, the mystery clients reached out to the sellers on study tablets using SIM cards purchased specifically for this study. Mystery clients were able to use two SIM cards simultaneously, maintaining a conversation with five sellers per SIM card. Conversations with sellers lasted 2–3 days on average, and mystery clients changed SIM cards after conversations with five sellers were complete. If a duplicate seller was undetected, the rotation of phone numbers was intended to prevent a seller from becoming suspicious of receiving similar inquiries from the same phone number. Mystery clients did not obtain verbal or written consent from the online sellers to participate in the study, as that would have alerted the online sellers to the study, potentially changing their behaviours or responses, and introduced bias to the findings from the conversations. All research team members, including the people trained as mystery clients, signed a confidentiality agreement not to disclose any research information outside of the study team. Once the seller responded to the mystery client’s initial message, the mystery client engaged in a conversation based on the content of a predesigned survey in SurveyCTO. This survey included questions about how to use the medications and what to expect during the abortion; the mystery clients could ask for clarification from the sellers about the information the sellers provided. All contact took place over text message; mystery clients never spoke with the sellers via audio. Mystery clients documented any information requested or received from the sellers regarding pregnancy confirmation and gestational age, the types of pills offered and their prices, instructions on how to use the pills, including initial dosage amounts, repeat doses and administration route(s), information about expected physical effects, possible complications and any other instructions the seller provided about what to expect or do during the process of taking the pills. Between them, the three mystery clients were instructed to contact each seller using two separate profiles: once with a profile of a person with a pregnancy at 8 weeks gestational age and once with a profile of a person with a pregnancy at 16 weeks gestational age. Mystery client profiles included that they were 24 years of age, single, studying at university, experiencing their first pregnancy, and did not have any medical issues or contraindications (though this information was only provided to sellers if asked). Four sellers were mistakenly contacted twice with the same gestational age profile, so we retained the first contact with these sellers. Additionally, as fieldwork progressed, mystery clients did not always update the date of their last menstrual period (LMP) to align with the specified gestational age they were meant to be presenting in their profile. In about half of the 41 conversations in which the mystery clients used the first trimester profile and attempted a purchase, they provided a gestational age between 10 and 12 weeks, while use of the second trimester profile was more consistent. Therefore, for accuracy in the analysis, we report the gestational age categories as 8–12 weeks and 16–17 weeks. The survey was not meant to be filled out in any specific order by the mystery clients, as we could not predict the order in which sellers would discuss survey items. For example, some sellers told the mystery clients they would only send instructions on how to use the pills once a purchase was completed or the pills were received. In these cases, the mystery clients went back and filled in the information after purchase or receipt. Sellers wrote out instructions about expected physical effects or complications in messages to mystery clients and provided links to third-party websites and videos as well. Images, videos, and website links with information about the abortion process from third-party sources (most of which were in Spanish) were generally counted as information provided to the client. The exceptions were cases where the seller explicitly said that nothing would occur (eg, the woman would experience no physical effects or complications as a result of taking the pills) but also sent information from third-party sources. For these, we considered the seller as not providing information. Mystery clients were instructed not to continue conversations with sellers after purchase or delivery of the pills, except when sellers said they would send instructions only after selling or shipping the pills. Some sellers offered accompaniment services which, as noted above, are generally characterised by provision of medications, information and support before, during and after the abortion process, especially by trained feminist activists guided by a focus on women’s own decision-making power and autonomy. We instructed mystery clients not to accept any offer of accompaniment because they did not intend to actually use the pills or go through the abortion process. Mystery clients took screenshots of the full conversation with each seller on the study tablets assigned to them and saved the complete exchanges on a secure shared drive. Once the fieldwork phase was complete, Oriéntame’s technical team restored the tablets to their factory settings, deleting any saved files. Mystery clients were trained to go through with making a purchase of the pills if the seller offered up to a ceiling of Colombian peso (COP) $100 000 (US$24.53) for four pills, COP$120 000 (US$29.44) for six pills, COP$150 000 (US$36.79) for eight pills and COP$200 000 (US$49.06) for 12 pills. (Conversions to US$ used the median exchange rate for the fieldwork period of 23 July to 6 September 2023 of 4076.68. ) These prices were the average prices for the specified number of pills from a pretest conducted in July 2023. Partway through the study, we raised the maximum accepted price for each of the pill amounts by COP$15 000 (US$3.68) after determining that the prices specified by social media sellers were higher than the prices specified by website sellers. We also made exceptions to surpass the budget limit in three cases: when one seller offered a treatment of 14 misoprostol pills and when two sellers offered packages with pills other than misoprostol, including mifepristone, painkillers or antibiotics. Mystery clients were instructed to negotiate the price only once, when the price of the medications offered by the seller was higher than our prespecified budget limit. The pills were paid for using five different bank accounts in Colombia. Using this payment strategy avoided possible detection by sellers if using one single bank account or popular cash transfer application like Nequi or Daviplata, which display identifiable information like bank account numbers and names. To protect the mystery clients and avoid arousing suspicion if additional duplicate sellers existed and were contacted, the study team provided 10 different physical addresses in two different cities (Bogotá and Medellín) at which to receive pills. When a purchase was made, all sellers but one informed mystery clients when the package had been shipped and the expected time of arrival. A tracking number was given by the seller only in cases when the package was shipped from a different city via certified mail. When packages were delivered to one of the study addresses, they were forwarded to the first author (DA) who retrieved and opened each package. He then recorded the contents of each package and any additional instructions for use that were sent in a preprogrammed SurveyCTO form. Specifically, he recorded what pill brand was received (if visible), how many pills arrived and the condition of the packaging (eg, if a blister package was punctured or opened). Abortion pills acquired in this study were disposed of in accordance with Fundación Oriéntame’s guidelines for this process, that follow Colombia’s hazardous waste management policy for disposal of unused medications. Analysis First, responses to the three surveys (the listing survey describing the seller’s website/social media profile, the mystery client survey with the details of the seller contact attempt and the package survey documenting the content of packages received) were merged using the prespecified anonymised seller ID. Most of the instructions the sellers provided on physical effects, complications and how to take the pills were recorded as free text in the mystery client survey due to the diversity of information provided. This free text was translated from Spanish to English and then quantitatively coded into descriptive categories. We present descriptive statistics on the conversations by gestational age profile, including the frequency that sellers screened for pregnancy confirmation, gestational age, and contraindications, and mentioned physical effects and possible complications. We then report if pills were received, their condition and what information sellers gave to mystery clients on dosages and administration routes, by gestational age. Finally, we summarise how often sellers provided instructions to clients on other aspects of using the pills: pain medication, using the bathroom, eating and drinking specific foods, and physical activity. All analyses were conducted using Stata V.18 (College Station, Texas, USA).
First, responses to the three surveys (the listing survey describing the seller’s website/social media profile, the mystery client survey with the details of the seller contact attempt and the package survey documenting the content of packages received) were merged using the prespecified anonymised seller ID. Most of the instructions the sellers provided on physical effects, complications and how to take the pills were recorded as free text in the mystery client survey due to the diversity of information provided. This free text was translated from Spanish to English and then quantitatively coded into descriptive categories. We present descriptive statistics on the conversations by gestational age profile, including the frequency that sellers screened for pregnancy confirmation, gestational age, and contraindications, and mentioned physical effects and possible complications. We then report if pills were received, their condition and what information sellers gave to mystery clients on dosages and administration routes, by gestational age. Finally, we summarise how often sellers provided instructions to clients on other aspects of using the pills: pain medication, using the bathroom, eating and drinking specific foods, and physical activity. All analyses were conducted using Stata V.18 (College Station, Texas, USA).
Universe of online sellers and responses to initial contacts Of the original 161 online sellers identified, we intended to contact each of the 65 unduplicated, unique sellers twice, for a total of 130 contact attempts . The majority of these sellers were from websites or Facebook profiles. Ultimately, mystery clients made 122 initial contacts with sellers (61 with the 8–12 weeks gestational age profile; 61 with the 16–17 weeks gestational age profile). Mystery clients were unable to engage in conversation when the site/profile was no longer live, the seller was identified as a duplicate prior to initial contact, the contact information was not valid, the seller did not respond to the initial message, or the site provided information only and did not sell pills (n=35). Nine conversations were cut short before discussing a purchase. For clients with the 8–12 weeks profile, one seller asked to meet in person and two stopped responding. For clients with the 16–17 weeks profile, one seller asked to meet in person, one stopped responding, and four told the mystery clients that they were too far along in their pregnancies to provide them with pills (all four of these sellers reported 12 weeks as their maximum). Seller rapport with mystery clients For the mystery clients who were able to initiate contact with sellers (n=87), the conversations generally had an informal tone. Sellers often used slang and emojis in their messages and about one-third of sellers referred to the mystery clients with pet names such as nena (babe), linda (beauty), reina (queen) and amor (love), which are commonly used in casual conversations, but not appropriate in customer service, sales or medical encounters. About one-quarter of sellers offered “accompaniment” via WhatsApp throughout the abortion process (which was not accepted by the mystery clients), and this was slightly more common in conversations where mystery clients used the 16–17 weeks profile than the 8–12 weeks profile (25.6% vs 20.5%; not shown). Twelve sellers (13.8%) sent screenshots of conversations with other customers, which contained information about the physical effects those customers experienced, in addition to descriptions of the success that those customers had in terminating their pregnancies using the sellers’ pill products. Seven of these 12 sellers shared personal or identifying information of previous customers through the images, including names, profile pictures, phone numbers, banking account numbers, and addresses. Screening and information provided about physical effects and complications In the contact attempts where the mystery client initiated a conversation with the seller (n=87), communication most commonly commenced with sellers attempting to screen the mystery clients for pregnancy confirmation and gestational age . Few sellers asked if the mystery client had taken a pregnancy test (11.5%); rather, most attempted to screen for gestational age, either asking for the date of the last menstrual period (92.0%) or how many weeks pregnant the clients were (57.5%). Almost a quarter of sellers (24.1%) spontaneously mentioned that having an ectopic pregnancy was a contraindication for using medication abortion pills and a smaller proportion flagged an allergy to misoprostol or other medications (14.9%) as another reason not to take them. Five sellers incorrectly mentioned conditions such as anaemia as contraindications for use, while none mentioned conditions such as haemorrhagic disorder (not shown). Sellers with whom the conversation progressed to the point of the mystery client attempting to make a purchase (n=78; with an overlap of 30 sellers from whom a purchase attempt was made with both profiles) mentioned a range of physical effects the mystery client should expect when using medication abortion pills. Bleeding and menstrual-like cramping were most frequently mentioned for both the 8–12 weeks and 16–17 weeks profiles . One-quarter of sellers specified that both strong cramps and bleeding that is heavier than menstruation (possibly with clots of blood) were normal and expected physical effects of the abortion process, while another 15.4% described symptoms as being similar to those of a normal period (not shown). Other physical effects, including vomiting, nausea and mild fever, were mentioned by less than half of sellers. In general, all physical effects were mentioned more frequently for the 8–12 weeks profile, except for passing products of conception and pain, which were both mentioned in 40.5% of 16–17 weeks conversations vs 36.6% and 22.0% of 8–12 weeks conversations, respectively. About 8% of sellers did not give any information on any physical effects. Overall, <20% of sellers mentioned any complication of medication abortion for which the mystery clients should seek medical attention. Sellers mentioned ways to determine the abortion was complete during 85.4% of attempted purchase conversations with the 8–12 weeks profile, compared with 73.0% of conversations with the 16–17 weeks profile (not shown). The most common methods sellers suggested for determining a complete abortion were ultrasound (69.9%), expulsion of the products of conception (24.6%) and taking a pregnancy test (14.5%). Pills purchased and condition of pills received Mystery clients made 45 purchases: 27 with the 8–12 weeks gestational age profile and 18 with the 16–17 weeks gestational age profile. These purchases were from 35 unique sellers; two purchases were made from from 10 sellers. The median prices paid, including any shipping or delivery costs, were COP$150 000 (US$36.79) (8–12 weeks profiles) and COP$200 000 (US$49.06) (16–17 weeks profiles) (not shown). Price negotiations that mystery clients attempted resulted in one of three outcomes: (1) the seller refused to reduce their price and the conversation ended; (2) the seller agreed to reduce their price for the same number of pills that they originally offered; or (3) the seller reduced both the price and the number of pills that they were willing to offer. Among packages purchased, mystery clients were able to obtain discounts between COP$5000 and COP$30 000 (not shown). The most common reason for not completing a purchase was that the price the seller offered was higher than the budget limit the study team had at the time of the conversation (n=28). In four instances (two for each profile), mystery clients did not purchase pills because the seller told them that the number of pills that they could afford would not work compared with what the seller had offered them for their gestational age (6 and 8 pills for the 8–12 weeks profiles and 12 and 14 pills for the 16–17 weeks profiles, respectively). Ultimately, we received 40 packages containing pills (88.9% of total purchases): 23 packages with the 8–12 weeks profile (equivalent to 37.7% of initial contacts with this profile) and 17 packages with the 16–17 weeks profile (27.9% of initial contacts). One additional seller sent a package, but it contained a pair of shorts and a T-shirt instead of pills (a possible mix-up). Four other packages for which payment for medication abortion pills was sent were never received. When the mystery clients followed up with the four sellers about the status of the missing deliveries, they either did not respond or attributed the delays to shipping mistakes or issues with the package carriers. None of these sellers sent a second package of pills. Nine of the packages were delivered on the same day that the medication was ordered and overall, three-quarters of packages received were delivered within 2 days of initiating contact (not shown). Of the 40 packages of medication received, all contained Cytotec brand 200 μg misoprostol pills in aluminium blister packs labelled with Pfizer manufacturing information and sanitary registry approval from Ecuador and Peru (Cytotec is not registered in Colombia). Misoprostol pills in all the blister packs were stamped with the 1461 pharmaceutical code for Cytotec. In addition to Cytotec, two of the 40 sellers also sent painkillers, antibiotics and unspecified pills that we believed to be mifepristone, since both of those sellers had offered mifepristone to the mystery client as part of the treatment. While over 77% of all packages included misoprostol in a blister pack with no signs of damage, this was much more common for the 16–17 weeks purchases (94.1%) than the 8–12 weeks purchases (65.2%) . The other blisters had the foil around one pill carefully cut (the sellers sometimes sent pictures of the pills to the mystery clients) or in two cases, the sellers sent an undamaged blister along with additional pills wrapped in a paper napkin. Expiration dates were visible on 60% of the blisters (n=24) and all visible dates were in the future (2024, 2025 or 2026). Blister packs appeared to have been cut to send the specific number of pills that the clients and sellers agreed on, with some expiration dates removed as a result. Instructions for use provided to those who received pills The number of misoprostol pills most frequently received for the 8–12 weeks profiles was eight (69.6%), compared with 12 pills (76.5%) for the 16–17 weeks profiles . Over 80% of the 23 sellers interacting with mystery clients using the 8–12 weeks profile instructed them to take an initial dose of misoprostol that was in line with the Colombia MOH guidelines for first trimester medication abortion (800 μg or four pills). Thirteen per cent gave instructions for more than that amount and only one seller gave instructions below the recommended amount. All 17 sellers that sold to mystery clients using the 16–17 weeks profile instructed them to take an initial dose that was above the MOH-recommended amount of 400 μg (two pills) for second trimester medication abortion. All but three sellers from whom packages were received directed the mystery clients to take repeated doses of misoprostol (92.5%; n=37). The most common instruction was two doses of four pills for the 8–12 weeks profiles and three doses of four pills for the 16–17 weeks profiles. The majority of sellers (80.0%) advised the mystery clients to use MOH-recommended routes of administration of the pills (vaginal, sublingual and/or buccal). However, sellers more often mentioned non-recommended routes of administration (oral) for the 16–17 weeks profiles (23.5% compared with 17.4% for the 8–12 weeks profiles). About a third of sellers from whom packages were received provided contradictory information about the instructions for use or the maximum gestational ages at which medication abortion pills could be taken. Ultimately, just over half of the 40 packages that were purchased and received included manufacturer-labelled, non-damaged packaging, along with instructions to take at least the minimum recommended initial dose using recommended administration route(s). In their instructions to mystery clients about the abortion process, many sellers also included recommendations that were not necessary for the abortion to be successful and although not always harmful, could result in additional and unnecessary pain and physical distress for the person having an abortion. For example, while over half of sellers told the mystery clients to take pain medications during their abortions, 12.5% explicitly told them not to take pain medications, stating that taking pain medications could impact the efficacy of the abortion pills . Over half of sellers also gave instructions for the clients not to use the bathroom or eat or drink for varying numbers of hours during the abortion process (which can often last several days). Sellers also told clients specific foods to eat or drink (eg, light soups, orange juice, rue tea) or to specifically avoid (eg, citrus, dairy, alcohol). Fifteen per cent of sellers told clients to take antibiotics during the abortion (not shown), even though antibiotics are not routinely recommended for medication abortions.
Of the original 161 online sellers identified, we intended to contact each of the 65 unduplicated, unique sellers twice, for a total of 130 contact attempts . The majority of these sellers were from websites or Facebook profiles. Ultimately, mystery clients made 122 initial contacts with sellers (61 with the 8–12 weeks gestational age profile; 61 with the 16–17 weeks gestational age profile). Mystery clients were unable to engage in conversation when the site/profile was no longer live, the seller was identified as a duplicate prior to initial contact, the contact information was not valid, the seller did not respond to the initial message, or the site provided information only and did not sell pills (n=35). Nine conversations were cut short before discussing a purchase. For clients with the 8–12 weeks profile, one seller asked to meet in person and two stopped responding. For clients with the 16–17 weeks profile, one seller asked to meet in person, one stopped responding, and four told the mystery clients that they were too far along in their pregnancies to provide them with pills (all four of these sellers reported 12 weeks as their maximum).
For the mystery clients who were able to initiate contact with sellers (n=87), the conversations generally had an informal tone. Sellers often used slang and emojis in their messages and about one-third of sellers referred to the mystery clients with pet names such as nena (babe), linda (beauty), reina (queen) and amor (love), which are commonly used in casual conversations, but not appropriate in customer service, sales or medical encounters. About one-quarter of sellers offered “accompaniment” via WhatsApp throughout the abortion process (which was not accepted by the mystery clients), and this was slightly more common in conversations where mystery clients used the 16–17 weeks profile than the 8–12 weeks profile (25.6% vs 20.5%; not shown). Twelve sellers (13.8%) sent screenshots of conversations with other customers, which contained information about the physical effects those customers experienced, in addition to descriptions of the success that those customers had in terminating their pregnancies using the sellers’ pill products. Seven of these 12 sellers shared personal or identifying information of previous customers through the images, including names, profile pictures, phone numbers, banking account numbers, and addresses.
In the contact attempts where the mystery client initiated a conversation with the seller (n=87), communication most commonly commenced with sellers attempting to screen the mystery clients for pregnancy confirmation and gestational age . Few sellers asked if the mystery client had taken a pregnancy test (11.5%); rather, most attempted to screen for gestational age, either asking for the date of the last menstrual period (92.0%) or how many weeks pregnant the clients were (57.5%). Almost a quarter of sellers (24.1%) spontaneously mentioned that having an ectopic pregnancy was a contraindication for using medication abortion pills and a smaller proportion flagged an allergy to misoprostol or other medications (14.9%) as another reason not to take them. Five sellers incorrectly mentioned conditions such as anaemia as contraindications for use, while none mentioned conditions such as haemorrhagic disorder (not shown). Sellers with whom the conversation progressed to the point of the mystery client attempting to make a purchase (n=78; with an overlap of 30 sellers from whom a purchase attempt was made with both profiles) mentioned a range of physical effects the mystery client should expect when using medication abortion pills. Bleeding and menstrual-like cramping were most frequently mentioned for both the 8–12 weeks and 16–17 weeks profiles . One-quarter of sellers specified that both strong cramps and bleeding that is heavier than menstruation (possibly with clots of blood) were normal and expected physical effects of the abortion process, while another 15.4% described symptoms as being similar to those of a normal period (not shown). Other physical effects, including vomiting, nausea and mild fever, were mentioned by less than half of sellers. In general, all physical effects were mentioned more frequently for the 8–12 weeks profile, except for passing products of conception and pain, which were both mentioned in 40.5% of 16–17 weeks conversations vs 36.6% and 22.0% of 8–12 weeks conversations, respectively. About 8% of sellers did not give any information on any physical effects. Overall, <20% of sellers mentioned any complication of medication abortion for which the mystery clients should seek medical attention. Sellers mentioned ways to determine the abortion was complete during 85.4% of attempted purchase conversations with the 8–12 weeks profile, compared with 73.0% of conversations with the 16–17 weeks profile (not shown). The most common methods sellers suggested for determining a complete abortion were ultrasound (69.9%), expulsion of the products of conception (24.6%) and taking a pregnancy test (14.5%).
Mystery clients made 45 purchases: 27 with the 8–12 weeks gestational age profile and 18 with the 16–17 weeks gestational age profile. These purchases were from 35 unique sellers; two purchases were made from from 10 sellers. The median prices paid, including any shipping or delivery costs, were COP$150 000 (US$36.79) (8–12 weeks profiles) and COP$200 000 (US$49.06) (16–17 weeks profiles) (not shown). Price negotiations that mystery clients attempted resulted in one of three outcomes: (1) the seller refused to reduce their price and the conversation ended; (2) the seller agreed to reduce their price for the same number of pills that they originally offered; or (3) the seller reduced both the price and the number of pills that they were willing to offer. Among packages purchased, mystery clients were able to obtain discounts between COP$5000 and COP$30 000 (not shown). The most common reason for not completing a purchase was that the price the seller offered was higher than the budget limit the study team had at the time of the conversation (n=28). In four instances (two for each profile), mystery clients did not purchase pills because the seller told them that the number of pills that they could afford would not work compared with what the seller had offered them for their gestational age (6 and 8 pills for the 8–12 weeks profiles and 12 and 14 pills for the 16–17 weeks profiles, respectively). Ultimately, we received 40 packages containing pills (88.9% of total purchases): 23 packages with the 8–12 weeks profile (equivalent to 37.7% of initial contacts with this profile) and 17 packages with the 16–17 weeks profile (27.9% of initial contacts). One additional seller sent a package, but it contained a pair of shorts and a T-shirt instead of pills (a possible mix-up). Four other packages for which payment for medication abortion pills was sent were never received. When the mystery clients followed up with the four sellers about the status of the missing deliveries, they either did not respond or attributed the delays to shipping mistakes or issues with the package carriers. None of these sellers sent a second package of pills. Nine of the packages were delivered on the same day that the medication was ordered and overall, three-quarters of packages received were delivered within 2 days of initiating contact (not shown). Of the 40 packages of medication received, all contained Cytotec brand 200 μg misoprostol pills in aluminium blister packs labelled with Pfizer manufacturing information and sanitary registry approval from Ecuador and Peru (Cytotec is not registered in Colombia). Misoprostol pills in all the blister packs were stamped with the 1461 pharmaceutical code for Cytotec. In addition to Cytotec, two of the 40 sellers also sent painkillers, antibiotics and unspecified pills that we believed to be mifepristone, since both of those sellers had offered mifepristone to the mystery client as part of the treatment. While over 77% of all packages included misoprostol in a blister pack with no signs of damage, this was much more common for the 16–17 weeks purchases (94.1%) than the 8–12 weeks purchases (65.2%) . The other blisters had the foil around one pill carefully cut (the sellers sometimes sent pictures of the pills to the mystery clients) or in two cases, the sellers sent an undamaged blister along with additional pills wrapped in a paper napkin. Expiration dates were visible on 60% of the blisters (n=24) and all visible dates were in the future (2024, 2025 or 2026). Blister packs appeared to have been cut to send the specific number of pills that the clients and sellers agreed on, with some expiration dates removed as a result.
The number of misoprostol pills most frequently received for the 8–12 weeks profiles was eight (69.6%), compared with 12 pills (76.5%) for the 16–17 weeks profiles . Over 80% of the 23 sellers interacting with mystery clients using the 8–12 weeks profile instructed them to take an initial dose of misoprostol that was in line with the Colombia MOH guidelines for first trimester medication abortion (800 μg or four pills). Thirteen per cent gave instructions for more than that amount and only one seller gave instructions below the recommended amount. All 17 sellers that sold to mystery clients using the 16–17 weeks profile instructed them to take an initial dose that was above the MOH-recommended amount of 400 μg (two pills) for second trimester medication abortion. All but three sellers from whom packages were received directed the mystery clients to take repeated doses of misoprostol (92.5%; n=37). The most common instruction was two doses of four pills for the 8–12 weeks profiles and three doses of four pills for the 16–17 weeks profiles. The majority of sellers (80.0%) advised the mystery clients to use MOH-recommended routes of administration of the pills (vaginal, sublingual and/or buccal). However, sellers more often mentioned non-recommended routes of administration (oral) for the 16–17 weeks profiles (23.5% compared with 17.4% for the 8–12 weeks profiles). About a third of sellers from whom packages were received provided contradictory information about the instructions for use or the maximum gestational ages at which medication abortion pills could be taken. Ultimately, just over half of the 40 packages that were purchased and received included manufacturer-labelled, non-damaged packaging, along with instructions to take at least the minimum recommended initial dose using recommended administration route(s). In their instructions to mystery clients about the abortion process, many sellers also included recommendations that were not necessary for the abortion to be successful and although not always harmful, could result in additional and unnecessary pain and physical distress for the person having an abortion. For example, while over half of sellers told the mystery clients to take pain medications during their abortions, 12.5% explicitly told them not to take pain medications, stating that taking pain medications could impact the efficacy of the abortion pills . Over half of sellers also gave instructions for the clients not to use the bathroom or eat or drink for varying numbers of hours during the abortion process (which can often last several days). Sellers also told clients specific foods to eat or drink (eg, light soups, orange juice, rue tea) or to specifically avoid (eg, citrus, dairy, alcohol). Fifteen per cent of sellers told clients to take antibiotics during the abortion (not shown), even though antibiotics are not routinely recommended for medication abortions.
This study is the first study in Colombia specifically and Latin America overall to examine medication abortion information provided by online sellers to clients and successful receipt by clients of medication abortion pills purchased online. Even though recent legal changes and MOH guidelines have outlined the processes for accessing abortion at health facilities, this study confirmed that medication abortion pills are still available for purchase from online sellers. This implies that sellers are meeting a demand that persists in Colombia. While people may choose to purchase from these sellers because they are perceived as more convenient and private in comparison with health facilities, their practices are not regulated by MOH guidelines. For mystery clients who were able to make purchases in this study, the majority of packages did arrive containing both the minimum necessary initial dose and often repeat doses of misoprostol for use of medication abortion as outlined in the Colombia MOH guidelines. However, particularly for mystery clients who contacted sellers with a second trimester (16–17 weeks) profile, the initial dose they were instructed to take was higher than the recommended amount. In addition, the instructions provided by sellers lacked information on contraindications, expected physical effects, potential complications, and how to determine that the abortion was complete, which are all recommended components of quality abortion care. For the 8–12 weeks profile contacts that resulted in completed purchases and receipt of the packages, most sellers instructed mystery clients to take an initial dose for first trimester abortions that was in line with Colombia’s MOH guidelines for pregnancies up to 12 weeks (four pills or 800 μg), and additionally sold and sent enough pills for repeat doses of misoprostol. There was only one seller who directed the mystery client to take less than the recommended initial dose of misoprostol for their gestational age. This would be a concern for an actual person seeking abortion, as it puts the individual at risk of continuing pregnancy. Compared with a similar quality of care assessment of online medication abortion pill sellers in Indonesia and sales at pharmacies without prescriptions in Mexico, online sellers in Colombia were much more likely to provide enough misoprostol, as well as provide instructions about the correct initial dosages. The large majority of 8–12 weeks purchases were of 8 or 12 pills of misoprostol, which is sufficient for repeat doses in line with Colombia’s MOH guidelines, but not all sellers advised on repeat doses. Additionally, just over one-fifth of sellers provided the number of pills and instructions in line with international guidelines that recommend three doses of 800 μg misoprostol for first trimester abortions. Based on this finding, it is likely that some real clients would need additional doses to complete their abortions, necessitating repeating the online purchasing process or seeking follow-up care at health facilities. For the 16–17 weeks profiles, sellers overwhelmingly sold and provided instructions for the use of more pills (four pills or 800 μg) for the initial dose than the two pills or 400 μg recommended by Colombia’s MOH (along with repeated doses). While this higher initial dose is not generally harmful, it may make the abortion experience additionally uncomfortable by inducing more physical symptoms. In addition, advising clients that more pills are needed overall comes with additional expense. Unfortunately, sellers did not usually provide details in their instructions about when repeat doses might be required or how to assess completeness of the abortion without ultrasound; rather, they recommended taking multiple doses to almost every mystery client. More accurate instructions about how people managing their abortion process should assess their need for repeat doses of misoprostol should be provided (eg, if no bleeding occurred), in line with MOH guidelines to take repeat doses as needed until pregnancy expulsion. The most common reason the mystery clients were not able to attempt a purchase was that the quoted price from the sellers was too high based on our study’s budget limit. As of 2023, the minimum monthly wage in Colombia was COP$1.16 million (US$276.61) ; however, 39.2% of the population lives below the national poverty line. Obtaining abortion pills from online sellers is likely cost-prohibitive for many individuals who cannot afford a large, unexpected expense that is almost 20% of the monthly minimum wage. Interestingly, a few sellers did not sell because they said the number of pills the clients could afford would not result in a complete abortion, when in many of the cases, the number would have been sufficient. In addition, when mifepristone was infrequently advertised, it was generally cost-prohibitive to purchase. This may be a result of its restricted availability, which makes it very difficult for non-clinicians to obtain and sell. A small but noteworthy percentage of purchases made during this study were also never delivered (10.6%) (a similar rate to online misoprostol sales in Indonesia), and real pregnant people may be unable to afford a second purchasing attempt. Abortion care is a mandatory health service in the public healthcare system and is available at no cost to those using health insurance or otherwise seeking care in public facilities. However, data from abortion patients served by Oriéntame in the year following the 2022 legal changes found that only 34% of abortion patients used their health insurance benefits to cover costs, 10% received subsidies from Oriéntame and 56% paid out of pocket. Bureaucratic delays from health insurance plans to access abortion care or a desire to avoid having abortion mentioned in their medical records may lead some people to choose to pay out-of-pocket. Online sellers often tried to present themselves in a more casual and friendly manner and not like health facility providers, which may be a reaction to women’s lack of trust in the healthcare system for abortion care. However, people’s privacy is a particular concern as online sellers not infrequently shared screenshots of other customers’ personal and private information. While sellers shared a variety of information and resources, the instructions they provided about how to take the misoprostol pills, contraindications, and possible physical effects and complications varied in accuracy and completeness. More than three-quarters of the sellers from whom pills were received directed mystery clients to use the misoprostol through recommended routes (vaginally, buccally and/or sublingually) and some also suggested lying down for short periods of time after inserting pills vaginally or taking analgesics to manage pain, in line with WHO guidelines for quality abortion care. However, few sellers mentioned possible contraindications or complications associated with medication abortion or what to do if clients experienced them. Avoiding the mention of complications, even if they are uncommon, might be a strategy sellers use to ensure a sale, or it may also be the case that sellers themselves do not have accurate information about the potential warning signs. Either way, the average person’s knowledge of warning signs that indicate a need to seek follow-up care are likely to still be low, especially for vulnerable populations, including younger and less educated people, as well as Venezuelan migrants who are more likely to be unfamiliar with Colombia’s abortion law and health system. Other complicated and contradictory instructions about diet, physical activity and using the toilet given to over half of the mystery clients who received pills were not necessarily harmful but could make the abortion process more cumbersome and uncomfortable than needed (similar complicated and unnecessary instructions were also mentioned by online sellers in the previous Indonesia study). While previous qualitative research in Colombia had implied that online sellers of abortion pills gave better instructions and information than other sources operating outside of the medical system or guidelines, we found them lacking in some areas, especially compared with recommended guidelines for provision of quality abortion care. Based on the results from this study, sellers did not consistently provide clients with information on accurate dosage amounts (especially for second trimester abortions), the conditions under which repeat doses of misoprostol are needed, and what physical effects and complications to expect and how to manage them. International websites that provided abortion information did come up in our website search and some sellers referred clients to external international sources of reliable information such as Ipas Mexico and safe2choose. Nonetheless, these findings highlight the need for compiling updated, medically accurate information about what to expect when having a medication abortion into one centralised, easy-to-locate online source that is Colombia-specific and disseminating it widely so that Colombians seeking self-managed abortion have clarity about what to expect. Accompaniment groups may also have a role to play in connecting with people ordering abortion pills online and providing them additional information that online sellers lack. In Colombia, there are several hotlines and feminist groups that provide instructions and reliable information on how to perform a medication abortion, as well as accompaniment and emotional support during the process. A quarter of sellers offered “accompaniment”, although we were unable to assess the quality of this support. Additionally, it was not the goal of this project to search for patient-facing materials or assess the quality of information available or being used by Colombian abortion patients. Therefore, more research is needed to determine what sources people ordering pills online are relying on for understanding the abortion process. Strengths and limitations The mystery client design of this study allowed us to simulate real customer experiences of pregnant people seeking medication abortion pills. This research is distinct from other studies of the online market of abortion medications in the USA and Indonesia (more restrictive environments than currently exists in Colombia ) and it provides insight into how online sellers behave in a country with greater legal access to abortion. Despite its strengths, this study also had several limitations. The market of online abortion pill sellers is constantly in flux. As a result, this study is only able to present a snapshot in time and is not a comprehensive picture of every online seller. We used a variety of search terms informed by Google Trends data, but the research team determined the keywords for the search, rather than asking potential clients what terms they would use. Even though we performed deduplication, the study team suspected that there may have been additional duplicates based on some similar sources of information and bank accounts shared with clients. We are not able to determine whether sellers are working together, independently, or where they obtain the information they share, and we are not able to comment on the qualifications of the sellers or where or how they are getting the pills they are selling. While purchasing medication abortion pills without a prescription and self-managing an abortion is not mentioned in the penal code, it is neither permitted to sell registered drugs without a prescription, nor to sell medicines that are not formally registered, which is the case of the Cytotec brand in Colombia. Therefore, we cannot comment on the legal risk the sellers are willing to take on and how that informs their advertising or behaviour with clients. While we initially instructed mystery clients to report gestational ages of either 8 or 16 weeks to sellers, during analysis we determined greater variation in dates of last menstrual period provided, especially for those intended 8-week profile conversations. Therefore, we classified the profiles according to ranges (8–12 and 16–17 weeks gestational age). This deviation in gestational ages could also have affected the number of pills that mystery clients were offered, but since Colombia’s MOH recommends 800 μg of misoprostol for misoprostol-only regimens for all pregnancies under 12 weeks gestational age and 400 μg for pregnancies between 12 and 24 weeks (with repeated doses as necessary), we were able to consistently evaluate the sellers’ dosing adequacy according to MOH guidelines within these two gestational age groupings for all but two interactions that provided an LMP equivalent to 12 weeks of pregnancy. While the MOH guidelines were updated in early 2023, previous versions of the guidelines had recommended the same misoprostol-only dose could be used up to 13 weeks of pregnancy. Therefore, we may have slightly overstated the extent to which sellers provided the correct instructions for taking the pills between 8 and 12 weeks in line with the current guidelines. Due to budgetary constraints, we were not able to purchase pills from all identified sellers. This may impact our conclusions if the more expensive sellers were likely to provide a different number of pills and different information as compared with the sellers from whom we bought. In addition, we did not test the pills received to verify their authenticity; rather, we based our assessments of the drugs on the package labelling with manufacturer information and the shape and pharmaceutical code stamped on the pills.
The mystery client design of this study allowed us to simulate real customer experiences of pregnant people seeking medication abortion pills. This research is distinct from other studies of the online market of abortion medications in the USA and Indonesia (more restrictive environments than currently exists in Colombia ) and it provides insight into how online sellers behave in a country with greater legal access to abortion. Despite its strengths, this study also had several limitations. The market of online abortion pill sellers is constantly in flux. As a result, this study is only able to present a snapshot in time and is not a comprehensive picture of every online seller. We used a variety of search terms informed by Google Trends data, but the research team determined the keywords for the search, rather than asking potential clients what terms they would use. Even though we performed deduplication, the study team suspected that there may have been additional duplicates based on some similar sources of information and bank accounts shared with clients. We are not able to determine whether sellers are working together, independently, or where they obtain the information they share, and we are not able to comment on the qualifications of the sellers or where or how they are getting the pills they are selling. While purchasing medication abortion pills without a prescription and self-managing an abortion is not mentioned in the penal code, it is neither permitted to sell registered drugs without a prescription, nor to sell medicines that are not formally registered, which is the case of the Cytotec brand in Colombia. Therefore, we cannot comment on the legal risk the sellers are willing to take on and how that informs their advertising or behaviour with clients. While we initially instructed mystery clients to report gestational ages of either 8 or 16 weeks to sellers, during analysis we determined greater variation in dates of last menstrual period provided, especially for those intended 8-week profile conversations. Therefore, we classified the profiles according to ranges (8–12 and 16–17 weeks gestational age). This deviation in gestational ages could also have affected the number of pills that mystery clients were offered, but since Colombia’s MOH recommends 800 μg of misoprostol for misoprostol-only regimens for all pregnancies under 12 weeks gestational age and 400 μg for pregnancies between 12 and 24 weeks (with repeated doses as necessary), we were able to consistently evaluate the sellers’ dosing adequacy according to MOH guidelines within these two gestational age groupings for all but two interactions that provided an LMP equivalent to 12 weeks of pregnancy. While the MOH guidelines were updated in early 2023, previous versions of the guidelines had recommended the same misoprostol-only dose could be used up to 13 weeks of pregnancy. Therefore, we may have slightly overstated the extent to which sellers provided the correct instructions for taking the pills between 8 and 12 weeks in line with the current guidelines. Due to budgetary constraints, we were not able to purchase pills from all identified sellers. This may impact our conclusions if the more expensive sellers were likely to provide a different number of pills and different information as compared with the sellers from whom we bought. In addition, we did not test the pills received to verify their authenticity; rather, we based our assessments of the drugs on the package labelling with manufacturer information and the shape and pharmaceutical code stamped on the pills.
These results demonstrate that online sellers selling medication abortion without a prescription continue to operate as a source for medication abortion pills in Colombia. Now that Colombia has a new progressive legal framework for abortion access, additional research with people who have purchased abortion pills online could provide more information on barriers to access and how and why the online abortion pill market continues to be appealing for some women. Similarly, additional studies of people who opt for medication abortion through telemedicine could provide further insight into the motivations for avoiding in-person care in this context. Continuing to assess barriers to expanding telemedicine for medication abortion through the formal health system could also highlight policy or operational recommendations for implementing those services more broadly. Finally, additional research on what information sources are available and used by people purchasing abortion pills online and through other informal sources outside of the medical system about how to use the pills, what to expect, and contraindications for use could shed light on what information individuals are referencing. People in Colombia should have access to complete and accurate information to ensure that they are well positioned to self-manage their abortions if they choose to do so.
10.1136/bmjopen-2024-086404 online supplemental file 1
|
Surgeon’s perceptions and preferences in the management of idiopathic macular hole | f2ba1e48-c738-4f4a-9b37-8ea50731001f | 11834932 | Surgical Procedures, Operative[mh] | A cross-sectional observational study was performed constituting a survey addressing the current practice patterns of vitreoretinal surgeons in India on macular hole surgery. The institutional review board approval was obtained. The study adhered to the tenets of the Declaration of Helsinki. A 12-item questionnaire was designed in the English language, keeping in mind the key aspects of macular hole surgery where inter-surgeon variation is more likely to be present. Most of the questions were close-ended with options to choose and only a few questions required descriptive answers. All 12 questions were optional to answer. The questionnaire was sent by personal correspondence (email) to 104 vitreoretinal specialists, actively practicing and performing iMH surgeries at various institutes across the country between October 2022 to November 2022, who then returned their responses on the same platform. A reminder was sent to those who had not responded by the end of December 2022. The responses were received till January 2023, and the data were exported to an Excel sheet. Statistical analysis was performed using the SPSS 23.0 software. The response to individual questions was analyzed for the whole cohort. The analyses of variables with continuous data were performed as mean (with standard deviation) and median (with the range), while the categorical data were expressed as frequency. Ninety-one retina specialists responded to the survey with a response rate of 87.5% (91/104). The median surgical load of the surgeons was 30 iMH surgeries per year (range: 5–150), and 52.9% (45/85) of the respondents performed more than 20 surgeries per year. Furthermore, 80% (73/91) of the participants were affiliated to tertiary care academic institutions, and the rest were stand-alone practitioners. The majority of the surgeons (84.3%, 74/89) performed surgery even in elderly patients >75 years, one-eyed patients (78.7%, 70/89), and chronic cases of more than 1-year duration of iMH (71.6%, 63/88). However, 60% (53/89) of surgeons did not perform surgery in flat MH with complete PVD. Nearly all the surgeons (97.8%, 89/91) preferred brilliant blue G (BBG) dye for staining the ILM, while the remaining two respondents preferred trypan blue dye. In phakic eyes with visually insignificant cataracts, 34.1% (31/91) of surgeons would choose to perform combined cataract surgery with MH surgery, while the rest would prefer to perform cataract surgery on follow-up when required. The preferred instruments for initiation of ILM peel and creating a flap were ILM forceps (82.2%, 74/90), Finesse loop (11.1%, 10/90), membrane scrapper (4.4%, 4/90), or microvitreoretinal (MVR) blade (2.2%, 2/90). The preferred approach to peeling ILM was “pinch and peel” (80%, 72/90), followed by “scrape and peel” (15.6%, 14/90) and “incise and peel” (3.3%, 3/90). However, the preferred approach recommended for beginners or trainee surgeons was “scrape and peel” (50.6%, 40/79), followed by “pinch and peel” (40.5%, 32/79) and “incise and peel” (6.3%, 5/79). The preferred distance to initiate peel was 2 disc-diameter (DD) from the MH (52.9%, 46/87), followed by 1 DD (34.5%, 30/87). Furthermore, 12.6% of respondents (11/87) chose variable distance depending on the individual case profile. While the majority did not prefer any quadrant to initiate the peel, 43.7% (38/87) of respondents preferred a fixed area to initiate the peel. Of these 38 respondents, 18 preferred the inferotemporal quadrant to initiate the peel. Nearly all the respondents (97.1%, 68/70) initiated peel away from the blood vessels. The average number of attempts before achieving a satisfactory ILM fracture was 1–2 in 60.5% (52/86) and >2 in 39.5% (34/86) respondents. Most surgeons did not consider massaging the hole edges (80.7%, 67/82) or draining through the MH (75.9%, 63/83). The preferred vitreous substitute for tamponade was sulfur hexafluoride (SF6) gas (52.7%, 48/90), followed by perfluoropropane (C3F8) gas (41.8%, 38/90) and air (4.4%, 4/90). Postoperative prone position was recommended for 3–7 days by the majority (89%, 81/91), followed by 1 day (8.8%, 8/91) and 6 hours (2.2%, 2/91). Many respondents considered MH as large when the minimum linear dimension on OCT was >600 microns (42.2%, 38/90). In addition, 31.1% of surgeons (28/910) considered MH >800 microns as large, while only 18.9% (17/90) considered it large if the size was above 400 microns. Four surgeons considered >1000 microns as the definition for large MH. The preferred approaches in large MH were classic inverted ILM flap (48.9%, 44/90), multilayer flap (26.7%, 24/90), temporal flap (11.1%, 10/90), and traditional ILM peel (8.9%, 8/90). The other responses obtained were traditional ILM peel with massage of MH edge, temporal large rhexis, radial ILM peel, and multilayer flap with application of platelet-rich plasma (one respondent each). The approaches preferred in persistent iMH despite surgery with ILM peeling were free ILM flap (49.4%, 44/89), repeat fluid-gas exchange (12.4%, 11/89), autologous retinal graft (6.7%, 6/89), macular detachment and tamponade (3.4%, 3/89), amniotic membrane graft (AMG, 2.2%, 2/89), platelet rich plasma application (1), and temporal large rhexis (1). Four surgeons considered AMG, and one considered lens capsule transplant as an additional but not preferred option for failed surgery. Furthermore, 24.7% of surgeons (22/89) did not consider further intervention in failed cases. Lastly, the three most important prognostic factors for visual outcomes after MH surgery (in order of importance) as perceived by survey participants were the duration of MH (46.2%, 42/91), preoperative vision (40.4%, 36/91), and MH size (37.5%, 33/91). OCT indices, status of PVD, size of peel, type of tamponade, and compliance to positioning were not considered among the three most important prognostic factors by most surgeons. The treatment for idiopathic MH is vitrectomy, ILM peeling, and gas tamponade. However, there exist different views on individual aspects of the surgery, and this study assessed the current practices in India. The survey included responses from vitreoretinal surgeons with good surgical experience and workload. Nearly 85% of surgeons would operate even on elderly patients >75 years. MH surgery can lead to significant improvement in the quality of life in elderly patients. Therefore, old age may not be considered a criterion to recommend against MH surgery. In the current times, the improvements in microsurgical techniques have made vitrectomy and ILM peeling a safe procedure. This could be the reason that nearly 80% of the surgeons operated on one-eyed patients as well. Chronic iMH has been described variably as MH with a duration of more than 6 months, 1 year, or 2 years. The outcome studies in chronic iMH, which date back to the 1990s, have shown MH closure rates varying from 63% to 95%. While some authors have cautioned against surgery in chronic MH, most studies have reported an average of 2–3 Snellen line improvement. Chronic iMH may thus benefit from surgery, and some useful vision could be obtained, and this would be the reason why nearly three-fourths of the respondents in the survey would operate on chronic iMH cases. Almost all surgeons used BBG dye for staining the ILM. Compared to indocyanine green, newer dyes such as BBG, trypan blue, or a combination of these with polyethylene glycol or deuterium are much safer. While trypan blue also stains epiretinal membranes, BBG is specific for ILM and remains the first choice for ILM staining. Vitrectomy for MH increases the risk of development and progression of cataract. Only one-third of surgeons performed phaco-vitrectomy in eyes with insignificant cataract. While a combined phaco-vitrectomy will avoid a second surgery, the combined approach may have a higher risk of intraoperative complications such as corneal edema, poor visualization, and iatrogenic retinal injury and postoperative issues such as posterior synechiae formation, high intraocular pressure, unpredictable refractive outcomes, and MH non-closure. A recently performed systematic review and meta-analysis found no significant difference in the outcomes and complications between combined and sequential surgery groups. However, most studies included in the meta-analysis were retrospective or low-moderate quality trials. ILM peel can be initiated by forming a flap with the help of forceps (pinch technique), MVR blade, pick, diamond-dusted membrane scraper, or serrated nitinol loop (Finesse loop). The degree of dissociated optic nerve fiber layer appearance with scrapping ILM is greater than after pinching it. Most of the surgeons in the survey used ILM forceps, and only a few preferred loop or scraper. However, many of them advocated the “scrape-and-peel” technique for beginners as experience and dexterity in hand control are required to avoid pinching the retinal tissue with forceps. For similar safety reasons perhaps, nearly all surgeons preferred initiating the peel away from blood vessels. Nearly half of the surgeons initiated ILM peel 2 DD away from the MH, and one-fifth of surgeons preferred the inferotemporal quadrant. While there is no standard location to initiate the peel, the ILM is thickest and has maximum rigidity at nearly 1 mm from the foveal center. Therefore, it may be easier and also safer to initiate the peel away from the foveal center. Massaging the MH edge with a scraper, Finesse loop, or soft-tip cannula and draining through the MH with a soft-tip extrusion cannula have been considered by some authors to mobilize the adjacent retina and assist in hole closure in chronic, large, or persistent MH. In 2007, Alpatov et al . described the technique of tapping the MH edges from periphery to center with a vitreal spatula; since then, more surgeons have explored its role either alone or with other adjuvant maneuvers such as macular detachment or drainage through the MH. However, retinal massage has been associated with retinal pigment epitheliopathy, and drainage through MH can also damage the underlying retinal pigment epithelium (RPE) and adjacent photoreceptors. Nearly 80% of the surgeons did not consider these techniques in their practice. The preferred gas for endotamponade was SF6, with C3F8 being a close second option. Gas tamponade helps by reducing the fluid flow across the MH and brings the edges of MH closer by the action of interfacial surface tension. Initially, the choice of gas preferred was C3F8 gas, which would maintain these functions for a longer period and perhaps improve the closure rates, but now there has been a gradual shift to the use of shorter-acting SF6 gas. The understanding behind hole closure has improved, and we now know that factors other than choice of tamponade also play a role, such as type and duration of postoperative posture, fill of gas, chronicity of MH, and retinal compliance. The current evidence is weak and does not support the use of a particular gas. Conventionally, a prone position is advised for a duration of 3 days–1 week after MH surgery, and this practice was also observed in this survey. There is no definitive evidence regarding the need and duration of prone or face-down positioning (FDP) after MH surgery. However, closure rates and visual gains are better in the subset of large MH (>400 microns) with FDP versus other positions. What seems more relevant is the fill of gas; if it is more than 50% fill for a desired duration, then adequate MH tamponade could be achieved with the propped-up position as well. Larger MH may need tamponade for a longer duration (5–7 days) and hence these cases benefit from FDP. Traditionally, MHs >400 microns in size have been considered as large, and additional maneuvers such as ILM flap and those aimed at improving retinal compliance have been advocated to improve closure rates in this category. In 2013, even the IVMT study group classified MH >400 microns as large. Recently, it has been noted the holes larger than this close to a certain extent without additional maneuvers with comparable visual gains. In 2018, the Manchester Large Macular Hole Study reported MH >650 microns as large as these failed to close or had type 2 closure with only ILM peel. In 2021, the BEAVRS Macular Hole Outcome group suggested using 500 microns as the cutoff for large MH as beyond this the success rate starts to decline. Most recently, the CLOSE study group reported that ILM peel is enough for MH <530 microns, while MH between 535 and 800 microns need ILM flap techniques, and MH >800 microns need more invasive maneuvers such as macular hydrodissection or detachment, amniotic membrane transplant, or retinal autografts. Along similar lines, most of the surgeons in the survey reported considering MH >600 microns as large, followed by a close second option of >800 microns. In the category of large MH, the respondents preferred inverted or multilayered ILM flaps. The inverted ILM flap technique, originally described by Michalewska et al ., provides a higher anatomical closure rate as well as visual gain in large MH than only ILM peel by providing a bridge of glial tissue that contracts and brings the MH edges together. Very large MH require more invasive techniques or a combination of these techniques aimed at increasing retinal compliance (arcade to arcade peel or arcuate temporal retinotomy), freeing the retina from underlying adhesion with RPE (macular hydrodissection or detachment), promoting further glial reaction (platelet-rich plasma or autologous serum application), replacing the dead space with retinal tissue (retinal autografts), and bridging the space with other tissues (amniotic membrane or lens capsule). For failed, persistent, or recalcitrant cases, all the surgical methods mentioned above for very large MH could be used. However, the more commonly used technique is the free ILM flap. In our study also, the surgeons preferred either free flaps or repeat fluid-air exchange with longer-acting gas tamponade. The visual prognosis in repeat MH surgery tends to remain suboptimal despite anatomical closure. This explains why close to one-fourth of the surgeons in the survey did not consider repeat intervention. Several factors have been identified as predictive of visual outcomes after successful MH surgery, including preoperative visual acuity, MH duration, age, lens status, MH size, OCT parameters, autofluorescence patterns, type of tamponade, and duration of FDP. Among these, the study participants perceived duration of MH, preoperative vision, and MH size as the most important predictors. In 2021, Fallico et al . reported symptom duration as the most important predictor, followed by preoperative visual acuity and MH size. These factors indirectly reflect the condition of the external limiting membrane, ellipsoid zone, and interdigitation zone, which are key structures integral to the visual function. Inherent to most surveys, the current study had limitations of coverage bias, sampling bias, non-response bias, and recall bias of the respondents. The survey does not reflect the practice of all vitreoretinal specialists in the country. Being cross-sectional, the data applies only to the current times as MH surgery practices are continuously evolving. The study surveyed only the surgical approach and not the anatomical and functional outcomes of MH surgery. This survey revealed the current practice patterns of experienced vitreoretinal surgeons performing MH surgery in India and found them in accordance with the available evidence. There is a need to revisit and reassess the existing classification system for MH size and patient selection criteria, standardize the ILM peel techniques, particularly for beginners, and determine the most beneficial additional surgical maneuvers as per MH size and configuration. Abbreviations iMH: idiopathic macular hole; ILM: internal limiting membrane; PVD: posterior vitreous detachment; IVMT: international vitreomacular traction; BBG: brilliant blue G; SF6: sulfur hexafluoride; C3F8: perfluoropropane; AMG: amniotic membrane graft; FDP: face-down positioning; RPE: retinal pigment epithelium; ELM: external limiting membrane; EZ: ellipsoid zone; IZ: interdigitation zone. Conflicts of interest: There are no conflicts of interest. iMH: idiopathic macular hole; ILM: internal limiting membrane; PVD: posterior vitreous detachment; IVMT: international vitreomacular traction; BBG: brilliant blue G; SF6: sulfur hexafluoride; C3F8: perfluoropropane; AMG: amniotic membrane graft; FDP: face-down positioning; RPE: retinal pigment epithelium; ELM: external limiting membrane; EZ: ellipsoid zone; IZ: interdigitation zone. There are no conflicts of interest. |
Findings and Challenges in Replacing Traditional Uterine Cervical Cancer Diagnosis with Molecular Tools in Private Gynecological Practice in Mexico | e1552993-093c-48e0-a0da-a51b3efeabe0 | 11209306 | Gynaecology[mh] | The specific detection of the human papillomavirus (HPV) subtype has favored the reduction in the mortality and morbidity rate in women due to cervical cancer (CC) worldwide . This reduction has been achieved due to the widespread use of liquid-based cytology preparations to obtain cervical samples and the implementation of new molecular technologies for HPV genotyping by polymerase chain reaction (PCR) . Globally, CC is the fourth most common cancer in women, with 604,000 new cases in 2020 . About 90% of the 342,000 deaths caused by CC occurred in low- and middle-income countries . The highest rates of this tumor incidence and mortality are in sub-Saharan Africa (SSA), Central America, and South-East Asia. Regional differences in the CC burden are related to inequalities in access to vaccination, screening and treatment services, risk factors (including human immunodeficiency virus (HIV) prevalence and social and economic determinants such as sex), gender biases, and poverty. Women living with HIV are six times more likely to develop CC compared to the general population, with an estimated 5% of all cases attributable to this infection. The contribution of HIV to CC disproportionately affects younger women, and as a result, 20% of children who lose their mother to cancer do so due to this neoplasia . In Mexico, this neoplasm represents the second cause of cancer in women, with 9440 new cases per year, and the second cause of death, with 4340 cases . Among women with invasive CC, around 70% are diagnosed with locally advanced disease, which highlights deficiencies in timely diagnosis . All these data reinforce the need to continue implementing early screening strategies for CC in Mexico and in other Latin American countries with similar incidences. In recent years, the diagnosis of HPV infections by detection of viral DNA and PCR genotyping of high-risk (HR) variants in cervical samples has replaced the traditional Pap smear due to its higher sensitivity . The improvement in the HR HPV detection capacity offered by PCR and its automation has led various countries, such as Australia, to abandon the Papanicolaou test as a CC screening tool . On the other hand, in Latin America, the efforts to implement HPV genotyping by PCR continue to be limited, and greater awareness of this serious female health problem is necessary in public institutions and private hospitals . HPV is a highly transmissible virus that leads to transient infections, with several factors increasing the risk of its persistence, including genetics, age, smoking, and the specific genomic sequence of the infecting virus . To date, more than 150 subtypes of papillomavirus viruses have been described that, based on their association with CC, have been classified as very HR (HPV-16 and HPV-18), twelve HR subtypes (HR12), and other low-risk (LR) types that are associated with benign mucosal lesions . The high- and very-HR HPV subtypes (16, 18, and HR12) are associated with malignant lesions and cause approximately 70% of CC cases worldwide . The cytological course caused by transient HPV infection begins with low-grade squamous intraepithelial lesions, of which 90% revert to healthy epithelium. However, HR-HPV-specific infections aggravated by host risk factors are more difficult to reverse, and their persistence leads to high-grade squamous intraepithelial lesions (moderate or high dysplasia) that can progress to CC . Cytological assessment by the expert cytotechnologist is highly relevant for predicting the course of HPV infection. However, it has its limitations, such as low sensitivity and the impossibility of distinguishing persistent infection from reinfection . These are examples in which molecular tools such as PCR, given their greater sensitivity and ability to identify the subtype of HPV infection, outperform traditional diagnostics . To date, seven PCR assays for the detection and genotyping of HPV from cervical cell samples have been validated. Three of them are tests aimed at amplifying a region of the L1 gene (Abbott Real-Time HR HPV Test, Anyplex II HPV HR Detection, and Cobas 4800 HPV Test). At the same time, another four are assays that amplify early-region genes (BD Onclarity HPV Assay, HPV-Risk Assay, PapilloCheck HPV-Screening Test, and Xpert HPV) . In Mexico and the rest of Latin America, strategies have been developed to implement these molecular PCR tests for HPV detection in public and private institutions . However, automated PCR methods, such as those used in our laboratories (Cobas 4800 HPV Test), stand out from other methods for their ability to process many simultaneous samples and their speed and diagnostic accuracy . Using an internal control for co-amplification of a human gene makes it possible to practically eliminate the analysis of invalid samples and the presence of false negatives . Although the specific detection of HPV genotyping by PCR is currently the “gold-standard” technique in the early diagnosis of CC, complementary technologies have been developed for diagnosing premalignant cervical lesions in liquid cytology samples . The p16INK4a (p16) protein is a regulatory protein of the cell cycle under normal physiological conditions . This biomarker is effective in histological samples and is widely used to improve the reproducibility of cervical biopsy assessment and accuracy in detecting premalignant lesions . Likewise, the simultaneous detection of p16 and Ki67 (a proliferation biomarker) within the same cervical epithelial cell has been proposed as a marker of cellular transformation mediated by infections with the 12 HR HPV genotypes (HR12) . This combination of biomarkers (p16/Ki67 dual-stain cytology, Cintec-Plus) has provided excellent results in cervical cytological samples where it has been used for the detection of premalignant and malignant lesions of CC . Given the proven advantages of the herein-described molecular tests, we have been implementing them in our laboratories and boosting their adoption among private practicing gynecologists in Mexico. 2.1. Clinical Samples and Diagnostic Algorithms A total of 4499 cervix samples received consecutively in our laboratories were analyzed using three strategies: (a) molecular PCR analyses (to screen for the presence of HPV infections); (b) liquid-based cytology to search for cellular alterations suggestive of HPV infections, and (c) finally, if the medical expert followed the triage, they requested the laboratory to perform dual-staining cytology (to assess if the cellular transformation process has already started). PCR was used to test all 4499 samples, 3806 were subjected to liquid-based cytology, and 567 samples were analyzed by p16/Ki67 dual-stain cytology. 2.2. Clinical Specimen Sampling Cervical samples from the cervix were taken by gynecologists in private clinical practice using a cervix brush and deposited in a transport medium (ThinPrep or Roche Cell Collection Medium). The vials with the samples were sent at room temperature to the laboratory, where they were stored and refrigerated (4 degrees Celsius) until processing. 2.3. Liquid-Based Cytology (PAP Test) Liquid-based cytology slides were prepared by a cytotechnologist and interpreted according to the Bethesda System for Reporting Cervical Cytology (third edition, 2017). In contrast, a pathologist reviewed 50% of negative samples and 100% of the positive ones for quality control and quality assurance. 2.4. HPV PCR Assay The COBAS 4800 HPV Test (Roche) is an FDA-approved and validated qualitative test device for detecting HPV DNA in swabs from the cervical canal. This test amplifies target DNA isolated from cervical epithelium by real-time PCR to detect HPV 16 and HPV 18, along with a simultaneous pooled result for 12 other HR genotypes in a single test. The entire procedure is automated, and the manufacturer’s instructions are followed. The COBAS 4800 HVPV Test Primers are used to amplify DNA from 14 HR-HPV types (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68) in a single analysis, where probes with four different reporter dyes screen different targets in the multiplex reaction: dye 1 screens 12 pooled HR-HPVs (31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68), dyes 2 and 3 screen for HPV 16 and 18, respectively, while dye 4 targets the human β-globin gene to provide a control for uterine cervical cell adequacy for extraction and amplification. 2.5. p16/Ki67 Dual-Stain Cytology One slide for each sample was prepared for PAP testing using a Cytospin chamber adapter and subjected to p16/Ki67 dual-stain cytology using the CINtec Plus Cytology Kit (Roche Laboratories, Indianapolis, USA), according to the manufacturer’s instructions. Immunohistochemistry staining was performed using a BenchMark GX Stainer followed by evaluation by a trained cytotechnologist. An initial evaluation was performed to confirm the presence of the minimal criteria for squamous cellularity defined by the Bethesda terminology. Subsequently, the slide was checked for the presence of double-immunoreactive cervical epithelial cells, that is, cells with simultaneous brown cytoplasmic p16 immunostaining and Ki67 red nuclear immunostaining, which were interpreted as positive by double-stained cytological analysis regardless of the morphological interpretation. A pathologist reviewed all cases with positive cells for double-staining to confirm the result. The PCR results were compared with the PAP findings using a contingency table and chi-square test. All data analyses were performed using Graph Pad Prism 10 (GraphPad Software, Inc. (Boston, MA, USA). A total of 4499 cervix samples received consecutively in our laboratories were analyzed using three strategies: (a) molecular PCR analyses (to screen for the presence of HPV infections); (b) liquid-based cytology to search for cellular alterations suggestive of HPV infections, and (c) finally, if the medical expert followed the triage, they requested the laboratory to perform dual-staining cytology (to assess if the cellular transformation process has already started). PCR was used to test all 4499 samples, 3806 were subjected to liquid-based cytology, and 567 samples were analyzed by p16/Ki67 dual-stain cytology. Cervical samples from the cervix were taken by gynecologists in private clinical practice using a cervix brush and deposited in a transport medium (ThinPrep or Roche Cell Collection Medium). The vials with the samples were sent at room temperature to the laboratory, where they were stored and refrigerated (4 degrees Celsius) until processing. Liquid-based cytology slides were prepared by a cytotechnologist and interpreted according to the Bethesda System for Reporting Cervical Cytology (third edition, 2017). In contrast, a pathologist reviewed 50% of negative samples and 100% of the positive ones for quality control and quality assurance. The COBAS 4800 HPV Test (Roche) is an FDA-approved and validated qualitative test device for detecting HPV DNA in swabs from the cervical canal. This test amplifies target DNA isolated from cervical epithelium by real-time PCR to detect HPV 16 and HPV 18, along with a simultaneous pooled result for 12 other HR genotypes in a single test. The entire procedure is automated, and the manufacturer’s instructions are followed. The COBAS 4800 HVPV Test Primers are used to amplify DNA from 14 HR-HPV types (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68) in a single analysis, where probes with four different reporter dyes screen different targets in the multiplex reaction: dye 1 screens 12 pooled HR-HPVs (31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68), dyes 2 and 3 screen for HPV 16 and 18, respectively, while dye 4 targets the human β-globin gene to provide a control for uterine cervical cell adequacy for extraction and amplification. One slide for each sample was prepared for PAP testing using a Cytospin chamber adapter and subjected to p16/Ki67 dual-stain cytology using the CINtec Plus Cytology Kit (Roche Laboratories, Indianapolis, USA), according to the manufacturer’s instructions. Immunohistochemistry staining was performed using a BenchMark GX Stainer followed by evaluation by a trained cytotechnologist. An initial evaluation was performed to confirm the presence of the minimal criteria for squamous cellularity defined by the Bethesda terminology. Subsequently, the slide was checked for the presence of double-immunoreactive cervical epithelial cells, that is, cells with simultaneous brown cytoplasmic p16 immunostaining and Ki67 red nuclear immunostaining, which were interpreted as positive by double-stained cytological analysis regardless of the morphological interpretation. A pathologist reviewed all cases with positive cells for double-staining to confirm the result. The PCR results were compared with the PAP findings using a contingency table and chi-square test. All data analyses were performed using Graph Pad Prism 10 (GraphPad Software, Inc. (Boston, MA, USA). Four thousand and four hundred ninety-nine liquid-based cervical samples were performed on samples from patients referred by private gynecologists mostly from central regions of Mexico. A molecular PCR study for genotyping of HPV was performed on all samples to detect the 14 most common HR genotypes of HPV to contribute to the early prevention of CC. In parallel, Pap tests were performed on 84.6% ( n = 3806) of samples. The global distribution of the results of the viral genotypes revealed by the PCR analysis is described in . In 99.4% of cases, the samples received were considered adequate for this test, given that they rendered positive during the amplification of the internal control gene, i.e., the genomic β-globin gene. In cases where samples did not pass this control (0.6%), a second sample was requested from the gynecologist. This request was met on 53.8% of occasions (14/26). Thus, the final number of PCR results was 4487, or 99.7% of the samples received. Out of the 4487 results obtained, 68% tested negative for any of the 14 HR HPV genotypes analyzed. Out of the 1438 samples that tested positive for HR HPV (32%), 1261 (87.7%) were positive for any of the 12 HR HPV subtypes present in the group. In 12.3% of cases, the positive result corresponded to HPV16 and/or HPV18 subtypes, considered very HR for CC. It was observed that 86% of positive cases presented an isolated viral genotype without involving genotypes 16 or 18, while 14% showed co-infections with one of the two most oncogenic HR HPVs (viral genotype 16 or 18). Most of the positive samples yielded positive PCR results for the group of 12 HR subtypes (HR12). Overall, 24.5% of patients had PCR results involving very HR viruses (HPV16 and/or HPV18), alone or in combination with HR12 viral types. The results of PCR viral genotyping distributed by age range are shown in . The data shown in correspond to 3826 samples for which data on the age of the patients were available. However, if the results are stratified by age ranges, the population under 25 years of age was the one that showed the highest percentage of overall positive HPV-PCR results: 44% (181/411) tested positive for one of the high or very HR subtypes. If the results distributed by viral subtype are analyzed, the population between 25 and 35 years of age always shows a higher risk of infection with high or very HR genotypes. More than 50% of the results were positive in this age range for any viral genotype alone or in co-infection: 50.1% (450/898) for HR12, 51.8% (57/110) for HPV16, 55.2% (16/29) for HPV18, and 54.5% (91/167) for a viral co-infection. If the different positive subtypes are analyzed by age range, the HR12 subgroup always shows the highest percentage: 84% (152/181) in patients younger than 25 years, 73.3% (450/614) for the age range of 25 to 35 years, and 72.4% (296/409) for those over 35 years. However, for the other subtypes, their behavior by age is vastly different. For HPV16, 5% (9/181) was obtained for those under 25 years of age, 9.3% (57/614) for the range of 25 to 35 years, and 10.8% (44/409) for others over 35 years of age. For HPV18, the data were 0.6% (1/181) for those under 25 years of age, 2.6% (16/614) for those 25 to 35 years of age, and 2.9% (12/409) for those over 35 years of age. The data for viral coinfections showed that 10.5% (19/181) correspond to those under 25 years of age, 14.8% (91/614) for the age range of 25 to 35 years, and 13.9% (57/409) for those over 35 years of age. Analyzing the results globally, it stands out that: (a) The population under 25 years of age presented a higher percentage of positive results for HR12 viral types (84%), compared to 77.3% of the group between 25 and 35 years of age and a figure of 72.4% for those over 35 years of age. (b) The population between 25 and 35 years of age presented a higher percentage of results for co-infection, with 14.8%, compared to 13.9% in the group over 35 years of age and 10.5% in those under 25 years of age. (c) The population over 35 years of age presented a higher percentage of positive results for HPV16 with 10.8%, compared to 9.3% for the age range of 25 to 35 and 5% for those under 25. Likewise, the population over 35 years of age presented a higher percentage of positive results for HPV18 (2.9%), compared to 2.6% for the age range of 25 to 35 and 0.6% for those under 25 years of age. The global distribution of the 3806 Pap test results, which represent 84.6% of the total of 4499 PCR tests, is shown in . The positive results were sub-classified as low grade (69.1%) or high grade (30.9%). All negative and positive samples from the PAP test underwent PCR testing for the detection of high or very high-risk (HR) HPVs. Using a contingency table, we compared the results of the gold standard for HPV detection (PCR) with those of the PAP test. Within these findings, we noted that the percentage of false positives in cytology was 5.6% ( n = 200) of the total samples assessed. Additionally, presents the results of HR HPV-PCR among the 3184 samples identified as negative in the PAP test. Notably, among these, 703 samples (22.1%) yielded positive results for viral genotypes of high or very HR for CC. Within this subset of false negatives from cytology for HPV detection, 85.9% ( n = 604) corresponded to the 12 HR subtypes (HR12), while 14.1% ( n = 99) were attributed to HPV16 or HPV-18, which are very HR genotypes. shows the distribution by age of the 622 samples with a positive result for PAP, and whether they are low- or high-grade. The most striking results correspond to patients over 35 years of age: 110 samples (45 low-risk and 65 HR) produced negative results in the PCR test, representing 45.6% of potential false positives (110/241). The percentages of the lack of correlation between PAP and HPV-PCR were lower in the other age ranges: 25.4% in the age range of 25 to 35 years [(53 + 21)/291] and 17.7% for those under 25 years [(10 + 6)/90]. shows the HR HPV-PCR genotyping results in the 422 double-positive samples in both the HPV-PCR and PAP tests ( n = 322 low-grade and n = 100 high-grade). A total of 61.8% of the samples were positive for HR12, 16% were positive for HR12 plus HPV16, and 15.2% were positive for isolated HPV16. Overall, 79.6% of the analyses obtained an isolated infection [(261 + 64 + 11)/422], and 20.4% corresponded to double or triple co-infections [(4 + 67 + 1 + 41)/422]. shows the results of the confirmation tests for CC performed using p16/Ki67 dual-stain cytology (CINTEC PLUS). A total of 536 samples (37.3%) from the subgroup of 1438 positive results for 14 HR HPV types were tested. A total of 31.7% were positive for dual staining. These results also confirm that 68.3% of the positive results for any of the HR HPV subtypes do not present cellular transformation, which highlights the importance of performing p16/Ki67 dual-stain cytology after PCR to avoid unnecessary colposcopies in the screening of CC patients. These samples came from private practice gynecologists in private hospitals. In 100% of the cases, the gynecologists continued to perform the traditional Pap test in parallel with HPV genotyping analysis using PCR (Cobas). Although this double screening for the early diagnosis of CC represents an extra cost for their patients, gynecologists argue that the traditional Pap test allows them to identify other gynecological lesions (suspected infections by bacteria, fungi, or other viruses) that the PCR test alone does not allow. In , we present a strategy to encourage the increased adoption of molecular methods, highlighting the benefits of employing these tools throughout the entire process, from sample collection to patient reporting. In our experience, gynecologists are aware of the good sensitivity and specificity of the molecular PCR test and the costs of the study for their patients are the only reason for not requesting the HPV genotyping test together with the traditional PAP test. Regarding the possible reasons why the dual-stain test is not widely requested within the global detection program for CC in the private gynecology sector, we can include the following: (1) Extra costs for many of the patients, for whom the cost of the PCR study already represents a large increase over the expense of the traditional PAP; (2) Ignorance on the part of the patients of the objective and scope of the staining in CC screening, which implies a longer explanation time in private medical consultations; and (3) Reticence from a subgroup of gynecologists (colposcopists) since the positive/negative result of the dual-stain test determines (or relativizes) the need for surgical intervention (colposcopy) in their patients. This result seems to interfere with their area of expertise and business, which may be one of the causes of the low return rate in the confirmatory dual-stain test. The results presented here correspond to 4499 samples received in our laboratories during three years for CC screening. The distribution of HPV subpopulations in all samples is comparable to other results published internationally. However, this study shows for the first time the results in the Mexican population within the private medicine market . It should be noted that if we analyze the results within each age range, the population under 25 years of age was the one that showed the highest percentage of overall positive HR HPV-PCR results: 44% tested positive for one of the risk subtypes, high or very high, compared to 35.6% of positive results in the population between 25 and 35 years of age and 24.2% of positive results in the population over 35 years of age. Likewise, our results make it possible to highlight that the population between 25 and 35 years of age presented a higher percentage of results for viral co-infections with high or very HR for CC viral types at 14.8% compared to 13.9% in the group over 35 years of age and 10.5% in those under 25 years of age. On the other hand, the population over 35 years of age was the one that presented a higher percentage of positive results for HPV16 and HPV18. Once we organized the PCR results into a contingency table alongside the PAP results, we identified 5.6% false positives and 18.5% false negatives for the PAP results. These findings strongly support the advantages of using PCR as the standard test for detecting the human papillomavirus and preventing CC. The most striking results correspond to patients older than 35 years, who show 45.6% potential false positives. The percentages of PAP vs. HR HPV-PCR non-correlation were lower in the other age ranges: 25.4% in the age range of 25 to 35 years and 17.7% for those younger than 25 years. These results may suggest the existence of an age bias when interpreting PAP results in the older population, who are theoretically more prone to presenting with CC. Likewise, our data allowed us to detect a percentage of 68.3% negative results using confirmatory p16/Ki67 dual-stain cytology in the population of positive patients in the HR HPV group. These results made it possible to reassure the patients who had initially received a worrying result from the PCR test and to reduce the performance of unnecessary invasive tests (for example, colposcopy). Although the PCR test for HPV screening is well known by gynecologists, p16/Ki67 dual-stain cytology is not a widely requested test for CC screening. Out of the total HR-HPV-positive samples, only 37.3% were claimed for re-analysis with confirmatory dual staining. PCR tests detect molecular changes within the cell that are not manifested in morphological changes. This premise supports the promotion of PCR and p16/Ki67 dual-stain cytology as primary screening methods for HPV detection and CC prevention, as they provide greater diagnostic sensitivity to the physician. It is worth noting that, although there are other sensitive methods for HPV detection, such as NGS (Next-Generation Sequencing), this option is costly and would take longer to obtain results (even weeks). In addition, this method is not validated for the clinical diagnosis of CC. We acknowledge the limitations of our study in enhancing the robustness of our results. One limitation is the absence of clinical information or pathological tissue analysis of the patients, which is crucial for determining the sensitivity and specificity parameters, particularly within our study and the evaluated population sample. However, the most important challenges encountered after carrying out these works are: (1) Continue promoting, among gynecologists, that the papillomavirus PCR technique (HR HPV-PCR) is the best method for CC screening due to the lower presence of false negatives and false positives. (2) Stimulate the study of dual staining in HR12-positive patients to focus colposcopy procedures only on those patients who have cancer cells. (3) Promote the establishment of new protocols agreed between the Medical Societies of specialists (Gynecology and/or Colposcopy) for the better diagnosis and treatment of patients with CC . (4) Recently, a new global WHO strategy to achieve the elimination of CC as a public health problem by 2030 has been launched. It is called “Strategy 90-70-90”: 90% of girls (and boys in countries where resources allow) are to be fully vaccinated with the HPV vaccine by age 15, 70% of people are to be screened with a high-performance test, and 90% of persons identified with cervical disease will have received treatment . However, there are large inequities still to be resolved. Significant disparities continue to increase between countries and within different regions of the same country. Large differences persist between sociodemographic groups, household income, and children’s access to health insurance. The aspects in which we should improve to achieve these objectives are: (a) Accelerate the implementation of HPV screening programs, (b) Take advantage of diagnostic and therapeutic innovations, and (c) Focus on equity . To address these challenges, we are considering and proposing to (1) Disseminate in gynecology congresses in our countries the benefits of molecular diagnosis of HR HPV by PCR, insisting on its greater speed, specificity, and sensitivity. (2) Carry out dissemination campaigns of these new technologies in social networks aimed at the neediest female population in the prevention of CC. (3) Inform and achieve scientific discussion in specialized societies on the clinical risks associated with the unnecessary performance of many colposcopy procedures, for example, infertility or miscarriages. (4) Implement two innovations in our countries: (a) Promote self-collection programs for HR HPV-PCR screening, and (b) Test and validate new detection procedures close to the patient, such as point-of-care (POC) testing. The self-collection programs have already been validated by many countries around the world . The sensitivity and specificity of self-sample HPV tests are like provider-collected HPV tests, and some devices are excellently accepted by women from very different countries . Regarding the accessibility of less expensive and convenient methodologies than automated laboratory machines, especially for developing countries, there is great interest in POC testing instruments that could be fast, low-cost, and with a minimal training requirement . The HR HPV genotyping results by PCR show the low specificity and selectivity of the PAP test for CC screening. Up to 47.9% false positives have been observed in samples with a cytology diagnosed as high-grade positive. Likewise, 22.1% of false negatives have been observed in samples diagnosed as negative in the PAP test. The results of p16/Ki67 dual-stain cytology in positive PCR samples for any of the HR HPV subtypes show that 68.3% do not present with cancer cells, highlighting the importance of performing these tests to avoid unnecessary colposcopies after the screening of possible patients with CC. The results of our interactions with practicing gynecologists can be summarized as follows. There is a deeply embedded societal reluctance to abandon traditional PAP tests given the long history of public relations campaigns to convince women and their doctors of their value as the preferred prevention tool for the diagnosis of CC. In the opinion of some specialists, colposcopy remains preferred to dual-staining protocols despite the invasive and frequently unnecessary practice. Eradicating CC still is a challenge in Mexico due to poor prevention education and the inefficacy of current PAP technology. Our goal is to educate doctors and their patients about the benefits of the new molecular screening technologies for effective primary screening for HPV and the diagnosis of CC. |
Electrophysiology fellowship experience and requirements: an EHRA survey | 02f16bdc-45e1-42ff-a121-029eb4fa1e49 | 10487282 | Physiology[mh] | The field of cardiac electrophysiology (EP) is a rapidly growing subspecialty in cardiology, which requires understanding of mechanisms of cardiac arrhythmias to become familiar with different treatment strategies. It includes a wide range of conventional and complex catheter ablation procedures for various cardiac arrhythmias and the implantation and follow-up of cardiac implantable electronic devices. The Accreditation Committee of the European Heart Rhythm Association (EHRA) has previously introduced specific curricula in 2009, defining requirements for both training centres and trainees. However, a recent EHRA survey showed that considerable heterogeneity exists with respect to certification processes and standardized fellowship programmes during the EP training across European Society of Cardiology (ESC) member countries. This may particularly impact career building in countries, where EP centres are not broadly established, yet. Additionally, although the number of catheter ablation procedures is expected to grow, training opportunities are limited, and, depending on national regulations, not all centres fulfil recommended requirements for ablation centres. In this EHRA survey, we assessed the current educational experience and individual requirements and challenges of young electrophysiologists for EP education. To map EP fellowship experience and requirements in several EHRA countries, the EHRA e-Communication Committee and the Scientific Initiatives Committee prepared a questionnaire on SurveyMonkey. The official EHRA website, the EHRA newsletter, and the EHRA Young EP network as well as Twitter, LinkedIn, Facebook and personal mailing lists were used to disseminate the survey among relevant colleagues. The survey consisted of 22 questions in two blocks (see ): The first block consisted of general questions regarding personal information and demographics including gender, age, working position, working environment, and main specialty. The second block assessed the EP fellowship experience and individual requirements and challenges of young electrophysiologist for EP education using Likert scales. Statistical analysis Continuous variables were expressed as mean and standard deviation, and categorical variables were presented as numbers and percentages. Comparisons between groups were performed using Student’s t -tests or Mann–Whitney U tests for continuous variables as appropriate and chi-square test for categorical variables. Statistical analysis was performed using SPSS 25.0 for Windows (SPSS Inc., Chicago, IL, USA) and R (R Foundation for Statistical Computing, Vienna, Austria). Values of P < 0.05 (two-tailed) were considered as statistically significant. Continuous variables were expressed as mean and standard deviation, and categorical variables were presented as numbers and percentages. Comparisons between groups were performed using Student’s t -tests or Mann–Whitney U tests for continuous variables as appropriate and chi-square test for categorical variables. Statistical analysis was performed using SPSS 25.0 for Windows (SPSS Inc., Chicago, IL, USA) and R (R Foundation for Statistical Computing, Vienna, Austria). Values of P < 0.05 (two-tailed) were considered as statistically significant. General characteristics of respondents The characteristics of respondents are shown in Table . Two hundred and forty-three responders from 35 countries (32% female, age 38 ± 6 years old) completed the survey. Of those, 46% were fully trained electrophysiologists, 41% were electrophysiologists in training and 12% were non-electrophysiologists. In total, 16%, 9%, and 8% of all respondents graduated from universities in Germany, Italy, and Spain, respectively. Additionally, 20%, 9%, and 6% of respondents are currently working in Germany, Italy, and Spain, respectively. The majority (62%) of respondents were working in university hospitals; 15% were working in specialized public cardiology centres followed by 10% working in public community hospitals and private hospitals. Electrophysiology fellowship: in general Figure shows what activities the respondents consider to be important in learning EP. The three most important activities are ‘performing cases in hospital/hands-on’ (88.8%, very important), ‘seeing cases in hospital’ (74% very important), and ‘reading books/literature’ (54% very important). ‘International educational courses’ were considered to be important, followed by ‘webinars’ and ‘industry courses’, which ranked both lower. Respondents were asked to identify benefits ( Figure ) and challenges/obstacles ( Figure ) of undergoing an EP fellowship abroad. The main benefit for the respondents of undergoing an EP fellowship abroad is to get to know another country and to build up a network. Other benefits are related to the level of expertise and availability of a structured fellowship programme not available in their country. All surveyed challenges (finances, family, access to positions, time commitments, and board registrations) have been scored as significant and very significant by the majority (more than 50%) of all respondents. Electrophysiology fellowship: own experience Almost 70% of all respondents have either already completed or are currently partaking in a dedicated EP fellowship programme; 7% focus on implantable cardiac devices, 36% focus solely on EP, and 57% focus on both. Fifty-nine per cent participated in the EHRA courses/exams, 46% took a national certification exam (69% took a certification exam, either national or EHRA), and 41% of all respondents took part in an educational fellowship programme (most organized by industry). The main motivations for the choice of the fellowship institution were the reputation and volume of the centre, followed by the availability of a structured fellowship programme and research opportunities in the centre. Personal reasons are mentioned as well but were considered weaker on the motivation scale ( Figure ). The respondents were overall satisfied with their fellowships. Activities in invasive EP and basics in EP particularly fulfilled the expectations in about 80% of all respondents. Activities in non-invasive EP and cardiac pacing were satisfactory for 70% of all respondents ( Figure ). This was also reflected in the question regarding the trainees’ confidence in performing procedures independently after the fellowship. The respondents feel secure in diagnostic EP procedures, interventional EP procedures, cardiac pacing procedures, and checks/programming of cardiac devices. However, they expressed an ongoing lack of confidence in conduction system pacing and cardiac resynchronization therapy implantation. Almost 80% of respondents performed research during their fellowship. Most of them performed clinical research (77.4%), while only 4.1% performed basic science research. Less than 20% of respondents performed both, clinical and basic science research. An EP fellowship should be at least 2 years with 12 months focus on invasive EP, 7–12 months focus on cardiac pacing, and 3–6 months focus on non-invasive cardiac EP, if the goal is to cover all facets of EP ( Figure ). The characteristics of respondents are shown in Table . Two hundred and forty-three responders from 35 countries (32% female, age 38 ± 6 years old) completed the survey. Of those, 46% were fully trained electrophysiologists, 41% were electrophysiologists in training and 12% were non-electrophysiologists. In total, 16%, 9%, and 8% of all respondents graduated from universities in Germany, Italy, and Spain, respectively. Additionally, 20%, 9%, and 6% of respondents are currently working in Germany, Italy, and Spain, respectively. The majority (62%) of respondents were working in university hospitals; 15% were working in specialized public cardiology centres followed by 10% working in public community hospitals and private hospitals. Figure shows what activities the respondents consider to be important in learning EP. The three most important activities are ‘performing cases in hospital/hands-on’ (88.8%, very important), ‘seeing cases in hospital’ (74% very important), and ‘reading books/literature’ (54% very important). ‘International educational courses’ were considered to be important, followed by ‘webinars’ and ‘industry courses’, which ranked both lower. Respondents were asked to identify benefits ( Figure ) and challenges/obstacles ( Figure ) of undergoing an EP fellowship abroad. The main benefit for the respondents of undergoing an EP fellowship abroad is to get to know another country and to build up a network. Other benefits are related to the level of expertise and availability of a structured fellowship programme not available in their country. All surveyed challenges (finances, family, access to positions, time commitments, and board registrations) have been scored as significant and very significant by the majority (more than 50%) of all respondents. Almost 70% of all respondents have either already completed or are currently partaking in a dedicated EP fellowship programme; 7% focus on implantable cardiac devices, 36% focus solely on EP, and 57% focus on both. Fifty-nine per cent participated in the EHRA courses/exams, 46% took a national certification exam (69% took a certification exam, either national or EHRA), and 41% of all respondents took part in an educational fellowship programme (most organized by industry). The main motivations for the choice of the fellowship institution were the reputation and volume of the centre, followed by the availability of a structured fellowship programme and research opportunities in the centre. Personal reasons are mentioned as well but were considered weaker on the motivation scale ( Figure ). The respondents were overall satisfied with their fellowships. Activities in invasive EP and basics in EP particularly fulfilled the expectations in about 80% of all respondents. Activities in non-invasive EP and cardiac pacing were satisfactory for 70% of all respondents ( Figure ). This was also reflected in the question regarding the trainees’ confidence in performing procedures independently after the fellowship. The respondents feel secure in diagnostic EP procedures, interventional EP procedures, cardiac pacing procedures, and checks/programming of cardiac devices. However, they expressed an ongoing lack of confidence in conduction system pacing and cardiac resynchronization therapy implantation. Almost 80% of respondents performed research during their fellowship. Most of them performed clinical research (77.4%), while only 4.1% performed basic science research. Less than 20% of respondents performed both, clinical and basic science research. An EP fellowship should be at least 2 years with 12 months focus on invasive EP, 7–12 months focus on cardiac pacing, and 3–6 months focus on non-invasive cardiac EP, if the goal is to cover all facets of EP ( Figure ). The main findings of this EHRA fellowship survey are as follows: Hands-on participation and observation of EP procedures are very important. The main motivations for the choice of the fellowship institution are the reputation and volume of the centre, as well as the availability of a structured fellowship programme. The majority of respondents participated in the EHRA exam and/or took a national certification exam. Respondents are overall satisfied with their fellowship. However, a degree of lack of confidence in certain areas of complex cardiac pacing procedures remains after the completion of the fellowship. The majority of respondents performed research during their fellowship. But just 4.1% performed purely basic science research and 18.6% performed basic science research in combination with clinical research. The optimal duration of an EP fellowship should be a bit above 2 years, if the goal is to cover all facets of EP. Doing fellowships abroad is beneficial, but significant obstacles exist. The results of this survey indicate that the most preferred way of learning EP is hands-on participation and observation of EP procedures, ideally in high-volume centres with a good reputation. In addition to the active and passive involvement in clinical cases and procedures, educational programmes are also used widely. A previous ESC survey showed considerable heterogeneity with respect to certification processes and standardized fellowship programmes during the EP training. Our survey shows that most of the trainees voluntarily take the certification exam by EHRA or their national societies. Of note, besides the EHRA courses/exams and national courses, which are completed by 69%, a large proportion of fellows also participated in industry-sponsored educational courses. Although these industry-sponsored educational courses were established by the wider EP community, the role and responsibility of scientific societies such as EHRA in the delivery and quality of this form of training should be further explored. Additionally, the digital and social media transformation of cardiac EP education has revolutionized the way education is currently delivered by hybrid in-person and virtual modalities providing electrophysiologists with the flexibility to choose the best option to suit their individual needs and preferences for continuing education. On top of a theoretical evaluation, such the one performed by the EHRA exam, also a practical evaluation, which should be not just focus on time and the number of procedures, but also on the assessment of skills, might make sense. The EHRA is offering different types and levels of training and education opportunities such as webinars and simulation village at the EHRA Congress 2023. Additionally, also simulator training may significantly improve the independent trainees’ performance, particularly during the early phase of the trainees’ learning curve. Another interesting result was that research is frequently incorporated in the EP fellowships. The largest proportion of respondents focused on clinical research, while basic science research was just performed by a minority. This is in line with a recent EHRA survey on research activities. Therefore, centres providing EP fellowship positions should also facilitate access to a supportive research environment, to ensure the most optimal output of the research activities during an EP fellowship. Overall, respondents were satisfied with their own fellowship; however, a majority of respondents did not feel confident in performing interventional EP and device implantation (particularly cardiac resynchronization therapy and conduction system pacing implantation). This finding has been already shown in other national surveys before. While the conduction system pacing is a growing field and one may argue that there are few centres that are actual experts, cardiac resynchronization therapy has been around for years and there is no reason why the teaching of these techniques should be insufficient during training. Based on our EHRA survey, an EP fellowship should be at least 2 years with 12 months focus on invasive EP, 7–12 months focus on cardiac pacing, and 3–6 months focus on non-invasive cardiac EP, if the goal is to cover all facets of EP. The EHRA provides support through the ‘EHRA Training Fellowships’ and the ‘EHRA Observational Training Program’ focusing on clinical EP with emphasis on catheter ablation and cardiac pacing with emphasis on implantable cardioverter defibrillator/cardiac resynchronization therapy. These programmes are offered to allow physicians to gain specialized training in clinical EP in an ESC member country preferably outside their home country, for example in the EHRA Recognised Training Centers (ERTC) ( https://www.escardio.org/Education/Career-Development/Grants-and-fellowships/EHRA-training-fellowships ). However, as indicated in previous surveys, training opportunities are limited in many EHRA countries [mean number of centres accredited for EP training per country was 10 ± 31 (range 0–149)]. Additionally, although respondents see benefit in doing their EP fellowship abroad in a country, where training positions are available, several obstacles exist. The EHRA can play an active and important role in ensuring access to training in EP and cardiac pacing. Also, participation in an educational programme such as the ‘Diploma of Advanced Studies in Cardiac Arrhythmia Management’ (DAS-CAM) may provide opportunities to build networks, which has been mentioned as one of the main benefits of undergoing an EP fellowship abroad. The EHRA Certification Committee has previously introduced a specific curriculum in 2009, defining requirements both for training centres and trainees. Currently, a committee from the EHRA Certification Committee is updating the EHRA core curriculum. The new version will be presented at EHRA 2024. Limitations As it is the case for all surveys, there might be a responder bias that cannot be neglected. Moreover, the respondent’s geographical distribution is focused on Europe with EHRA countries as the main source of replies. Therefore, caution should be made in generalizing the results of the present survey to other regional settings. As it is the case for all surveys, there might be a responder bias that cannot be neglected. Moreover, the respondent’s geographical distribution is focused on Europe with EHRA countries as the main source of replies. Therefore, caution should be made in generalizing the results of the present survey to other regional settings. This EHRA fellowship survey showed that (i) hands-on participation and observation of EP procedures are very important; (ii) the main motivations to choose a fellowship institution are the reputation and volume of the centre as well as the availability of a structured EP fellowship programme; (iii) the majority of respondents took the EHRA exam and/or took a national certification exam; (iv) lack of confidence upon completion of an EP fellowship remains in performing conduction system pacing and cardiac resynchronization therapy implantation independently; (v) the majority of respondents performed research during the fellowship; (vi) the optimal duration of an EP fellowship should be above 2 years; and (vii) doing fellowships abroad is beneficial, but significant obstacles exist. The results of this EHRA survey, including own experiences and expectations, may help to refine current EP fellowship programmes to improve the quality of EP training and early career building of young electrophysiologists. euad249_Supplementary_Data Click here for additional data file. |
General practice staff and patient experiences of a multicomponent intervention for people at high risk of poor health outcomes: a qualitative study | 5d716867-1121-4654-bebd-39febc796b6c | 10775450 | Family Medicine[mh] | An efficient and adequately resourced primary health care sector is recognised as critical for improved population health outcomes and for health funding to be sustainable. Australia faces an aging population, rising rates of chronic and complex disease and a growing demand for hospital and other expensive healthcare services. These challenges are not unique to Australia and have encouraged policy makers both locally and internationally to consider wide ranging reforms to the organisation and funding models of their primary care sector. These primary care reforms have included improved access to and continuity of care (typically through patient registration), expanding multidisciplinary care and better coordination and integration between primary and hospital care . In Australia general practice is funded by the federal government on a fee-for-service basis with most services eligible for rebates from the universal health insurance scheme known as Medicare. In contrast to other countries where patients register to a particular GP or to a specific practice Australian patients are free to consult with multiple GPs, including those at different general practices . Despite the lack of a formal registration system most Australians with (85%) or without (70%) a long term health condition report having a preferred GP although multiple practice attendance, particularly for younger people, is not uncommon . A strength of the Australian model is that promotes patient choice and encourages provider competition but the lack of a formalised relationship between GPs and their patients does not encourage continuity of health care which is widely recognised as a core principle of primary care . Systematic reviews suggest that there is an association between continuity of care and patient satisfaction and health service utilisation but this evidence has mainly been generated from cross-sectional studies with very few interventional trials. Observational studies from countries that have introduced primary care registration systems to promote continuity of care are suggestive of better patient outcomes and reduced costs but overall the evidence base is considered weak . The Australian fee-for-service funding model incentivises GPs to provide a higher volume of services as opposed to providing higher value care . In addition, the Medicare rebate structure discourages longer consultations by setting decreasing (on a per minute basis) patient rebates as appointment length increases encouraging GPs to focus on speed over need . By international standards Australia’s average length of GP consultation of just under 15 min would not be considered short . The interpretation of this however is complicated by the fact that Australia’s fee for service model has acted as a barrier for the development of general practice multidisciplinary teams . As a result, compared to GPs in other countries, Australian GPs have less opportunity to delegate administrative tasks and basic clinical work . While general practice is funded and managed by the federal government, hospital services are administrated separately by the eight State and Territory governments. This provides little incentive for intersectoral collaboration or communication to deliver the best possible patient care. The transition between hospital to home and GP care is associated with high risk for adverse events and avoidable hospital readmissions, particularly for older people with complex needs . The extent to patients receive timely GP follow-up after hospital discharge depends on a set of complex factors relating to patient’s perceptions of the value of GP follow-up, GPs receiving notification of the hospitalisation and hospitals communicating the need for GP follow-up to both the patient and the GP . From the limited Australian research available it appears that in the order of around one-third of patients discharged from hospital do not see a GP within 14 days . In 2018 the Royal Australian College of General Practitioners, in collaboration with the federal government, funded a clustered randomised trial (titled the Flinders Quality Enhanced general practice Services Trial: Flinders QUEST) to test a multicomponent general practice intervention aimed at improving health outcomes and health service use for patients at high risk of poor outcomes. The multicomponent intervention was designed to improve continuity of care (defined as general practice appointments with patient’s regular preferred GP), access to long appointments and timely (with 7 days) general practice follow-up if patients experienced a hospital care episode. Flinders QUEST was conducted in 20 general practices located in the metropolitan area of Adelaide, South Australia. Ninety-two participating GPs were provided a list of their active patients (three or more visits in the previous two years) drawn from three cohorts (1) children and young people aged 0 to 17 years; (2) adults aged 18–64 with two or more chronic illnesses; (3) older people aged 65 years and above. GPs were asked to identify 18 patients who they believed were at risk of poor health outcomes and who may potentially benefit from the intervention. GPs were asked to use their clinical judgement to prioritise patients who were not too low a risk of poor health outcomes nor those who were so seriously ill that it was too late for the intervention to work. The implementation of the intervention and quantitative results, at 12-month outcome assessment, have been reported elsewhere . In brief the intervention was implemented to a reasonable standard and there were statistically significant improvements to continuity of care and the number of longer length of appointments. There was a greater likelihood of follow-up after emergency department or hospital care episodes but this was not statistically significant. The intervention was not found to improve self-rated health (the primary outcome of the trial), nor were there any statistically significant intervention effects for health service utilisation. The economic evaluation found that the intervention was more effective in terms of Quality Adjusted Life Years (QALY) but considering the payment (A$1,000 per patient) intervention practices received for providing the intervention the intervention was not cost-effective. In a pre-specified exploratory sub-group analysis of older people (69% of the total sample), the intervention was found to be cost-effective primarily due to a reduction in hospital usage. There is increasing recognition of the potential benefits of qualitative research within randomised clinical trials and Flinders QUEST included a qualitative component to complement the quantitative evaluations. Consistent with the most frequent use of qualitative research in clinical trials, particularly for complex interventions , the present qualitative study was focussed on the multicomponent intervention. Specifically, we wished to better understand participants perspectives on each of the components of the intervention, gather their opinions on whether the intervention had improved general practice services and may have resulted in hospital avoidance and whether practice staff believed the intervention would be sustainable after the trial had completed. It was intended that the findings from the qualitative study would complement the and enrich the understanding of the main quantitative results from the trial.
This article was written in accordance with the standards for reporting qualitative research (SRQR). Context and setting This study was conducted with general practice staff (Practice Managers [PMs], GPs, Practice Nurses [PNs) and their patients in Adelaide, South Australia. Study design We conducted a qualitative study using semi-structured interviews. Separate interview guides were developed by the research team for practice staff and intervention group patients. Practice staff interviews explored perspectives of each of the components of the intervention (including factors that facilitated or constrained its implementation in practices), the mechanisms through which the intervention may have improved patient health outcomes and health service usage and finally the sustainability of the intervention after the trial had completed. A draft interview guide was created by the first author (SJ) and then developed iteratively with the input from the Flinders QUEST chief investigator (RR) and trial manager (LR). The guide was designed to elicit participants perspectives (GP staff and patients) about the components of the intervention and whether the intervention had (in their opinion) improved general practice services (relating to continuity of care, appointment length and general practice follow-up after a hospitalisation) during the trial period. A key secondary outcome of the trial was whether the multicomponent intervention resulted in reduced hospital use. In the trial this was assessed quantitively from hospital administrative records and in the qualitative interviews we wished to better understand the mechanism through which any reduction might have occurred. For this reason we specifically targeted patients who had experienced one or more hospital care episodes during the intervention period. Given the small number of children and young people in the trial they were not included in the qualitative study. Participants For general practice staff, we aimed to conduct one interview with the PM in each control group practice and three interviews (PM, GP, PN) in each intervention group practice. We included control group PMs because we were interested in their experience of the research process (e.g. patient recruitment, data provision, working with the research team) and also whether control group practices had engaged in any other quality improvement activities during the intervention period. For intervention group practices, we invited the PM to nominate a GP and a PN who had played an active role in the trial to take part in the interviews. A participant information sheet and consent form were forwarded to practice staff and the interview sessions arranged. Of the 10 control group practices all PMs (one PM was the manager at two practices) agreed to be interviewed. Of the 10 intervention group practices all the PMs and their nominated GPs and PNs agreed to be interviewed. In one intervention group practice two PNs who had been involved in the implementation of the study expressed a desire to be interviewed jointly. In another intervention group practice an Administrative Officer who had played a key role in the implementation of the study was interviewed along with the PM at the PM’s request. Patient recruitment to the qualitative study was conducted in three stages. Overall, 1044 patients drawn from three cohorts took part in Flinders QUEST: children and young people under the age of 18 years ( n = 58); adults aged between 18 and 64 years with two or more chronic diseases ( n = 315) and older people aged 65 years and above ( n = 671). In the first stage, intervention group patients (in the adults and older adults cohorts) who had indicated a willingness via a response to a question in a six month follow-up questionnaire to take part in an interview about their experiences in the trial were identified. From 468 potentially eligible patients 391 (84.6%) had responded positively to this question. This was further refined to 188 (47.5%) patients who had reported one or more hospital care episodes (emergency department presentation or hospital admission) during the preceding 12 months. A purposeful sample of 55 patients was selected to ensure representation by practice, cohort (adults with two or more chronic diseases and older people aged > 64 years) and gender. An initial phone call was made to potential interviewees to provide background information about the qualitative study and for those expressing an interest in receiving further details a participant information sheet and consent form was posted to them. Of the 55 patients approached, 45 agreed to be interviewed with written consent completed at the time of the interview. Patients received a A$20 gift voucher for participating in the interview. Data collection The practice staff and patient interviews were conducted by the first author, an experienced qualitative researcher, between November 2019 and March 2020, which was after the 12-month intervention period of the trial had completed. Practice staff interviews were conducted in general practices and took between 30 and 45 min. The patient interviews mostly occurred in patient’s homes and took between 30 and 60 min. The study was not affected by COVID-19; the first recorded case in South Australia was reported on 1 February 2020, by which time most of the interviews had been completed. Data analysis The interviews were audio-recorded for transcription and further analysis. Interview files were transcribed by a professional transcribing service and imported to the qualitative analysis software (NVivo 12). A coding framework was developed using the key themes that emerged from the interviews and discussed in the research team.
This study was conducted with general practice staff (Practice Managers [PMs], GPs, Practice Nurses [PNs) and their patients in Adelaide, South Australia.
We conducted a qualitative study using semi-structured interviews. Separate interview guides were developed by the research team for practice staff and intervention group patients. Practice staff interviews explored perspectives of each of the components of the intervention (including factors that facilitated or constrained its implementation in practices), the mechanisms through which the intervention may have improved patient health outcomes and health service usage and finally the sustainability of the intervention after the trial had completed. A draft interview guide was created by the first author (SJ) and then developed iteratively with the input from the Flinders QUEST chief investigator (RR) and trial manager (LR). The guide was designed to elicit participants perspectives (GP staff and patients) about the components of the intervention and whether the intervention had (in their opinion) improved general practice services (relating to continuity of care, appointment length and general practice follow-up after a hospitalisation) during the trial period. A key secondary outcome of the trial was whether the multicomponent intervention resulted in reduced hospital use. In the trial this was assessed quantitively from hospital administrative records and in the qualitative interviews we wished to better understand the mechanism through which any reduction might have occurred. For this reason we specifically targeted patients who had experienced one or more hospital care episodes during the intervention period. Given the small number of children and young people in the trial they were not included in the qualitative study.
For general practice staff, we aimed to conduct one interview with the PM in each control group practice and three interviews (PM, GP, PN) in each intervention group practice. We included control group PMs because we were interested in their experience of the research process (e.g. patient recruitment, data provision, working with the research team) and also whether control group practices had engaged in any other quality improvement activities during the intervention period. For intervention group practices, we invited the PM to nominate a GP and a PN who had played an active role in the trial to take part in the interviews. A participant information sheet and consent form were forwarded to practice staff and the interview sessions arranged. Of the 10 control group practices all PMs (one PM was the manager at two practices) agreed to be interviewed. Of the 10 intervention group practices all the PMs and their nominated GPs and PNs agreed to be interviewed. In one intervention group practice two PNs who had been involved in the implementation of the study expressed a desire to be interviewed jointly. In another intervention group practice an Administrative Officer who had played a key role in the implementation of the study was interviewed along with the PM at the PM’s request. Patient recruitment to the qualitative study was conducted in three stages. Overall, 1044 patients drawn from three cohorts took part in Flinders QUEST: children and young people under the age of 18 years ( n = 58); adults aged between 18 and 64 years with two or more chronic diseases ( n = 315) and older people aged 65 years and above ( n = 671). In the first stage, intervention group patients (in the adults and older adults cohorts) who had indicated a willingness via a response to a question in a six month follow-up questionnaire to take part in an interview about their experiences in the trial were identified. From 468 potentially eligible patients 391 (84.6%) had responded positively to this question. This was further refined to 188 (47.5%) patients who had reported one or more hospital care episodes (emergency department presentation or hospital admission) during the preceding 12 months. A purposeful sample of 55 patients was selected to ensure representation by practice, cohort (adults with two or more chronic diseases and older people aged > 64 years) and gender. An initial phone call was made to potential interviewees to provide background information about the qualitative study and for those expressing an interest in receiving further details a participant information sheet and consent form was posted to them. Of the 55 patients approached, 45 agreed to be interviewed with written consent completed at the time of the interview. Patients received a A$20 gift voucher for participating in the interview.
The practice staff and patient interviews were conducted by the first author, an experienced qualitative researcher, between November 2019 and March 2020, which was after the 12-month intervention period of the trial had completed. Practice staff interviews were conducted in general practices and took between 30 and 45 min. The patient interviews mostly occurred in patient’s homes and took between 30 and 60 min. The study was not affected by COVID-19; the first recorded case in South Australia was reported on 1 February 2020, by which time most of the interviews had been completed.
The interviews were audio-recorded for transcription and further analysis. Interview files were transcribed by a professional transcribing service and imported to the qualitative analysis software (NVivo 12). A coding framework was developed using the key themes that emerged from the interviews and discussed in the research team.
Forty-one face-to-face interviews were conducted with practice staff (19 PMs, 1 Administrative staff, 11 PNs and 10 GPs). Forty-five interviews (3 by phone and 42 face-to face) were conducted with patients (25 female and 20 male). Most patients interviewed were drawn from the trial’s older adults cohort (69.0%) and the overall mean age was 71 years (SD = 11.3). Most patients were married (73.3%), retired (60.0%) and had a yearly income of less than A$60,000 (63.6%). Patients reported a mean of 3.9 (SD = 1.6) chronic diseases with the most frequent types being cardiovascular (57.8%) and musculoskeletal (68.9%) disorders. At baseline (the start of the intervention) patients reported a mean of 2.5 (SD = 2.6) emergency department presentations and 1.4 (SD = 1.9) hospital admissions during the preceding 12 month period. Practice staff and patient perspectives of the multicomponent intervention Continuity of care Practice staff valued GP continuity of care for service quality, trust and building relationships: “Continuity of care is that you don’t have to cover off that big chunk of their life every time. You just know what’s going on, just by them walking in the room before they even said anything…You get that connection with people. They will always tell you more with that connection. You’ve got more time because you don’t have to cover all that other history that you already know. Yeah, you can kind of nuance things a bit more, because you can pick the little things out that aren’t quite right with what’s going on.” (Intervention practice, GP). While continuity of care with a regular GP was generally supported, some practice staff drew a distinction between patients being ‘loyal to the practice’ versus being ‘loyal to a regular GP’. Strategies such as communication between GPs within a practice, systems in place enabling GPs to share patient’s notes and information and having a second regular GP were viewed by some as more likely to lead to a sustainable general practice model: “I should be selling the practice, the quality practice where all the doctors write good notes and knowledgeable caring and are on the same page. That’s what I’d like to do is sell the practice rather than an individual GP because if I go away for three weeks or something then what do they do? We all have holidays.” (Intervention practice, GP). There were also perceived clinical benefits from having a variety of different GPs involved in patient care: “I don’t want any of my patients to be absolutely dependent on me. I actually think it’s healthy for them to see another doctor because I might miss something that another doctor picks up. I might have some level of expertise in one area and another doctor might have a level of expertise in another area, so I actually think that concept of a preferred GP needs modification.” (Intervention practice, GP). A PM also commented: “You sometimes get a patient so dependent on one doctor that if that doctor’s away, they won’t see anyone else. To me, that’s putting their health at risk. It’s hard sometimes to convince people that, ‘Yes, even though he’s away, please come and see’ – and sometimes another doctor can shine a different light on the problem.” (Control practice, PM). Continuity of care was valued by patients for largely the same reasons as practice staff. Patients believed it enabled them to build trusted relationships with their GP and helped their GP to better understand their family and social circumstances: “Well, familiarity I guess and just I suppose the information, the things that you discuss with him then, I mean he’s got a fuller picture of a person as a patient, rather than just a random doctor here and there. They don’t have the whole picture.” (Female, 75 yrs. Old). Continuity of care was particularly valued by patients with complex health problems, mental illnesses or for other ‘personal’ issues: “I said to her [doctor] I didn’t feel like having to explain everything to everybody; especially when you’re losing blood from the bowel region, it’s embarrassing.” (Male, 66 yrs. Old). Several patients also raised the idea of a second regular GP or a practice-based GP service. For some, particularly those with long-term connections with their general practice, being able to see any GP within the practice was a way to fill in the gaps in care continuity with their preferred GP. Patients also reported the benefits of visits with other GPs, for example, consultations with a female GP for gender specific health issues or screenings or with GPs who are specialised in specific health conditions such as skin cancer. Patient perceived barriers to continuity of care included GPs who worked part-time, particularly, younger aged female doctors and GPs absences due to sickness. Longer appointments Longer appointments were viewed by practice staff as being particularly advantageous for patients with chronic and complex health conditions because they enabled more comprehensive care to be provided: “Rather than just looking at what the presenting complaint may be, it’s doing a full thorough check on someone. It’s being able to take that time to properly talk to them, to properly give education, to provide information, make sure that the understanding is there and answer any questions and that sort of thing, which is really important.” (Intervention practice, PN). On the other hand, some practice staff noted that patients with chronic and complex health conditions were more likely to have frequent appointments and this might be more beneficial compared with less frequent longer appointments: “If they’re seen once a month, a 15-minute appointment once a month that would be more valuable to them than a half-hour appointment every three months…and if they’re checking in with the nurse every three months, monthly 15-minute appointment with the doctor, they’re going to get everything that they need.” (Intervention practice, PM). Some GPs noted that the current Medicare rebate structure discourages longer appointments, and this requires changes at broader policy and system levels: “The way the schedule fee is based, you’re actually rewarded for shorter appointments rather than longer appointments. I think ultimately that needs to change, because the population at large is getting older, the complexity of patients is getting more difficult, and patients are becoming more knowledgeable. They don’t want to come in and just be given a script and told to do this or do that, they want more of an explanation of what’s going on and the rest of it.” (Intervention practice, GP). Longer appointments were valued by patients because it was felt they provided an opportunity to discuss complex health problems thoroughly (especially in the case of mental health issues), and to review medications. Longer appointment times were also thought to allow time for more questions to be asked. One patient described standard length appointments as: “… I think this business about working on a 15-minute appointment, to my mind, doesn’t work well, right…I call it shop medicine, right, you go in and you buy something and you go out.” (Male, 89 yrs. Old). However, patients also reported that appointment length should be based on need and that a long appointment was not always required for example for prescription renewals, test results or for simple problems: “Sometimes it’s not needed. Sometimes I think 30 minutes was too much, but there was other times when it was good to have that extra time.” (Female, 58 yrs. Old). Patients also raised the importance of doctor-patient communication as being equally important to appointment length: “What is important is the communication between the doctor and patient. You may have a longer appointment but if the doctor doesn’t communicate well, the longer appointment does not work.” (Male, 69 yrs. Old). Timely follow-up after a hospital care episode Patient follow-up after a hospital care episode was considered by practice staff to be valuable but challenging to achieve consistently due to poor communication between hospitals and general practices which led to delays in practices receiving patient discharge summaries: “I think the biggest thing is that the discharge information, once they’ve been in hospital and struggling to pin people down.” (Intervention practice, PM). “It really disrupts the continuity because eventually you’ll say, they need to have follow up bloods done three days post discharge. Well, they’ve been home for a week now.” (Intervention practice, PN). “Private hospitals don’t have their own resident medical staff. And it’s a medical handover issue, they don’t see that as their problem. As opposed to a public hospital where they actually have a resident medical staff so you can address the concern or something to that staff as an entity. And that’s the problem I think, that’s where it breaks down unfortunately.” (Intervention practice, GP). Similarly, patients reported examples of their GP not being aware of their hospital care episode: “He [GP] said “When did you get back?” and I said “Mate, I’ve just been in hospital”… I told him! So you know, that’s where the system falls down.” (Male, 65 yrs. Old). Another patient reported: “I think there is a problem with the hospital and their follow-ups. Whenever I came out from a stay in hospital, I’d do a follow-up appointment with the doctors and usually they’d have no idea that you’ve been in there. They haven’t received follow-up reports or anything like that… the run of the mill is that you’re told you should follow-up within a fortnight of leaving hospital. So we’d make that appointment and you’d see them [GP] and they’d say “Oh we haven’t received anything” so there is a lag time there.” (Male, 69 yrs. Old). Practice staff and patient views on the impact of the intervention on general practice services The extent to which the intervention had provided general practice care that was different from usual care was one of the key themes that emerged from the practice staff interviews. Noting that the patients in the trial had been identified as at high risk of poor outcomes, practice staff often felt that they had strategies in place prior to the trial to ensure high levels of continuity of care and access to longer appointments: “It was along the same path of what we’ve already done, been trying to do something to help the target group in this case… If we managed to get into intervention, then the work that we do with our patients is not dissimilar to what we do already, in terms of that patient care and trying to give that extra bit which the QUEST was all about.” (Control practice, PM). “We are already doing the job, but now we had more incentive, we’ve got appointments and more nurses and staff were knowing that we need to take care of people, those people better and more efficiently… Yeah, special treatment, it made it a bit more systematic.” (Intervention practice, GP). There was a common perception that the intervention had largely been a part of usual care and therefore there were not significant changes because of the trial: “As a practice, that’s how we like to run things anyway. We’re very chronic-diseased focused, which you’d want to see the same GP if you can, not always possible, but most of the time we will try and make that happen anyway.” (Intervention practice, PM) . “We’re already implementing a lot of things before you even start. So to some degree it was a bit the icing on the cake.” (Intervention practice, GP) . On the other hand, some intervention group practice staff reported that participating in the trial had increased their level of awareness about proactively addressing the needs of high risk patients: “I’ve become more proactive with patients’ problems. Not only with the patients registered in the trial, but even with other patients coming in with similar problems. I initiated the same sort of practices so that they can benefit also, like giving them more time, looking into more preventative care before the problem started.” (Intervention practice, GP) . The benefits of longer appointment times were appreciated by some GPs who noted that: “It gives you time to not rush and be able to let them open up and talk about what’s going on, and look up their bloods and bone densities and all those kinds of things that in 15 minute-appointments, I find it to be too short, sharp. Because 15 is really 10 to 12, by the time you get notes and things done, it doesn’t give a lot of time. So, personally, double appointments I think are more beneficial, and reduce stress levels as well.” (Intervention practice, GP). As noted, earlier patient follow-up after a hospital care episode was viewed by practice staff as challenging to implement and efforts to encourage patients to inform the practice of their hospital episode not always successful: “ … we try to tell the QUEST patients to ring us, to get a relative to ring us if they’ve been hospitalised, but they don’t because you don’t think of it, and they assume that the hospital is telling them. They rock up here and say, “I’ve been in hospital,” and you go, “Really?” (Intervention practice, PM). There were contrasting views amongst patients on the impact of the intervention on the general practice services they received. Some patients believed that there had been little or minimal changes noticeable during the intervention period. Patients with chronic and complex health conditions believed that due to their special circumstances the practice had always offered priority services including longer GP appointments and regular check-ups: “It’s always been really good, that surgery, which is why I’ve been there for 20 years.” (Male, 62 yrs. Old). “So it’s pretty hard to say that QUEST made him that way. I think it’s just in general, he’s just a good old fashioned GP who wants to spend time with the patient.” (Male, 57 yrs. Old). “Because I’m so complex and one thing can happen and I’ll just drop and be really sick, I’ve always been well first priority. Yeah so I always feel looked after and there’s support there if I need it.” (Female, 53 yrs. Old). On the other hand, there were those who felt that general practice services improved during the intervention period. These improvements included the improvements to the waiting times for appointments, an increased awareness with participant’s health and healthcare and better access to long GP appointments: “ … prior to it (the trial) we did have a couple of occasions where he was fully booked up for a week. I never got that once we started the QUEST program, yeah. Whenever you’d ring up and say “Okay next available is?” “Is it urgent?” “Yeah” so you’d get in if not that day, the following day. It’s encouraged them to lift their game.” (Male, 73 yrs. Old). “Being part of this program, I got enhanced medical treatment… I am really seriously ill and I guess this program allowed my surgery practice to actually streamline me.” (Male, 66 yrs. Old). “QUEST trial has made him [doctor] more aware that he has to be quite thorough, even though he is thorough, but I think it is in the back of his mind.” (Female, 74 yrs. Old). “I’ve felt less pressured to get in and get out. I’ve felt like it’s okay to come in, and take a breath, and say, “Okay, this is what I’m doing, and this is how I’m feeling… it’s been nice to be able to sit there with her and go, “Look, I’m coming down. I can feel my depression deepening.” To be free enough to talk to her about that without thinking, “I’ve got to be out of there in five minutes” So it’s been better in that respect.” (Female, 50 yrs. Old). The role of the PN in the trial was raised by several intervention group practice staff and patients. In some practices the PN had played a very limited role confined to patient recruitment while in other practices the PN had a more active role in implementing the intervention and this appeared to facilitate a more team based approach to care: “We trialled a pod kind of thing, so we allocated each patient with their own doctor, their own nurse, and their own admin team, so we tried to do the pod environment…we have a pharmacist on the team as well.” (Intervention practice, PN). This team-based approach was viewed as being a strong enabler to the successful implementation of the intervention as well as improving patient satisfaction and engagement: “I guess the QUEST patients loved having their nurse. They absolutely loved it. They take ownership. Like, “That’s my nurse.” … It was actually nice and then they’ve got someone that they can call. Sometimes they feel like they don’t want to bother the doctor, or it might be a silly problem. Then they know that they can just call up, have a conversation with the nurse.” (Intervention practice, GP). This model of team based care was viewed by practice staff as a more ‘ sustainable practice model’ that reduces ‘ burn-out’ and ensures continuity of care: ‘ continuity isn’t necessarily with the one person [doctor]. It’s with a team.’ (Intervention practice, GP) . Patients also noted a stronger involvement of PNs during the trial which was viewed positively: “The only change that I saw in there was with the nurse. – for the good, not bad. The nurse, you know, it came up, “You have an appointment with the nurse.” I go there and she take the blood pressure and measure the asthma thing, all the stuff. Some time I had appointment just with the nurse, and then she gathered up the stuff and put it in a folder, and they put it into the computer for the doctor. That was a new thing, that was a good thing.” (Female, 77 yrs. Old). “But I must admit I have, since the QUEST program I suppose, I’ve got even more time with also a nurse, check my blood pressure and my weight and anything like that. And then I go and see her [doctor].” (Female, 68 yrs. Old). Practice staff and patient views on the impact of the intervention on hospital service use Overall, from the practice staff and patient interviews it was difficult to establish a direct link between the intervention and hospital care episodes. Indirectly it appeared that for practice staff their participation in the trial and the fact that they (GPs) had identified trial patients as at high risk of poor health outcomes raised awareness for the potential to reduce the risk of avoidable hospitalisations: “I’ve thought of those patients differently. When they came I used to think, what could lead you into hospital and how can we avoid that.” (Intervention practice, GP) . From the patient perspective, however, the most frequently cited reason for hospital care as opposed to a GP appointment was to seek after-hours care: “We did call the ambulance a couple of times but that was because of the hour of the day and they [general practice] weren’t working, you know, it was early in the morning or something.” (Male, 69 yrs. Old). “Well, it’s often in the evening, I mean I know I’ve got access to the after-hours service, but how do I get there? That sort of thing, for me that’s an issue, my disability…much easier to call an ambulance. “ (Female, 63 yrs. Old). Other perceptions included receiving more comprehensive care in hospitals including access to specialists, radiology or other services that are not available in general practices: “The surgery couldn’t help me, the GP will see you straightaway but sometimes it’s like you try and book in for a scan or X-rays and things like that, you might have a two-week waiting list now and that sort of scared me a bit. I was thinking, you know, if I’ve got a blockage or something like that, or a twisted bowel. They [at hospital] do it straightaway.” (Female, 71 yrs. Old). A few patients highlighted financial issues as an incentive for attending hospitals instead of visiting a GP: “I mean, you go in there [general practice] and before they’d even look at you you’ve got to pay up – money up front. Even if you’re a private patient or if you’re – whatever you are you’ve still got to pay them cash up front. Then you sit and wait and they come along and have a look at you and send you off for an X-ray and so you go to an X-ray and that costs you about another three or four hundred.” (Male, 64 yrs. Old). Sustainability of the intervention Intervention group practice staff were generally positively disposed to continuing to provide the intervention after the trial had completed but it was acknowledged that providing the intervention without the additional funding from the trial would be difficult. For example, improvements to continuity of care (appointments with the preferred GP) had been facilitated during the trial by intervention group practices reserving appointment slots for trial participants. But if the appointment time was not booked this could financially disadvantage the GP and practice: “… we just don’t have free appointments to be able to hold a few back. Yeah, I don’t think it would be a sustainable thing for us. We would love to be able to do it, for the most vulnerable patients and have those appointments free to be able to do it, but financially doctors aren’t going to hold appointments back, just in case they don’t get booked.” (Intervention practice, PM). Similarly for long appointments, intervention group practices were asked not to charge co-payments to trial patients to ensure that patients were not financially disadvantaged when receiving long appointments which are often associated with higher co-payment charges. Practice staff reported that bulkbilling for long appointments would be challenging to sustain and as a result the number of long appointments would likely reduce: “We will continue with a lot of those interventions, as normal. But the double appointments might be a different story. But we do have patients that aren’t specific QUEST patients who it is highlighted to preferred double appointments, and that is based on acuity and how chronic they are and problems going on. So that practice may reduce somewhat with those extra, longer appointments .” (Intervention practice, PN). Practices that had used the trial funding to support a greater emphasis on team based care also indicated that this would be difficult to sustain financially: The funding will prevent us from doing so. We employed more nurses, so we’ve got more nursing staff, but if we can’t fund that then where do we go? But the nurses take the pressure off the doctors and the doctors are under enormous pressure, so you want to maintain your doctors, stop them burning out because that’s not good for anyone, and also make sure you’ve got patients coming in. The doctor workforce is going to be tighter and tighter, so operating a good practise where you look after them and keeping them is key, from a workforce point of view.” (Intervention practice, PM) .
Continuity of care Practice staff valued GP continuity of care for service quality, trust and building relationships: “Continuity of care is that you don’t have to cover off that big chunk of their life every time. You just know what’s going on, just by them walking in the room before they even said anything…You get that connection with people. They will always tell you more with that connection. You’ve got more time because you don’t have to cover all that other history that you already know. Yeah, you can kind of nuance things a bit more, because you can pick the little things out that aren’t quite right with what’s going on.” (Intervention practice, GP). While continuity of care with a regular GP was generally supported, some practice staff drew a distinction between patients being ‘loyal to the practice’ versus being ‘loyal to a regular GP’. Strategies such as communication between GPs within a practice, systems in place enabling GPs to share patient’s notes and information and having a second regular GP were viewed by some as more likely to lead to a sustainable general practice model: “I should be selling the practice, the quality practice where all the doctors write good notes and knowledgeable caring and are on the same page. That’s what I’d like to do is sell the practice rather than an individual GP because if I go away for three weeks or something then what do they do? We all have holidays.” (Intervention practice, GP). There were also perceived clinical benefits from having a variety of different GPs involved in patient care: “I don’t want any of my patients to be absolutely dependent on me. I actually think it’s healthy for them to see another doctor because I might miss something that another doctor picks up. I might have some level of expertise in one area and another doctor might have a level of expertise in another area, so I actually think that concept of a preferred GP needs modification.” (Intervention practice, GP). A PM also commented: “You sometimes get a patient so dependent on one doctor that if that doctor’s away, they won’t see anyone else. To me, that’s putting their health at risk. It’s hard sometimes to convince people that, ‘Yes, even though he’s away, please come and see’ – and sometimes another doctor can shine a different light on the problem.” (Control practice, PM). Continuity of care was valued by patients for largely the same reasons as practice staff. Patients believed it enabled them to build trusted relationships with their GP and helped their GP to better understand their family and social circumstances: “Well, familiarity I guess and just I suppose the information, the things that you discuss with him then, I mean he’s got a fuller picture of a person as a patient, rather than just a random doctor here and there. They don’t have the whole picture.” (Female, 75 yrs. Old). Continuity of care was particularly valued by patients with complex health problems, mental illnesses or for other ‘personal’ issues: “I said to her [doctor] I didn’t feel like having to explain everything to everybody; especially when you’re losing blood from the bowel region, it’s embarrassing.” (Male, 66 yrs. Old). Several patients also raised the idea of a second regular GP or a practice-based GP service. For some, particularly those with long-term connections with their general practice, being able to see any GP within the practice was a way to fill in the gaps in care continuity with their preferred GP. Patients also reported the benefits of visits with other GPs, for example, consultations with a female GP for gender specific health issues or screenings or with GPs who are specialised in specific health conditions such as skin cancer. Patient perceived barriers to continuity of care included GPs who worked part-time, particularly, younger aged female doctors and GPs absences due to sickness. Longer appointments Longer appointments were viewed by practice staff as being particularly advantageous for patients with chronic and complex health conditions because they enabled more comprehensive care to be provided: “Rather than just looking at what the presenting complaint may be, it’s doing a full thorough check on someone. It’s being able to take that time to properly talk to them, to properly give education, to provide information, make sure that the understanding is there and answer any questions and that sort of thing, which is really important.” (Intervention practice, PN). On the other hand, some practice staff noted that patients with chronic and complex health conditions were more likely to have frequent appointments and this might be more beneficial compared with less frequent longer appointments: “If they’re seen once a month, a 15-minute appointment once a month that would be more valuable to them than a half-hour appointment every three months…and if they’re checking in with the nurse every three months, monthly 15-minute appointment with the doctor, they’re going to get everything that they need.” (Intervention practice, PM). Some GPs noted that the current Medicare rebate structure discourages longer appointments, and this requires changes at broader policy and system levels: “The way the schedule fee is based, you’re actually rewarded for shorter appointments rather than longer appointments. I think ultimately that needs to change, because the population at large is getting older, the complexity of patients is getting more difficult, and patients are becoming more knowledgeable. They don’t want to come in and just be given a script and told to do this or do that, they want more of an explanation of what’s going on and the rest of it.” (Intervention practice, GP). Longer appointments were valued by patients because it was felt they provided an opportunity to discuss complex health problems thoroughly (especially in the case of mental health issues), and to review medications. Longer appointment times were also thought to allow time for more questions to be asked. One patient described standard length appointments as: “… I think this business about working on a 15-minute appointment, to my mind, doesn’t work well, right…I call it shop medicine, right, you go in and you buy something and you go out.” (Male, 89 yrs. Old). However, patients also reported that appointment length should be based on need and that a long appointment was not always required for example for prescription renewals, test results or for simple problems: “Sometimes it’s not needed. Sometimes I think 30 minutes was too much, but there was other times when it was good to have that extra time.” (Female, 58 yrs. Old). Patients also raised the importance of doctor-patient communication as being equally important to appointment length: “What is important is the communication between the doctor and patient. You may have a longer appointment but if the doctor doesn’t communicate well, the longer appointment does not work.” (Male, 69 yrs. Old). Timely follow-up after a hospital care episode Patient follow-up after a hospital care episode was considered by practice staff to be valuable but challenging to achieve consistently due to poor communication between hospitals and general practices which led to delays in practices receiving patient discharge summaries: “I think the biggest thing is that the discharge information, once they’ve been in hospital and struggling to pin people down.” (Intervention practice, PM). “It really disrupts the continuity because eventually you’ll say, they need to have follow up bloods done three days post discharge. Well, they’ve been home for a week now.” (Intervention practice, PN). “Private hospitals don’t have their own resident medical staff. And it’s a medical handover issue, they don’t see that as their problem. As opposed to a public hospital where they actually have a resident medical staff so you can address the concern or something to that staff as an entity. And that’s the problem I think, that’s where it breaks down unfortunately.” (Intervention practice, GP). Similarly, patients reported examples of their GP not being aware of their hospital care episode: “He [GP] said “When did you get back?” and I said “Mate, I’ve just been in hospital”… I told him! So you know, that’s where the system falls down.” (Male, 65 yrs. Old). Another patient reported: “I think there is a problem with the hospital and their follow-ups. Whenever I came out from a stay in hospital, I’d do a follow-up appointment with the doctors and usually they’d have no idea that you’ve been in there. They haven’t received follow-up reports or anything like that… the run of the mill is that you’re told you should follow-up within a fortnight of leaving hospital. So we’d make that appointment and you’d see them [GP] and they’d say “Oh we haven’t received anything” so there is a lag time there.” (Male, 69 yrs. Old).
Practice staff valued GP continuity of care for service quality, trust and building relationships: “Continuity of care is that you don’t have to cover off that big chunk of their life every time. You just know what’s going on, just by them walking in the room before they even said anything…You get that connection with people. They will always tell you more with that connection. You’ve got more time because you don’t have to cover all that other history that you already know. Yeah, you can kind of nuance things a bit more, because you can pick the little things out that aren’t quite right with what’s going on.” (Intervention practice, GP). While continuity of care with a regular GP was generally supported, some practice staff drew a distinction between patients being ‘loyal to the practice’ versus being ‘loyal to a regular GP’. Strategies such as communication between GPs within a practice, systems in place enabling GPs to share patient’s notes and information and having a second regular GP were viewed by some as more likely to lead to a sustainable general practice model: “I should be selling the practice, the quality practice where all the doctors write good notes and knowledgeable caring and are on the same page. That’s what I’d like to do is sell the practice rather than an individual GP because if I go away for three weeks or something then what do they do? We all have holidays.” (Intervention practice, GP). There were also perceived clinical benefits from having a variety of different GPs involved in patient care: “I don’t want any of my patients to be absolutely dependent on me. I actually think it’s healthy for them to see another doctor because I might miss something that another doctor picks up. I might have some level of expertise in one area and another doctor might have a level of expertise in another area, so I actually think that concept of a preferred GP needs modification.” (Intervention practice, GP). A PM also commented: “You sometimes get a patient so dependent on one doctor that if that doctor’s away, they won’t see anyone else. To me, that’s putting their health at risk. It’s hard sometimes to convince people that, ‘Yes, even though he’s away, please come and see’ – and sometimes another doctor can shine a different light on the problem.” (Control practice, PM). Continuity of care was valued by patients for largely the same reasons as practice staff. Patients believed it enabled them to build trusted relationships with their GP and helped their GP to better understand their family and social circumstances: “Well, familiarity I guess and just I suppose the information, the things that you discuss with him then, I mean he’s got a fuller picture of a person as a patient, rather than just a random doctor here and there. They don’t have the whole picture.” (Female, 75 yrs. Old). Continuity of care was particularly valued by patients with complex health problems, mental illnesses or for other ‘personal’ issues: “I said to her [doctor] I didn’t feel like having to explain everything to everybody; especially when you’re losing blood from the bowel region, it’s embarrassing.” (Male, 66 yrs. Old). Several patients also raised the idea of a second regular GP or a practice-based GP service. For some, particularly those with long-term connections with their general practice, being able to see any GP within the practice was a way to fill in the gaps in care continuity with their preferred GP. Patients also reported the benefits of visits with other GPs, for example, consultations with a female GP for gender specific health issues or screenings or with GPs who are specialised in specific health conditions such as skin cancer. Patient perceived barriers to continuity of care included GPs who worked part-time, particularly, younger aged female doctors and GPs absences due to sickness.
Longer appointments were viewed by practice staff as being particularly advantageous for patients with chronic and complex health conditions because they enabled more comprehensive care to be provided: “Rather than just looking at what the presenting complaint may be, it’s doing a full thorough check on someone. It’s being able to take that time to properly talk to them, to properly give education, to provide information, make sure that the understanding is there and answer any questions and that sort of thing, which is really important.” (Intervention practice, PN). On the other hand, some practice staff noted that patients with chronic and complex health conditions were more likely to have frequent appointments and this might be more beneficial compared with less frequent longer appointments: “If they’re seen once a month, a 15-minute appointment once a month that would be more valuable to them than a half-hour appointment every three months…and if they’re checking in with the nurse every three months, monthly 15-minute appointment with the doctor, they’re going to get everything that they need.” (Intervention practice, PM). Some GPs noted that the current Medicare rebate structure discourages longer appointments, and this requires changes at broader policy and system levels: “The way the schedule fee is based, you’re actually rewarded for shorter appointments rather than longer appointments. I think ultimately that needs to change, because the population at large is getting older, the complexity of patients is getting more difficult, and patients are becoming more knowledgeable. They don’t want to come in and just be given a script and told to do this or do that, they want more of an explanation of what’s going on and the rest of it.” (Intervention practice, GP). Longer appointments were valued by patients because it was felt they provided an opportunity to discuss complex health problems thoroughly (especially in the case of mental health issues), and to review medications. Longer appointment times were also thought to allow time for more questions to be asked. One patient described standard length appointments as: “… I think this business about working on a 15-minute appointment, to my mind, doesn’t work well, right…I call it shop medicine, right, you go in and you buy something and you go out.” (Male, 89 yrs. Old). However, patients also reported that appointment length should be based on need and that a long appointment was not always required for example for prescription renewals, test results or for simple problems: “Sometimes it’s not needed. Sometimes I think 30 minutes was too much, but there was other times when it was good to have that extra time.” (Female, 58 yrs. Old). Patients also raised the importance of doctor-patient communication as being equally important to appointment length: “What is important is the communication between the doctor and patient. You may have a longer appointment but if the doctor doesn’t communicate well, the longer appointment does not work.” (Male, 69 yrs. Old).
Patient follow-up after a hospital care episode was considered by practice staff to be valuable but challenging to achieve consistently due to poor communication between hospitals and general practices which led to delays in practices receiving patient discharge summaries: “I think the biggest thing is that the discharge information, once they’ve been in hospital and struggling to pin people down.” (Intervention practice, PM). “It really disrupts the continuity because eventually you’ll say, they need to have follow up bloods done three days post discharge. Well, they’ve been home for a week now.” (Intervention practice, PN). “Private hospitals don’t have their own resident medical staff. And it’s a medical handover issue, they don’t see that as their problem. As opposed to a public hospital where they actually have a resident medical staff so you can address the concern or something to that staff as an entity. And that’s the problem I think, that’s where it breaks down unfortunately.” (Intervention practice, GP). Similarly, patients reported examples of their GP not being aware of their hospital care episode: “He [GP] said “When did you get back?” and I said “Mate, I’ve just been in hospital”… I told him! So you know, that’s where the system falls down.” (Male, 65 yrs. Old). Another patient reported: “I think there is a problem with the hospital and their follow-ups. Whenever I came out from a stay in hospital, I’d do a follow-up appointment with the doctors and usually they’d have no idea that you’ve been in there. They haven’t received follow-up reports or anything like that… the run of the mill is that you’re told you should follow-up within a fortnight of leaving hospital. So we’d make that appointment and you’d see them [GP] and they’d say “Oh we haven’t received anything” so there is a lag time there.” (Male, 69 yrs. Old).
The extent to which the intervention had provided general practice care that was different from usual care was one of the key themes that emerged from the practice staff interviews. Noting that the patients in the trial had been identified as at high risk of poor outcomes, practice staff often felt that they had strategies in place prior to the trial to ensure high levels of continuity of care and access to longer appointments: “It was along the same path of what we’ve already done, been trying to do something to help the target group in this case… If we managed to get into intervention, then the work that we do with our patients is not dissimilar to what we do already, in terms of that patient care and trying to give that extra bit which the QUEST was all about.” (Control practice, PM). “We are already doing the job, but now we had more incentive, we’ve got appointments and more nurses and staff were knowing that we need to take care of people, those people better and more efficiently… Yeah, special treatment, it made it a bit more systematic.” (Intervention practice, GP). There was a common perception that the intervention had largely been a part of usual care and therefore there were not significant changes because of the trial: “As a practice, that’s how we like to run things anyway. We’re very chronic-diseased focused, which you’d want to see the same GP if you can, not always possible, but most of the time we will try and make that happen anyway.” (Intervention practice, PM) . “We’re already implementing a lot of things before you even start. So to some degree it was a bit the icing on the cake.” (Intervention practice, GP) . On the other hand, some intervention group practice staff reported that participating in the trial had increased their level of awareness about proactively addressing the needs of high risk patients: “I’ve become more proactive with patients’ problems. Not only with the patients registered in the trial, but even with other patients coming in with similar problems. I initiated the same sort of practices so that they can benefit also, like giving them more time, looking into more preventative care before the problem started.” (Intervention practice, GP) . The benefits of longer appointment times were appreciated by some GPs who noted that: “It gives you time to not rush and be able to let them open up and talk about what’s going on, and look up their bloods and bone densities and all those kinds of things that in 15 minute-appointments, I find it to be too short, sharp. Because 15 is really 10 to 12, by the time you get notes and things done, it doesn’t give a lot of time. So, personally, double appointments I think are more beneficial, and reduce stress levels as well.” (Intervention practice, GP). As noted, earlier patient follow-up after a hospital care episode was viewed by practice staff as challenging to implement and efforts to encourage patients to inform the practice of their hospital episode not always successful: “ … we try to tell the QUEST patients to ring us, to get a relative to ring us if they’ve been hospitalised, but they don’t because you don’t think of it, and they assume that the hospital is telling them. They rock up here and say, “I’ve been in hospital,” and you go, “Really?” (Intervention practice, PM). There were contrasting views amongst patients on the impact of the intervention on the general practice services they received. Some patients believed that there had been little or minimal changes noticeable during the intervention period. Patients with chronic and complex health conditions believed that due to their special circumstances the practice had always offered priority services including longer GP appointments and regular check-ups: “It’s always been really good, that surgery, which is why I’ve been there for 20 years.” (Male, 62 yrs. Old). “So it’s pretty hard to say that QUEST made him that way. I think it’s just in general, he’s just a good old fashioned GP who wants to spend time with the patient.” (Male, 57 yrs. Old). “Because I’m so complex and one thing can happen and I’ll just drop and be really sick, I’ve always been well first priority. Yeah so I always feel looked after and there’s support there if I need it.” (Female, 53 yrs. Old). On the other hand, there were those who felt that general practice services improved during the intervention period. These improvements included the improvements to the waiting times for appointments, an increased awareness with participant’s health and healthcare and better access to long GP appointments: “ … prior to it (the trial) we did have a couple of occasions where he was fully booked up for a week. I never got that once we started the QUEST program, yeah. Whenever you’d ring up and say “Okay next available is?” “Is it urgent?” “Yeah” so you’d get in if not that day, the following day. It’s encouraged them to lift their game.” (Male, 73 yrs. Old). “Being part of this program, I got enhanced medical treatment… I am really seriously ill and I guess this program allowed my surgery practice to actually streamline me.” (Male, 66 yrs. Old). “QUEST trial has made him [doctor] more aware that he has to be quite thorough, even though he is thorough, but I think it is in the back of his mind.” (Female, 74 yrs. Old). “I’ve felt less pressured to get in and get out. I’ve felt like it’s okay to come in, and take a breath, and say, “Okay, this is what I’m doing, and this is how I’m feeling… it’s been nice to be able to sit there with her and go, “Look, I’m coming down. I can feel my depression deepening.” To be free enough to talk to her about that without thinking, “I’ve got to be out of there in five minutes” So it’s been better in that respect.” (Female, 50 yrs. Old). The role of the PN in the trial was raised by several intervention group practice staff and patients. In some practices the PN had played a very limited role confined to patient recruitment while in other practices the PN had a more active role in implementing the intervention and this appeared to facilitate a more team based approach to care: “We trialled a pod kind of thing, so we allocated each patient with their own doctor, their own nurse, and their own admin team, so we tried to do the pod environment…we have a pharmacist on the team as well.” (Intervention practice, PN). This team-based approach was viewed as being a strong enabler to the successful implementation of the intervention as well as improving patient satisfaction and engagement: “I guess the QUEST patients loved having their nurse. They absolutely loved it. They take ownership. Like, “That’s my nurse.” … It was actually nice and then they’ve got someone that they can call. Sometimes they feel like they don’t want to bother the doctor, or it might be a silly problem. Then they know that they can just call up, have a conversation with the nurse.” (Intervention practice, GP). This model of team based care was viewed by practice staff as a more ‘ sustainable practice model’ that reduces ‘ burn-out’ and ensures continuity of care: ‘ continuity isn’t necessarily with the one person [doctor]. It’s with a team.’ (Intervention practice, GP) . Patients also noted a stronger involvement of PNs during the trial which was viewed positively: “The only change that I saw in there was with the nurse. – for the good, not bad. The nurse, you know, it came up, “You have an appointment with the nurse.” I go there and she take the blood pressure and measure the asthma thing, all the stuff. Some time I had appointment just with the nurse, and then she gathered up the stuff and put it in a folder, and they put it into the computer for the doctor. That was a new thing, that was a good thing.” (Female, 77 yrs. Old). “But I must admit I have, since the QUEST program I suppose, I’ve got even more time with also a nurse, check my blood pressure and my weight and anything like that. And then I go and see her [doctor].” (Female, 68 yrs. Old).
Overall, from the practice staff and patient interviews it was difficult to establish a direct link between the intervention and hospital care episodes. Indirectly it appeared that for practice staff their participation in the trial and the fact that they (GPs) had identified trial patients as at high risk of poor health outcomes raised awareness for the potential to reduce the risk of avoidable hospitalisations: “I’ve thought of those patients differently. When they came I used to think, what could lead you into hospital and how can we avoid that.” (Intervention practice, GP) . From the patient perspective, however, the most frequently cited reason for hospital care as opposed to a GP appointment was to seek after-hours care: “We did call the ambulance a couple of times but that was because of the hour of the day and they [general practice] weren’t working, you know, it was early in the morning or something.” (Male, 69 yrs. Old). “Well, it’s often in the evening, I mean I know I’ve got access to the after-hours service, but how do I get there? That sort of thing, for me that’s an issue, my disability…much easier to call an ambulance. “ (Female, 63 yrs. Old). Other perceptions included receiving more comprehensive care in hospitals including access to specialists, radiology or other services that are not available in general practices: “The surgery couldn’t help me, the GP will see you straightaway but sometimes it’s like you try and book in for a scan or X-rays and things like that, you might have a two-week waiting list now and that sort of scared me a bit. I was thinking, you know, if I’ve got a blockage or something like that, or a twisted bowel. They [at hospital] do it straightaway.” (Female, 71 yrs. Old). A few patients highlighted financial issues as an incentive for attending hospitals instead of visiting a GP: “I mean, you go in there [general practice] and before they’d even look at you you’ve got to pay up – money up front. Even if you’re a private patient or if you’re – whatever you are you’ve still got to pay them cash up front. Then you sit and wait and they come along and have a look at you and send you off for an X-ray and so you go to an X-ray and that costs you about another three or four hundred.” (Male, 64 yrs. Old).
Intervention group practice staff were generally positively disposed to continuing to provide the intervention after the trial had completed but it was acknowledged that providing the intervention without the additional funding from the trial would be difficult. For example, improvements to continuity of care (appointments with the preferred GP) had been facilitated during the trial by intervention group practices reserving appointment slots for trial participants. But if the appointment time was not booked this could financially disadvantage the GP and practice: “… we just don’t have free appointments to be able to hold a few back. Yeah, I don’t think it would be a sustainable thing for us. We would love to be able to do it, for the most vulnerable patients and have those appointments free to be able to do it, but financially doctors aren’t going to hold appointments back, just in case they don’t get booked.” (Intervention practice, PM). Similarly for long appointments, intervention group practices were asked not to charge co-payments to trial patients to ensure that patients were not financially disadvantaged when receiving long appointments which are often associated with higher co-payment charges. Practice staff reported that bulkbilling for long appointments would be challenging to sustain and as a result the number of long appointments would likely reduce: “We will continue with a lot of those interventions, as normal. But the double appointments might be a different story. But we do have patients that aren’t specific QUEST patients who it is highlighted to preferred double appointments, and that is based on acuity and how chronic they are and problems going on. So that practice may reduce somewhat with those extra, longer appointments .” (Intervention practice, PN). Practices that had used the trial funding to support a greater emphasis on team based care also indicated that this would be difficult to sustain financially: The funding will prevent us from doing so. We employed more nurses, so we’ve got more nursing staff, but if we can’t fund that then where do we go? But the nurses take the pressure off the doctors and the doctors are under enormous pressure, so you want to maintain your doctors, stop them burning out because that’s not good for anyone, and also make sure you’ve got patients coming in. The doctor workforce is going to be tighter and tighter, so operating a good practise where you look after them and keeping them is key, from a workforce point of view.” (Intervention practice, PM) .
This study has reported the experiences of general practice staff and their patients who took part in a clustered randomised controlled trial of a multicomponent general practice intervention. The elements of the multicomponent intervention were broadly supported by practice staff and patients who recognised benefits for patient care from appointments with a regular GP, having sufficient time in appointments to discuss complex health problems and receiving timely general practice follow-up after hospital care episodes. That support however was not unqualified with practice staff raising potential advantages for the involvement of different GPs in a patient’s care and poorer outcomes for patients who might inappropriately delay treatment until they could make an appointment with their preferred GP. From the patient perspective, particularly those with complex health problems, continuity of care was highly valued but at that same time it was appreciated that in practice it was unlikely that appointments were always going to be able to be made with one’s preferred GP. Both practice staff and patients recognised that not all GP appointments needed to be longer length and that for people with chronic illnesses there might be benefit from more frequently occurring pre-planned standard length appointments. Providing timely (within 7 days) follow-up after patients experienced a hospital care episode was reported by practice staff as challenging. General practice follow-up after hospital care episodes involves complex dependencies between the patient, the GP and the hospital . Practices rely on discharge information being sent to them from hospitals or patients independently notifying them of their hospital care episode. But even when informed, intervention group practices were required to take a proactive approach initially to assess whether a follow-up appointment was clinically warranted and then to contact the patient to discuss whether they wished to attend for a face-to-face appointment. From the patient perspective there appeared to be element of surprise that there was not a seamless transfer of information between hospitals and general practices and that they themselves shared some of the responsibility for informing their general practices of any hospital care episodes. That intervention group practice staff experienced challenges ensuring timely follow-up of patients after a hospital care episode is not surprising given prior research both in Australia and internationally identifying the timeliness and quality of discharge letters as potential sources of conflict between GPs and hospital doctors. From the GP and patient perspectives very clear follow-up instructions are required to facilitate the provision of appropriate care to patients after their discharge from hospital . The assumption of a seamless transfer of information between hospital and general practices is a risky one, particularly given that systematic reviews of the quality of discharge letters have found that key components including information about follow-up and management plans are often lacking . In light of this initiatives designed to empower patients to have a better understanding of their follow-up recommendations through a simplified set of discharge instructions are likely to be of potential value . Many practice staff and patients reported that the intervention did not provide markedly different care than usual. Practice staff indicated that for the patients in the trial, who had been identified as at high risk of poor health outcomes, they had existing strategies in place to ensure high levels of continuity of care and patients had always had ready access to longer appointments. Similarly, patients often did not report major changes from usual care, but some reported a higher degree of thoroughness taken by their GP, more timely appointments, being routinely offered longer appointments and a greater involvement of the PN in their care. These findings are consistent with the quantitative results from the process indicators collected during the trial that showed only limited improvements to continuity of care, appointment length, and hospital follow-up. The role of the PN in the trial was variable. In some intervention group practices the PN role was confined to assisting with patient recruitment and data collection. In other practices the PN took on a coordinating role in the implementation of the intervention. For two intervention group practices the trial appear to act as a catalyst for a more formalised team based approach and there was an increase in the amount of contact time between patients and PNs. A team based approach generally was thought by practice staff to a be a more sustainable model for general practice. This is consistent with the view that team based care is a critical component for high quality primary care . From the patient perspective the adoption of a more team based approach was viewed very positively and indeed for some patients, the most noticeable (and positive) change for them during the trial was increased level of PN care that they received. This finding is broadly consistent with prior Australian research showing that patients who attended practices where PNs worked with broad scopes of practice were more likely to be more satisfied than those attending practices where PNs worked with narrow scopes of practice and low levels of autonomy . While a focus on team based care was not one of the four elements of the multicomponent intervention, the significant payments made to intervention group practices supported practices to increase the amount of PN care provided to trial patients. The trial payments were A$1,000 per intervention group patient and with an average of around 50 patients per practice this equated to average practice payments in the order of A$50,000 (noting that the two largest intervention group practices had around 100 patients each). These payments were addition to the usual Medicare fee for service that practices received. For the duration of the trial, for some patients at the practice (i.e. those in the trial), the trial payments reduced the dominance of the fee for service payment structure which has been identified as a barrier to implementing team-oriented primary care . In the economic evaluation, the intervention was found to be cost-effective in a pre-specified exploratory analysis of older patients primarily due to a reduction in hospital usage. The practice staff qualitative interviews suggested that the act of identifying a sub-set of patients as at high risk of poor health outcomes and being aware that hospital was an important outcome in the trial heightened awareness to the risk of potentially avoidable hospitalisations for trial patients. The exact mechanism through which this heightened awareness may have translated to reductions in hospitalisations however is unclear. From the patient perspective the most frequently cited reasons for hospital care as opposed to a GP appointment were to seek after-hours care and to receive a more comprehensive and timely level of care could be provided in the hospital setting. Patients also nominated financial considerations observing that public hospital services are usually provided at low (or nil) direct cost to patients. Notably these key patient-nominated drivers of hospital use were not addressed by the intervention. Practice staff expressed a general desire to continue to provide the intervention after the trial completed but they also indicated that changes they had made to operationalise the intervention in their practice, such as reserving appointments for trial patients to facilitate continuity of care, would be difficult to sustain financially when the significant trial payments ceased. The general feeling of many practice staff interviewed was that the financial challenges associated with implementing practice change were often underappreciated by people outside the general practice setting. Importantly the concerns by practice staff about the financial challenges associated with practice change were expressed just prior to the COVID-19 pandemic which appeared in South Australia within weeks of the completion of most of the qualitative interviews. The impact of COVID-19 on general practices across Australia was substantial and included a decrease in the number of face to face to consultations, the rapid adoption of telehealth consultations, workforce shortages and financial pressures . Chronic disease management is likely to have suffered during the early phases of the pandemic due to general practice service disruptions and patients with chronic diseases avoiding healthcare appointments to minimise the risk of contracting COVID-19 . It is unknown how COVID-19 impacted on the individual patients and practice staff that were interviewed for the present study and whether post the initial phases of COVID-19 they may have different perceptions about the value of the intervention. The Australian government’s Strengthening Medicare report published in April 2023 recommended a broad primary care reform agenda including several elements to the Flinders QUEST multicomponent intervention. These included encouraging GP continuity of care through the introduction of voluntary patient registration, additional funding for longer consultations, supporting multidisciplinary team-based care and better integration and coordination between primary and secondary care. The findings of the present study suggest that these reforms would enjoy broad support from patients and qualified support from general practice staff on the proviso that any reforms were adequately funded. Limitations The key limitations of this study relate to the generalisability of the findings. The practice staff who took part in Flinders QUEST were from practices who were part of an academic practice research network and who at the time of trial had the capacity to make changes to their systems of care. In addition, the GPs and PNs interviewed were nominated by the Practice Manager and their views may not be representative of clinical staff at their practice. Trial patients were drawn from a general practice patient population at risk of poor health outcomes, and this led a high proportion of older people who tended to have a long term relationship with their practice and GP in the trial. For the interviews this was further refined to those who had reported a recent hospital care episode. Younger patients, those without a regular GP, and in good health, might expected to have experienced the intervention very differently. Finally, the intervention itself was designed around perceived weaknesses in Australia’s primary health care system and the findings reported here may not be applicable in other countries.
The key limitations of this study relate to the generalisability of the findings. The practice staff who took part in Flinders QUEST were from practices who were part of an academic practice research network and who at the time of trial had the capacity to make changes to their systems of care. In addition, the GPs and PNs interviewed were nominated by the Practice Manager and their views may not be representative of clinical staff at their practice. Trial patients were drawn from a general practice patient population at risk of poor health outcomes, and this led a high proportion of older people who tended to have a long term relationship with their practice and GP in the trial. For the interviews this was further refined to those who had reported a recent hospital care episode. Younger patients, those without a regular GP, and in good health, might expected to have experienced the intervention very differently. Finally, the intervention itself was designed around perceived weaknesses in Australia’s primary health care system and the findings reported here may not be applicable in other countries.
The multicomponent intervention was supported by practice staff and patients who appreciated the value of GP continuity of care, longer GP appointment times and GP follow-up after a hospital care episode. Some patients reported an improvement in their care during the trial, but many did not notice significant changes from the intervention. Practice staff generally viewed the intervention as providing usual care but in a more systematic and rigorous manner. Practice staff expressed a desire to continue the intervention after the trial had completed but noted that it would be difficult to sustain financially.
|
Thoracoabdominal Normothermic Regional Perfusion and Donation After Circulatory Death Lung Use | 4acb2fb2-2c42-44a3-b926-c15d0ebd642f | 11833517 | Surgical Procedures, Operative[mh] | Donation after circulatory death (DCD) heart transplantation has increased in recent years and offers an important opportunity to expand the donor pool for cardiac transplantation. Growth in DCD heart transplantation can be partially attributed to improved procurement techniques broadly classified as thoracoabdominal normothermic regional perfusion (TA-NRP) or direct procurement, followed by either static cold storage or ex vivo machine perfusion. , In both approaches, expeditious heart procurement is prioritized over lung procurement, as the heart is more vulnerable to ischemic injury. The organ procurement organization (OPO) and transplant center decide jointly to perform TA-NRP vs direct procurement prior to organ procurement. Prior work from members of our group suggests that compared with direct procurement, TA-NRP may be associated with suboptimal lung allograft function in the immediate postoperative period but similar graft survival at 1 year. Despite these data supporting conscientious use of DCD lung allografts after TA-NRP, the effect of concomitant heart procurement on overall DCD lung use rates remains unclear. Rates of DCD lung use have been increasing annually, making comparison with historical controls biased. Therefore, contemporaneous and controlled comparison is necessary. Several aspects of concomitant heart procurement may injure lungs from DCD donors, potentially discouraging otherwise viable allografts. First, quick entry to the thoracic cavity alongside DCD donor coagulopathy may cause excessive bleeding, necessitating high-volume transfusion, which is associated with recipient primary graft dysfunction and mortality. , Second, while use of venoarterial extracorporeal membrane oxygenation (VA-ECMO) has been associated with left ventricular dilation and pulmonary congestion in other settings, this use remains a theoretical concern with TA-NRP, in which VA-ECMO is sometimes used. Furthermore, variations in procuring team experience with TA-NRP may contribute to differences in lung quality and the decision to use the lungs for transplantation. Despite these concerns, several studies in the US have demonstrated comparable rates of DCD lung use with use of TA-NRP and direct procurement. , , However, these studies are limited by small cohorts and have not adjusted for donor risk factors, highlighting the need for larger, risk-adjusted analyses to fully address this question. To better characterize the association of simultaneous heart procurement and TA-NRP with DCD lung use, we conducted a risk-adjusted analysis using a national cohort of DCD donors. Specifically, we identified controlled DCD donors in the United Network for Organ Sharing (UNOS) database and examined the observed to expected (O:E) ratios of lung use among 4 groups: noncardiac DCD donors, cardiac DCD donors, and, via subgroup analysis, cardiac DCDs procured with TA-NRP vs direct procurement. We hypothesized that cardiac donation is adversely associated with lung DCD use and that TA-NRP is associated with lower DCD lung use compared with direct procurement. Data Source We performed a retrospective cohort study using the UNOS Scientific Registry of Transplant Recipients (SRTR) database, which includes all organ donors in the US since October 1, 1987. The reporting of results adheres to the guidelines set out in the Strengthening the Reporting of Observational Studies in Epidemiology ( STROBE ) reporting guideline. This study was approved by the Institutional Review Board at Duke University, which deemed the study not human participants research, and thus the requirement for obtaining informed consent was waived. Study Population All controlled DCD donors were considered for inclusion. Donors before January 1, 2019, were excluded, as the first DCD heart transplant in the US occurred in 2019. Data through September 30, 2024, were collected. We excluded non-DCD donors, donors under the age of 18 years, and donors with unclear lung disposition. Study Design The primary objective of this analysis was to evaluate the association of DCD heart procurement with lung use for transplantation. The cohort was stratified into donors from whom a heart was recovered (cardiac donor), regardless of whether the heart was transplanted, and donors from whom a heart was not recovered (noncardiac donor). Rates of DCD lung use were compared between cardiac DCD and noncardiac DCD groups. The second objective of this study was to evaluate the association of the DCD heart procurement technique (TA-NRP or direct procurement) with lung use. As procurement technique is not discretely captured in the dataset, the interval from time of death to time of thoracic aorta cold flush was used to define procurement technique. Within the cardiac DCD cohort, donors with a time interval of greater than 15 minutes were defined as TA-NRP; those with elapsed time less than 15 minutes were defined as direct procurement, as previously described. Donors with missing aorta clamp time or time of death were excluded from further analysis. Unadjusted rates of DCD lung use were compared between TA-NRP and direct procurement groups. Estimating Expected Organ Yield The SRTR releases a risk-adjusted estimate of expected donor yield twice annually that assesses the likelihood of successful transplantation for a specific donor’s organs. Using logistic regression modeling, risk-adjusted estimates of organ-specific donor yield are developed using the most current data collected by the Organ Procurement and Transplantation Network (OPTN). These models have been validated and used in the literature to calculate expected yield of organs transplanted but, to our knowledge, have not been applied in lung transplantation. , , Once a model is chosen for a specific organ, the organ-specific expected yield is calculated algebraically using the model coefficients publicly shared by the SRTR and the OPTN UNOS database. In this way, lung use from each donor is adjusted for donor characteristics, allowing for comparison between observed and expected rates of lung use. For this analysis, we used the most recent model released in January 2024. The observed yield was divided by the expected yield. A ratio of 1 indicated equal observed yield and expected yield; ratios greater or less than 1 suggested that observed yield exceeded or fell short of the expected yield, respectively. We sought to determine whether the observed organ yield in each group was statistically different from the expected yield and from the observed yield in the comparator groups. To do so, we computed the O:E ratio as observed yield divided by expected yield. Bootstrapping techniques were employed to develop 95% CIs and 2-sided P values around the O:E ratios under a null hypothesis positing that observed yield was exactly equal to expected yield (ie, O:E ratio of 1.0). We used 1000 bootstrapped samples for all cohorts. Subanalysis The TA-NRP use among OPOs was also characterized to better understand variations in practice. Trends in TA-NRP use and O:E ratios of DCD lung use were examined over time using the Cochran-Armitage test. Statistical Analysis Nonparametric Kruskal-Wallis and χ 2 tests were used to compare donor and recipient characteristics as well as perioperative outcomes for lung transplant recipients. Fisher exact tests were used if sample sizes did not meet the assumptions for χ 2 testing. The Kaplan-Meier log-rank survival analysis was used to evaluate recipient mortality. As already described, bootstrapping was used to calculate 95% CIs and P values for reported O:E ratios. Statistical analyses were performed using RStudio, version 4.2.3 (R Project for Statistical Computing), with statistical significance set as a 2-sided P < .05. We performed a retrospective cohort study using the UNOS Scientific Registry of Transplant Recipients (SRTR) database, which includes all organ donors in the US since October 1, 1987. The reporting of results adheres to the guidelines set out in the Strengthening the Reporting of Observational Studies in Epidemiology ( STROBE ) reporting guideline. This study was approved by the Institutional Review Board at Duke University, which deemed the study not human participants research, and thus the requirement for obtaining informed consent was waived. All controlled DCD donors were considered for inclusion. Donors before January 1, 2019, were excluded, as the first DCD heart transplant in the US occurred in 2019. Data through September 30, 2024, were collected. We excluded non-DCD donors, donors under the age of 18 years, and donors with unclear lung disposition. The primary objective of this analysis was to evaluate the association of DCD heart procurement with lung use for transplantation. The cohort was stratified into donors from whom a heart was recovered (cardiac donor), regardless of whether the heart was transplanted, and donors from whom a heart was not recovered (noncardiac donor). Rates of DCD lung use were compared between cardiac DCD and noncardiac DCD groups. The second objective of this study was to evaluate the association of the DCD heart procurement technique (TA-NRP or direct procurement) with lung use. As procurement technique is not discretely captured in the dataset, the interval from time of death to time of thoracic aorta cold flush was used to define procurement technique. Within the cardiac DCD cohort, donors with a time interval of greater than 15 minutes were defined as TA-NRP; those with elapsed time less than 15 minutes were defined as direct procurement, as previously described. Donors with missing aorta clamp time or time of death were excluded from further analysis. Unadjusted rates of DCD lung use were compared between TA-NRP and direct procurement groups. The SRTR releases a risk-adjusted estimate of expected donor yield twice annually that assesses the likelihood of successful transplantation for a specific donor’s organs. Using logistic regression modeling, risk-adjusted estimates of organ-specific donor yield are developed using the most current data collected by the Organ Procurement and Transplantation Network (OPTN). These models have been validated and used in the literature to calculate expected yield of organs transplanted but, to our knowledge, have not been applied in lung transplantation. , , Once a model is chosen for a specific organ, the organ-specific expected yield is calculated algebraically using the model coefficients publicly shared by the SRTR and the OPTN UNOS database. In this way, lung use from each donor is adjusted for donor characteristics, allowing for comparison between observed and expected rates of lung use. For this analysis, we used the most recent model released in January 2024. The observed yield was divided by the expected yield. A ratio of 1 indicated equal observed yield and expected yield; ratios greater or less than 1 suggested that observed yield exceeded or fell short of the expected yield, respectively. We sought to determine whether the observed organ yield in each group was statistically different from the expected yield and from the observed yield in the comparator groups. To do so, we computed the O:E ratio as observed yield divided by expected yield. Bootstrapping techniques were employed to develop 95% CIs and 2-sided P values around the O:E ratios under a null hypothesis positing that observed yield was exactly equal to expected yield (ie, O:E ratio of 1.0). We used 1000 bootstrapped samples for all cohorts. The TA-NRP use among OPOs was also characterized to better understand variations in practice. Trends in TA-NRP use and O:E ratios of DCD lung use were examined over time using the Cochran-Armitage test. Nonparametric Kruskal-Wallis and χ 2 tests were used to compare donor and recipient characteristics as well as perioperative outcomes for lung transplant recipients. Fisher exact tests were used if sample sizes did not meet the assumptions for χ 2 testing. The Kaplan-Meier log-rank survival analysis was used to evaluate recipient mortality. As already described, bootstrapping was used to calculate 95% CIs and P values for reported O:E ratios. Statistical analyses were performed using RStudio, version 4.2.3 (R Project for Statistical Computing), with statistical significance set as a 2-sided P < .05. Study Population and Perioperative Outcomes During the study period, 24 431 controlled DCD donors (8553 [35.0%] female and 15 878 [65.0%] male; median [IQR] age, 49.0 [37.0-58.0] years) met inclusion criteria . Of these donors, 22 607 were classified as noncardiac DCD donors (8232 [36.4%] female and 14 375 [63.6%] male; median [IQR] age, 51.0 [39.0-58.0] years) and 1824 as cardiac DCD donors (321 [17.6%] and 1503 [82.4%] male; median [IQR] age, 32.0 [26.0-38.0] years), with noncardiac DCD donors being significantly older ( P < .001), less likely to be male ( P < .001), and more likely to be smokers (6873 [30.4%] vs 227 [12.4%]; P < .001) and having lower median (IQR) arterial partial pressure of oxygen to fraction of inspired oxygen (Pa o 2 :Fi o 2 ) ratios (235 [147-343] vs 277 [182-387]; P < .001) compared with cardiac DCD donors. Recipients of noncardiac DCD lungs were more likely than cardiac DCD recipients to remain intubated 72 hours after the operation (411 [43.3%] vs 108 [36.9%]; P = .050), but both cohorts experienced similar hospital lengths of stay (24.0 [15.0-45.0] days vs 24.0 [15.5-40.0] days; P = .58) and rates of postoperative airway dehiscence (17 [1.8%] vs 9 [3.1%]; P = .18), respectively . The cardiac DCD recipients had better 30-day survival compared with noncardiac DCD (923 of 950 [97.2%] vs 291 of 293 [99.3%]; log-rank P = .04) but similar overall survival (28 of 293 [71.9%] vs 8 of 950 [44.4%]; log-rank P = .24). Of 1824 hearts recovered from cardiac DCD donors, 1698 (93.1%) were transplanted and 126 (6.9%) were not. Of 126 nontransplanted hearts, 9 were discarded locally, 14 shared and discarded, 16 submitted for research, 10 used for heart valves, and 77 discarded. The cardiac DCD cohort was stratified by procurement technique, with 712 donors undergoing direct procurement and 325 undergoing TA-NRP . In total, 787 (43.0%) cardiac DCD donors could not be defined as direct procurement or TA-NRP due to missing time of death or thoracic aorta cooling time. Direct procurement donors (median [IQR] age, 31.0 [26.0-38.0] years) were significantly younger than TA-NRP donors (median [IQR], 33.0 [26.0-40.0] years; P = .01); other donor demographics were similar between direct procurement and TA-NRP. However, the TA-NRP cohort had shorter median (IQR) lung ischemic times (6.07 [4.38-9.56] hours vs 8.12 [6.16-12.00] hours; P < .001) and distances between the donor and recipient hospital (222 [9-626] nautical miles vs 331 [159-521] nautical miles; P = .050). Direct procurement and TA-NRP recipients had similar rates of reintubation and acute rejection episodes, although direct procurement recipients had longer hospital stays. However, TA-NRP recipients’ 90-day mortality rate (0 of 62 vs 9 of 128 patients [7.0%]; P = .03) and overall survival (4 of 62 patients [6.5%] vs 21 of 128 [16.4%]; P = .04) were significantly better than for direct procurement recipients. Observed and Expected Organ Yields Among 22 607 noncardiac DCD donors, observed donor lung yield was 988 (4.4%), meaning 4.4% of lungs from noncardiac DCD donors were transplanted. By contrast, cardiac DCD donors had a significantly higher observed yield of 304 donors (16.7%) ( P < .001). When stratifying the cardiac DCD cohort by procurement technique, there was no significant difference in observed lung yield between cardiac DCD donors who underwent direct procurement vs TA-NRP for heart recovery (133 [18.7%] vs 62 [19.1%]; P = .88). Comparing O:E ratios for lung use revealed that all cohorts exceeded expected lung yields. The noncardiac DCD cohort had an O:E ratio of 1.29 (95% CI, 1.21-1.35; P < .001), while the cardiac DCD cohort O:E ratio was 1.79 (95% CI, 1.62-1.96; P < .001). Among cardiac DCD donors, TA-NRP donors had the highest O:E ratio (2.00 [95% CI, 1.60-2.43]), and the direct procurement cohort O:E ratio was 1.77 (95% CI, 1.52-1.99). All of these O:E ratios were statistically significant. However, while the cardiac DCD donor O:E ratio was higher than that for noncardiac DCD donors, the O:E ratios did not differ significantly between the TA-NRP and direct procurement groups ( P = .83). Subanalyses The O:E ratios for donor lung yields were further stratified by year to evaluate temporal trends in lung use. Using all available controlled DCD data (January 1, 1994, through September 30, 2024), the overall DCD (noncardiac DCD plus cardiac DCD) cohort O:E ratios showed significant change over time (Cochran-Armitage test of trend, P < .001) ( A). However, within our study period (January 1, 2019, through September 30, 2024), the DCD O:E ratios lacked a significant temporal trend (Cochran-Armitage test of trend, P = .92), despite significantly increased numbers of DCD donors per year ( P < .001 for trend). Similarly, among the noncardiac DCD, DCD, direct procurement, and TA-NRP (cohorts, O:E ratios remained stable over time despite significant increases in donor counts ( P < .001 for trend for all cohorts) ( B). Temporal examination of TA-NRP use revealed a steady increase in the number of TA-NRP procurements since 2019 ( P < .001 for trend) ( C). We also characterized variations in TA-NRP use among OPOs and UNOS regions. Most OPOs performed fewer than 10 TA-NRP procurements since 2019, while only 5 OPOs performed 20 or more, indicating significant variability in TA-NRP use . During the study period, 24 431 controlled DCD donors (8553 [35.0%] female and 15 878 [65.0%] male; median [IQR] age, 49.0 [37.0-58.0] years) met inclusion criteria . Of these donors, 22 607 were classified as noncardiac DCD donors (8232 [36.4%] female and 14 375 [63.6%] male; median [IQR] age, 51.0 [39.0-58.0] years) and 1824 as cardiac DCD donors (321 [17.6%] and 1503 [82.4%] male; median [IQR] age, 32.0 [26.0-38.0] years), with noncardiac DCD donors being significantly older ( P < .001), less likely to be male ( P < .001), and more likely to be smokers (6873 [30.4%] vs 227 [12.4%]; P < .001) and having lower median (IQR) arterial partial pressure of oxygen to fraction of inspired oxygen (Pa o 2 :Fi o 2 ) ratios (235 [147-343] vs 277 [182-387]; P < .001) compared with cardiac DCD donors. Recipients of noncardiac DCD lungs were more likely than cardiac DCD recipients to remain intubated 72 hours after the operation (411 [43.3%] vs 108 [36.9%]; P = .050), but both cohorts experienced similar hospital lengths of stay (24.0 [15.0-45.0] days vs 24.0 [15.5-40.0] days; P = .58) and rates of postoperative airway dehiscence (17 [1.8%] vs 9 [3.1%]; P = .18), respectively . The cardiac DCD recipients had better 30-day survival compared with noncardiac DCD (923 of 950 [97.2%] vs 291 of 293 [99.3%]; log-rank P = .04) but similar overall survival (28 of 293 [71.9%] vs 8 of 950 [44.4%]; log-rank P = .24). Of 1824 hearts recovered from cardiac DCD donors, 1698 (93.1%) were transplanted and 126 (6.9%) were not. Of 126 nontransplanted hearts, 9 were discarded locally, 14 shared and discarded, 16 submitted for research, 10 used for heart valves, and 77 discarded. The cardiac DCD cohort was stratified by procurement technique, with 712 donors undergoing direct procurement and 325 undergoing TA-NRP . In total, 787 (43.0%) cardiac DCD donors could not be defined as direct procurement or TA-NRP due to missing time of death or thoracic aorta cooling time. Direct procurement donors (median [IQR] age, 31.0 [26.0-38.0] years) were significantly younger than TA-NRP donors (median [IQR], 33.0 [26.0-40.0] years; P = .01); other donor demographics were similar between direct procurement and TA-NRP. However, the TA-NRP cohort had shorter median (IQR) lung ischemic times (6.07 [4.38-9.56] hours vs 8.12 [6.16-12.00] hours; P < .001) and distances between the donor and recipient hospital (222 [9-626] nautical miles vs 331 [159-521] nautical miles; P = .050). Direct procurement and TA-NRP recipients had similar rates of reintubation and acute rejection episodes, although direct procurement recipients had longer hospital stays. However, TA-NRP recipients’ 90-day mortality rate (0 of 62 vs 9 of 128 patients [7.0%]; P = .03) and overall survival (4 of 62 patients [6.5%] vs 21 of 128 [16.4%]; P = .04) were significantly better than for direct procurement recipients. Among 22 607 noncardiac DCD donors, observed donor lung yield was 988 (4.4%), meaning 4.4% of lungs from noncardiac DCD donors were transplanted. By contrast, cardiac DCD donors had a significantly higher observed yield of 304 donors (16.7%) ( P < .001). When stratifying the cardiac DCD cohort by procurement technique, there was no significant difference in observed lung yield between cardiac DCD donors who underwent direct procurement vs TA-NRP for heart recovery (133 [18.7%] vs 62 [19.1%]; P = .88). Comparing O:E ratios for lung use revealed that all cohorts exceeded expected lung yields. The noncardiac DCD cohort had an O:E ratio of 1.29 (95% CI, 1.21-1.35; P < .001), while the cardiac DCD cohort O:E ratio was 1.79 (95% CI, 1.62-1.96; P < .001). Among cardiac DCD donors, TA-NRP donors had the highest O:E ratio (2.00 [95% CI, 1.60-2.43]), and the direct procurement cohort O:E ratio was 1.77 (95% CI, 1.52-1.99). All of these O:E ratios were statistically significant. However, while the cardiac DCD donor O:E ratio was higher than that for noncardiac DCD donors, the O:E ratios did not differ significantly between the TA-NRP and direct procurement groups ( P = .83). The O:E ratios for donor lung yields were further stratified by year to evaluate temporal trends in lung use. Using all available controlled DCD data (January 1, 1994, through September 30, 2024), the overall DCD (noncardiac DCD plus cardiac DCD) cohort O:E ratios showed significant change over time (Cochran-Armitage test of trend, P < .001) ( A). However, within our study period (January 1, 2019, through September 30, 2024), the DCD O:E ratios lacked a significant temporal trend (Cochran-Armitage test of trend, P = .92), despite significantly increased numbers of DCD donors per year ( P < .001 for trend). Similarly, among the noncardiac DCD, DCD, direct procurement, and TA-NRP (cohorts, O:E ratios remained stable over time despite significant increases in donor counts ( P < .001 for trend for all cohorts) ( B). Temporal examination of TA-NRP use revealed a steady increase in the number of TA-NRP procurements since 2019 ( P < .001 for trend) ( C). We also characterized variations in TA-NRP use among OPOs and UNOS regions. Most OPOs performed fewer than 10 TA-NRP procurements since 2019, while only 5 OPOs performed 20 or more, indicating significant variability in TA-NRP use . In this cohort study involving analyses of the UNOS registry, we used an SRTR risk-adjusted model to assess the association of simultaneous heart procurement with DCD lung use. We found that simultaneous heart procurement was associated with increased use of lungs recovered from DCD donors. Additionally, when comparing 2 distinct techniques for cardiac DCD, we found that TA-NRP and direct procurement were associated with similar rates of lung use with DCD. To our knowledge, this is the largest study evaluating the association of procurement techniques with lung use from DCD donors and the only study to develop risk-adjusted O:E ratios for this purpose. Furthermore, we observed improved survival among recipients after donor TA-NRP, contrasting with previous smaller studies showing similar survival between TA-NRP and direct procurement. Our results suggest that using TA-NRP to recover DCD hearts is not negatively associated with lung use and may, in fact, be associated with improved posttransplant survival. TA-NRP has gained traction in recent years as a method to reanimate hearts received through DCD and assess their viability for transplantation, demonstrating improved DCD liver and kidney transplantation outcomes. , Increased DCD liver use has also been demonstrated in recipients from donors with TA-NRP compared with donors undergoing direct procurement. Nevertheless, concerns persist that it may damage otherwise healthy lungs and negatively impact DCD lung use. Rapid access to the thoracic cavity and prolonged regional mechanical circulatory support can precipitate extensive blood loss and necessitate transfusions, while left ventricular dilation or increased left atrial pressures with VA-ECMO can cause pulmonary edema—both well-documented potential sources of injury. Despite these concerns, our analysis found that both direct procurement and TA-NRP cohorts had better-than-expected DCD lung use (O:E 1.77 vs 2.00; P < .001 for both). Notably, the 2 cohorts did not differ significantly, indicating both methods are reasonable options for heart recovery from a lung use perspective. Previous reports have cited raw percentages of lung use rates but have been limited by small cohorts and have not adjusted for donor characteristics that directly correlate with use. , , Simply comparing actual lung use between cohorts is not entirely meaningful, as it is unlikely that all donors were ideal lung donation candidates. The probability of successful organ yield varies across both donors and different organs within the same donor. Our analysis provides a detailed overview of lung yield in DCD donors by leveraging SRTR data to adjust for donor factors, allowing for comparison of observed DCD lung yields against reasonable standards, potentially alleviating previous concerns about whether TA-NRP interferes with lung recovery. Additionally, we demonstrated a substantial yearly increase in use of the TA-NRP technique ( P < .001 for trend) but stable O:E ratios over time (Cochran-Armitage test for trend, P = .24), further supporting the argument that TA-NRP does not negatively impact donor lung yield as experience has accumulated. Despite a TA-NRP O:E ratio that was greater than expected and comparable to the direct procurement group, anecdotal experiences from transplant teams still reveal concern about the potential of TA-NRP to damage lung allografts. These concerns may be partially reflected in variable OPO experience with TA-NRP. Our analysis found heterogeneity in OPO TA-NRP volumes: only 5 OPOs performed 20 or more TA-NRP procurements since 2019, while over half of OPOs performed under 10 in the same period. Although O:E ratios were similar between TA-NRP and direct procurement cohorts, variation may exist in TA-NRP O:E ratios at the OPO level. As low-volume OPOs increase TA-NRP volume and gain experience with the technique, DCD lung use may improve, potentially resulting in significantly higher O:E ratios nationally. Based on our analysis, anecdotal cases of poor lung quality following TA-NRP procurement should not deter transplant teams and OPOs from the technique. Instead, standardizing implementation of this technique should be prioritized, especially given the variations that currently exist. , , In fact, the argument to use TA-NRP extends beyond lung use. A study by Bakhtiyar et al demonstrated that TA-NRP is significantly cheaper and generates greater organ yield per donor compared with direct procurement followed by ex situ machine perfusion. Additionally, while perioperative outcomes were not the primary focus of our study, we also observed improved 90-day and overall survival in the TA-NRP cohort vs direct procurement, contrasting with prior work that has demonstrated comparable outcomes between the groups. , , The discrepancy likely stems from the larger sample size and more comprehensive analysis in our study. Further in-depth assessment of allograft and patient survival should be prioritized as case volume and experience accumulate. As prior analyses have shown a positive correlation between transplant center experience and posttransplant outcomes with ex vivo lung perfusion, future research should evaluate whether a similar association exists for TA-NRP. Limitations There are several limitations in our study. First, the UNOS SRTR registry does not explicitly track TA-NRP use. , , Consequently, we inferred use of TA-NRP from time elapsed between death and thoracic organ cooling. While we believe that a minimum threshold of 15 minutes between death and thoracic aorta cold flush most accurately captures TA-NRP donors, and this has been previously validated, there remains heterogeneity and a lack of consensus regarding the minimum threshold. Additionally, 43% of our 1824 cardiac donors were missing either time of death or thoracic aortic cooling time, necessitating exclusion and potentially introducing bias. The lack of a discrete field for these important data, compounded by missingness in other fields, highlights the importance of the OPTN modernization effort, especially with regard to data sciences. As TA-NRP, machine perfusion, and other technological advances become increasingly embedded in the national landscape of solid organ transplantation, the OPTN should make explicit provisions to collect important data for these donors to facilitate investigation of the impact of these changes on organ use and outcomes. These changes will potentially provide additional data to address the ethical concerns associated with NRP use as well. Furthermore, differences in lung use and outcomes between TA-NRP and abdominal-NRP remain unclear; improved data collection could help answer this question. Last, our use of the SRTR risk-adjustment model for donor yield presents a potential limitation. These models are trained on recent data and updated biannually, meaning that the 2024 model we used was based on data from the second half of 2023. Thus, our risk-adjusted analysis is based on current practices and trends, in essence using today’s expectations to evaluate yesterday’s donors, potentially underestimating the O:E ratios we developed. However, given that our observed yields exceeded expected yields, we believe this only strengthens our conclusion that TA-NRP does not impede lung use and may, in fact, be associated with improved recovery. There are several limitations in our study. First, the UNOS SRTR registry does not explicitly track TA-NRP use. , , Consequently, we inferred use of TA-NRP from time elapsed between death and thoracic organ cooling. While we believe that a minimum threshold of 15 minutes between death and thoracic aorta cold flush most accurately captures TA-NRP donors, and this has been previously validated, there remains heterogeneity and a lack of consensus regarding the minimum threshold. Additionally, 43% of our 1824 cardiac donors were missing either time of death or thoracic aortic cooling time, necessitating exclusion and potentially introducing bias. The lack of a discrete field for these important data, compounded by missingness in other fields, highlights the importance of the OPTN modernization effort, especially with regard to data sciences. As TA-NRP, machine perfusion, and other technological advances become increasingly embedded in the national landscape of solid organ transplantation, the OPTN should make explicit provisions to collect important data for these donors to facilitate investigation of the impact of these changes on organ use and outcomes. These changes will potentially provide additional data to address the ethical concerns associated with NRP use as well. Furthermore, differences in lung use and outcomes between TA-NRP and abdominal-NRP remain unclear; improved data collection could help answer this question. Last, our use of the SRTR risk-adjustment model for donor yield presents a potential limitation. These models are trained on recent data and updated biannually, meaning that the 2024 model we used was based on data from the second half of 2023. Thus, our risk-adjusted analysis is based on current practices and trends, in essence using today’s expectations to evaluate yesterday’s donors, potentially underestimating the O:E ratios we developed. However, given that our observed yields exceeded expected yields, we believe this only strengthens our conclusion that TA-NRP does not impede lung use and may, in fact, be associated with improved recovery. In this national cohort study of DCD donors, we used risk-adjusted models from the SRTR to evaluate the association of simultaneous heart procurement, including direct procurement and TA-NRP techniques, with use of DCD donor lungs. We found that simultaneous heart procurement was not adversely associated with DCD lung use, nor was use of TA-NRP for cardiac DCD associated with decreased use of lung allografts. Importantly, we identified significant heterogeneity in TA-NRP use and experience among OPOs, which may lead to varying rates of lung use when using TA-NRP. Our findings suggest that OPOs and transplant teams should not be deterred from pursuing cardiac DCD transplantation, regardless of procurement technique employed. Additionally, OPOs, transplant teams, and the OPTN should strive to standardize procurement techniques, especially for multiorgan DCD, to optimize all DCD organ yields. As UNOS data reporting improves and NRP experience accumulates in the US, future studies should evaluate the impact of ex vivo lung perfusion and abdominal NRP vs TA-NRP on lung use, as well as potential relationships between center volume of TA-NRP cases and clinical posttransplant outcomes. |
Educational video for self-care with arteriovenous fistula in renal patients: randomized clinical trial | 8d8d02a8-15d4-4054-ba4f-f784e1404866 | 11182605 | Patient Education as Topic[mh] | Chronic Kidney Disease (CKD) is one of the main causes of death from non-communicable diseases. Its prevalence has increased over time, so that it affects populations in different regions of the world unequally, probably as a result of differences in the demographic characteristics of the population, their comorbidities, and access to health resources . In Brazil, around 139,691 individuals were undergoing kidney dialysis as a treatment for the disease in 2019, which represented an average increase of 6,881 patients (5.43%) compared to the previous year . Among the types of kidney dialysis, hemodialysis (HD) is the most widely used modality in Brazil (93.2%), requiring vascular access . The Arteriovenous Fistula (AVF) is considered the most appropriate access when compared to grafts and catheters, as it has lower mortality rates and is related to a lower risk of infection . Despite being the most appropriate venous access for HD, the use of AVFs can be related to some complications, such as the presence of aneurysms, bleeding, ischemic neuropathies, lymphedema, and venous hypertensive events . In an attempt to reduce these complications, it is important for patients to develop self-care actions aimed at AVF . Even so, the implementation of these actions has been lower than expected in many studies . In order to help the patient, the main guidelines available recommend carrying out educational actions aimed at them about vascular access . In this context, educational technologies, as a potential strategy for professional practice, in addition to offering support, are able to provide the necessary support and guidance for care, as well as tending to increase the patient’s knowledge, with a view to promoting health and quality of life . Among educational technologies, video stands out as a strategy related to better health outcomes among diverse patient groups . Furthermore, the use of video can benefit professionals, patients, and the management of hemodialysis clinics in terms of the health education process. This is because it allows the content to be replayed for the patient without the need for a health professional, which can optimize the use of these human resources in the face of the demand for existing activities . However, in order to verify the effect of isolated educational interventions or the combination of different interventions on the development of correct AVF self-care behaviors, it is important that they be evaluated through randomized clinical trials . In this sense, some studies have shown the positive effect of educational interventions among renal patients with an approach to AVF and their effect on self-care practices with AVF . However, it is important to highlight the effect of these technologies not only on the practice of self-care but also on knowledge and attitude, since these concepts are interconnected. Knowledge refers to the apprehension of information with the potential to help maintain the functionality of the AVF; attitude represents the willingness to perform self-care, and is influenced by the patient’s convictions and feelings about access; and practice refers to actions developed by the patient, often based on their knowledge of the subject . In view of the above, this study sought to answer the research question: Is the educational video “Care of the arteriovenous fistula” effective in improving the knowledge, attitude, and practice of self-care in patients undergoing hemodialysis due to an arteriovenous fistula? To this end, the aim of this study was to evaluate the effect of an educational video on the knowledge, attitude, and practice of self-care with arteriovenous fistula of patients undergoing hemodialysis treatment.
Study type This is a randomized, controlled, two-arm, single-blind clinical trial. The study followed the recommendations of the Consolidated Standards of Reporting Trials (CONSORT) for non-pharmacological interventions and was registered in the Brazilian Clinical Trials Registry (ReBEC) database with the identification number U1111-1241-6730, which was approved in April 2020. Study variables The variables collected were classified into two groups: the Dependent Variable group (equivalent to the outcome of the intervention), which comprises the scores of knowledge, attitude and practice of self-care with AVF among patients undergoing hemodialysis treatment, assessed by the results of the KAP (Knowledge, Attitude, and Practice) survey on the self-care of patients with AVF; and the Independent Variables group (explanatory and descriptive), which comprises A) Sociodemographic variables: age (in years); gender (female/male); marital status (with partner/without partner); schooling (in years of study); occupation (self-employed; employed/unemployed/retired/student); monthly per capita income (in reais ); health care plan (public health network/private health network); and, B) Clinical variables: complications with the current AVF (yes/no); presence of previous AVF (yes/no); length of hemodialysis treatment (in months); length of hemodialysis treatment with AVF (in months); length of hemodialysis treatment with current AVF (in months). Intervention The intervention in this study was mediated by the use of the educational video “Care of the arteriovenous fistula”. This audiovisual resource was produced based on the precepts of Dorothea Orem’s General Nursing Theory and evaluated by experts represented by nurses and media professionals . The items evaluated by these experts were related to the concept of the idea, dramatic construction, rhythm, characters, dramatic potential, dialogues, visual style, target audience, and relevance. In addition to these, the media professionals also assessed the video’s functionality, usability, and efficiency. The items considered inappropriate were modified according to the experts’ suggestions . The educational technology is three minutes and seventeen seconds long and covers the self-care actions that patients should carry out before and after the arteriovenous fistula is made, during its use in hemodialysis, and in the prevention and monitoring of complications during access . The contents of the video regarding the pre-surgery of the AVF were the care taken to preserve the venous network of the arm chosen by the doctor. In the postoperative period, care is taken to dress the surgical wound and, during the use of the AVF for hemodialysis, the video illustrates actions that should be avoided on the arm, such as wearing watches, tight clothing, measuring blood pressure, sleeping on it, or carrying excess weight with it. For the prevention and monitoring of access complications, the precautions covered are routine checking of the AVF’s fremitus, washing the limb before HD, hemostasis at the end of therapy, and monitoring and treating complications such as hematomas, infections, steal syndrome, and thrombosis . Study location, period, and population The study was carried out in the hemodialysis services of a northeastern capital. To allocate them to the control and intervention groups, four clinics were selected by drawing lots among all 13 establishments offering hemodialysis treatment. In this way, all the clinics had the same probability of being selected for the study by drawing lots without the intervention of the researchers. The study population was made up of patients undergoing hemodialysis at these services between July and November 2021. Selection criteria, sample, and sampling The sample calculation equation for two experimental means was used to determine the sample size . A 95% confidence coefficient and 80% test power were defined. As for the means and standard deviation of knowledge, attitude, and practice of self-care with AVF, they were obtained from the application of a pilot study conducted in May 2021 with hemodialysis patients in the selected centers. To define the pilot study sample, the same sample calculation equation was used for two experimental means . However, we considered the average self-care behavior with AVF of 71% and a standard deviation of 13.6, as shown in the study by Sousa and collaborators (2017) . It is important to note that self-care behavior is equivalent to the practice assessed in this study and there are no studies assessing the dimensions of knowledge and attitude with representative samples. For the average number of self-care behaviors in the intervention group, the expectation of a 10% increase in the frequency of self-care was considered and the standard deviation was not altered by the Control Group (CG). To the sample value obtained, 77% was added, equivalent to possible losses during the continuation of the study. ET losses were estimated on the basis of a study and took into account the rate of death, kidney transplantation, and change of dialysis modality or leaving the original clinic. Thus, the total sample defined was 52 patients allocated to the CG and 52 to the Intervention Group (IG), while the sample for the pilot test was 6 patients for each group (equivalent to 10% of the calculation obtained). It should be noted that 10 patients were recruited for each of the groups in order to take account of losses during data collection and 7 completed the entire follow-up. Based on the results of the pilot study, a new sample calculation was made. As well as providing the necessary measurements for the sample calculation, the pilot study aimed to verify the need for adjustments to the data collection procedure. Patients undergoing hemodialysis at the clinics initially drawn for the control and intervention groups took part in this stage, with the aim of promoting conditions identical to the actual clinical trial in the pilot test. The data collected for the pilot study was kept for the experimental study, as the necessary changes in the conduct of the research did not affect the quality of the collection content. The means and standard deviation found in the pilot study are shown in . The sample calculation resulted in 47 patients for each of the groups, when considering knowledge, 10 patients for assessing attitude, and 12 for verifying practice. The inclusion criteria for the sample were: patients over the age of 18 using an AVF for hemodialysis treatment for at least six months, in order to allow patients to understand and carry out self-care actions with the AVF. Patients with some level of mental or cognitive disorder assessed by the Mini-Mental State Examination (MMSE) and patients with a knowledge score measured in the pre-test higher than 76, which represents 80% of the maximum score for this dimension, were not included. Patients with a diagnosis of total hypoacusis, as described in their medical records or signaled by the attending physician, were also excluded from the sample because they were unable to understand the educational video. Data collection instruments and procedures The first encounter with the patient took place during hemodialysis treatment, when cognitive status was assessed using the MMSE and then a socioeconomic and clinical questionnaire was administered to patients who met the inclusion and exclusion criteria. The evaluation of knowledge, attitude, and practice of self-care with AVF was carried out in the pre-test through telephone conversations and the application of a previously validated scale . The Scale of Knowledge, Attitude, and Practice of Self-Care with AVF (ECAPA-FAV, from its acronym in Portuguese) has 31 items with scores ranging from 1 to 5, so the higher the score, the more adequate the patient’s knowledge, attitude, and practice. The knowledge items seek to identify how much the patient knows about self-care with the AVF, with questions such as “What do you know about signs that the fistula is infected?”. The attitude scale aims to identify the importance the patient attaches to this self-care through questions such as “How important do you think it is to know how to care for the fistula?”. Finally, questions such as “How often do you ET the habit of checking whether the arm of the fistula is red, hot, or has a secretion?” are contained in the practice scale with the intention of identifying the self-care activities carried out by patients. In its validation process, the scale was applied to renal patients dialyzing by AVF and subjected to exploratory factor analysis, which enabled its final structure to be obtained with explained variance and McDonald’s Omega values of 40.4%/0.896, 60.7%/0.843, and 36.9%/0.702 for knowledge, attitude, and practice, respectively . Patients were randomized using conglomerates so that allocation to the control and intervention groups was defined by naturally occurring groupings. In the case of this study, the clusters were dialysis clinics. This type of randomization was chosen to avoid patients participating in the IG discussing the intervention with the CG members. Thus, the distance between the hemodialysis clinics allocated to the two groups was an obstacle to communication between the patients. Patients were allocated to the control or intervention groups in two stages. Firstly, the hemodialysis clinics that would be part of each group were drawn by lot, so that the first two clinics drawn made up the intervention group and the next two were allocated to the control group. Once the clinics participating in the study had been determined, a list of patients dialyzing by AVF was requested to enable the selection of the participants for the study. The patients selected for the final sample were then drawn at random from each of the clinics drawn in the first stage. The draw was carried out by the main researcher, under the supervision of two members of the research group, in order to guarantee the suitability of the process. After the pre-test, the intervention was carried out in IG. The video was played twice on the hemodialysis day following the pre-test, in order to optimize the assimilation of the information, since the patient may not be aware of all the instructions given on first contact. This took place during hemodialysis treatment, individually, using an electronic tablet with a 9.7-inch screen and a basic headset. Both the control group and the Intervention group were subjected to the usual health education actions carried out in the hemodialysis units themselves, represented by information delivered verbally to the patients by the health professionals during the preoperative consultation for the AVF and during the hemodialysis session. The content of the information provided was similar to that contained in the educational video used as an intervention in this study, although there was no specific script to guide the delivery of this information. The post-test was administered 7 and 14 days after the educational intervention in the case of patients allocated to the intervention group and 7 and 14 days after the pretest for patients in the control group. To this end, the same team responsible for the pretest contacted them by telephone and reapplied the ECAPA-FAV to the study participants. The face-to-face application of the scale in the pre-test lasted an average of 25 minutes, so the items on the scale were just read out to the participants, with no explanation of their content. If necessary, the person responsible for data collection was authorized to repeat the item as it was presented in the instrument. For the post-tests, the ECAPA-FAV was applied by telephone, in which the researcher identified herself, mentioned the pre-test stage, and proceeded to read out the instructions for completing each item, repeating them if requested. For each item read, she waited for the participant’s response and moved on to the next item on the scale, ending the call by thanking them for their participation, making the average phone call last twenty minutes. Patients who had been transplanted or who had died during the data collection period and patients who did not answer the phone to take the post-test after five attempts during the follow-up day and the three days afterward were considered to have dropped out or been lost. It should be noted that the researcher responsible for playing the video was not blinded, since she applied the intervention and managed the data collection team. However, in order not to compromise the research results, the team responsible for administering the pre and post-test and the statistician responsible for analyzing the data were blinded. Data processing and analysis The data was analyzed using the Statistical Package for the Social Sciences for Windows (SPSS) software, version 20.0. Initially, the relative and absolute frequencies of the qualitative variables and the statistical measures of minimum, maximum, mean, and standard deviation of the quantitative variables were calculated. The homogeneity between the representatives of the control and intervention groups at baseline and the assessment of differences between the knowledge, attitude, and practice of self-care between the groups were checked using the chi-square test for the qualitative variables and the Student’s t-test and Mann-Whitney test. The variables were quantitative. To check for normality, the Kolmogorov-Smirnov test was applied to the entire sample of patients (members of the control and intervention groups). The Friedman test was used to compare the scores for knowledge, attitude, and practice of self-care with AVF at baseline (D0), on the seventh (D7) and fourteenth (D14) days. In cases where the Friedman test showed statistical significance, a post-hoc analysis was carried out for multiple comparisons. The Shapiro-Wilk test was used to assess the normality of the scores in each of the groups. A 5% significance level was used for all conclusions. Ethical aspects This study was approved by the Research Ethics Committee (CEP) of the Federal University of Pernambuco, under opinion number 3.555.992. Data collection began only after approval from the REC and the signing of the Free and Informed Consent Term (FICT) by the study participants. It should also be noted that, as the study used headphones connected to the tablet to apply the intervention, certain precautions were taken to reduce discomfort caused by noise and the risk of infection caused by sharing the device. An intermediate volume was used for the sound during video playback, which could be changed according to the participant’s wishes. As for the risk of cross-infection, this was reduced by antisepsis with 70% alcohol before and after each use of the headphones.
This is a randomized, controlled, two-arm, single-blind clinical trial. The study followed the recommendations of the Consolidated Standards of Reporting Trials (CONSORT) for non-pharmacological interventions and was registered in the Brazilian Clinical Trials Registry (ReBEC) database with the identification number U1111-1241-6730, which was approved in April 2020.
The variables collected were classified into two groups: the Dependent Variable group (equivalent to the outcome of the intervention), which comprises the scores of knowledge, attitude and practice of self-care with AVF among patients undergoing hemodialysis treatment, assessed by the results of the KAP (Knowledge, Attitude, and Practice) survey on the self-care of patients with AVF; and the Independent Variables group (explanatory and descriptive), which comprises A) Sociodemographic variables: age (in years); gender (female/male); marital status (with partner/without partner); schooling (in years of study); occupation (self-employed; employed/unemployed/retired/student); monthly per capita income (in reais ); health care plan (public health network/private health network); and, B) Clinical variables: complications with the current AVF (yes/no); presence of previous AVF (yes/no); length of hemodialysis treatment (in months); length of hemodialysis treatment with AVF (in months); length of hemodialysis treatment with current AVF (in months).
The intervention in this study was mediated by the use of the educational video “Care of the arteriovenous fistula”. This audiovisual resource was produced based on the precepts of Dorothea Orem’s General Nursing Theory and evaluated by experts represented by nurses and media professionals . The items evaluated by these experts were related to the concept of the idea, dramatic construction, rhythm, characters, dramatic potential, dialogues, visual style, target audience, and relevance. In addition to these, the media professionals also assessed the video’s functionality, usability, and efficiency. The items considered inappropriate were modified according to the experts’ suggestions . The educational technology is three minutes and seventeen seconds long and covers the self-care actions that patients should carry out before and after the arteriovenous fistula is made, during its use in hemodialysis, and in the prevention and monitoring of complications during access . The contents of the video regarding the pre-surgery of the AVF were the care taken to preserve the venous network of the arm chosen by the doctor. In the postoperative period, care is taken to dress the surgical wound and, during the use of the AVF for hemodialysis, the video illustrates actions that should be avoided on the arm, such as wearing watches, tight clothing, measuring blood pressure, sleeping on it, or carrying excess weight with it. For the prevention and monitoring of access complications, the precautions covered are routine checking of the AVF’s fremitus, washing the limb before HD, hemostasis at the end of therapy, and monitoring and treating complications such as hematomas, infections, steal syndrome, and thrombosis .
The study was carried out in the hemodialysis services of a northeastern capital. To allocate them to the control and intervention groups, four clinics were selected by drawing lots among all 13 establishments offering hemodialysis treatment. In this way, all the clinics had the same probability of being selected for the study by drawing lots without the intervention of the researchers. The study population was made up of patients undergoing hemodialysis at these services between July and November 2021.
The sample calculation equation for two experimental means was used to determine the sample size . A 95% confidence coefficient and 80% test power were defined. As for the means and standard deviation of knowledge, attitude, and practice of self-care with AVF, they were obtained from the application of a pilot study conducted in May 2021 with hemodialysis patients in the selected centers. To define the pilot study sample, the same sample calculation equation was used for two experimental means . However, we considered the average self-care behavior with AVF of 71% and a standard deviation of 13.6, as shown in the study by Sousa and collaborators (2017) . It is important to note that self-care behavior is equivalent to the practice assessed in this study and there are no studies assessing the dimensions of knowledge and attitude with representative samples. For the average number of self-care behaviors in the intervention group, the expectation of a 10% increase in the frequency of self-care was considered and the standard deviation was not altered by the Control Group (CG). To the sample value obtained, 77% was added, equivalent to possible losses during the continuation of the study. ET losses were estimated on the basis of a study and took into account the rate of death, kidney transplantation, and change of dialysis modality or leaving the original clinic. Thus, the total sample defined was 52 patients allocated to the CG and 52 to the Intervention Group (IG), while the sample for the pilot test was 6 patients for each group (equivalent to 10% of the calculation obtained). It should be noted that 10 patients were recruited for each of the groups in order to take account of losses during data collection and 7 completed the entire follow-up. Based on the results of the pilot study, a new sample calculation was made. As well as providing the necessary measurements for the sample calculation, the pilot study aimed to verify the need for adjustments to the data collection procedure. Patients undergoing hemodialysis at the clinics initially drawn for the control and intervention groups took part in this stage, with the aim of promoting conditions identical to the actual clinical trial in the pilot test. The data collected for the pilot study was kept for the experimental study, as the necessary changes in the conduct of the research did not affect the quality of the collection content. The means and standard deviation found in the pilot study are shown in . The sample calculation resulted in 47 patients for each of the groups, when considering knowledge, 10 patients for assessing attitude, and 12 for verifying practice. The inclusion criteria for the sample were: patients over the age of 18 using an AVF for hemodialysis treatment for at least six months, in order to allow patients to understand and carry out self-care actions with the AVF. Patients with some level of mental or cognitive disorder assessed by the Mini-Mental State Examination (MMSE) and patients with a knowledge score measured in the pre-test higher than 76, which represents 80% of the maximum score for this dimension, were not included. Patients with a diagnosis of total hypoacusis, as described in their medical records or signaled by the attending physician, were also excluded from the sample because they were unable to understand the educational video.
The first encounter with the patient took place during hemodialysis treatment, when cognitive status was assessed using the MMSE and then a socioeconomic and clinical questionnaire was administered to patients who met the inclusion and exclusion criteria. The evaluation of knowledge, attitude, and practice of self-care with AVF was carried out in the pre-test through telephone conversations and the application of a previously validated scale . The Scale of Knowledge, Attitude, and Practice of Self-Care with AVF (ECAPA-FAV, from its acronym in Portuguese) has 31 items with scores ranging from 1 to 5, so the higher the score, the more adequate the patient’s knowledge, attitude, and practice. The knowledge items seek to identify how much the patient knows about self-care with the AVF, with questions such as “What do you know about signs that the fistula is infected?”. The attitude scale aims to identify the importance the patient attaches to this self-care through questions such as “How important do you think it is to know how to care for the fistula?”. Finally, questions such as “How often do you ET the habit of checking whether the arm of the fistula is red, hot, or has a secretion?” are contained in the practice scale with the intention of identifying the self-care activities carried out by patients. In its validation process, the scale was applied to renal patients dialyzing by AVF and subjected to exploratory factor analysis, which enabled its final structure to be obtained with explained variance and McDonald’s Omega values of 40.4%/0.896, 60.7%/0.843, and 36.9%/0.702 for knowledge, attitude, and practice, respectively . Patients were randomized using conglomerates so that allocation to the control and intervention groups was defined by naturally occurring groupings. In the case of this study, the clusters were dialysis clinics. This type of randomization was chosen to avoid patients participating in the IG discussing the intervention with the CG members. Thus, the distance between the hemodialysis clinics allocated to the two groups was an obstacle to communication between the patients. Patients were allocated to the control or intervention groups in two stages. Firstly, the hemodialysis clinics that would be part of each group were drawn by lot, so that the first two clinics drawn made up the intervention group and the next two were allocated to the control group. Once the clinics participating in the study had been determined, a list of patients dialyzing by AVF was requested to enable the selection of the participants for the study. The patients selected for the final sample were then drawn at random from each of the clinics drawn in the first stage. The draw was carried out by the main researcher, under the supervision of two members of the research group, in order to guarantee the suitability of the process. After the pre-test, the intervention was carried out in IG. The video was played twice on the hemodialysis day following the pre-test, in order to optimize the assimilation of the information, since the patient may not be aware of all the instructions given on first contact. This took place during hemodialysis treatment, individually, using an electronic tablet with a 9.7-inch screen and a basic headset. Both the control group and the Intervention group were subjected to the usual health education actions carried out in the hemodialysis units themselves, represented by information delivered verbally to the patients by the health professionals during the preoperative consultation for the AVF and during the hemodialysis session. The content of the information provided was similar to that contained in the educational video used as an intervention in this study, although there was no specific script to guide the delivery of this information. The post-test was administered 7 and 14 days after the educational intervention in the case of patients allocated to the intervention group and 7 and 14 days after the pretest for patients in the control group. To this end, the same team responsible for the pretest contacted them by telephone and reapplied the ECAPA-FAV to the study participants. The face-to-face application of the scale in the pre-test lasted an average of 25 minutes, so the items on the scale were just read out to the participants, with no explanation of their content. If necessary, the person responsible for data collection was authorized to repeat the item as it was presented in the instrument. For the post-tests, the ECAPA-FAV was applied by telephone, in which the researcher identified herself, mentioned the pre-test stage, and proceeded to read out the instructions for completing each item, repeating them if requested. For each item read, she waited for the participant’s response and moved on to the next item on the scale, ending the call by thanking them for their participation, making the average phone call last twenty minutes. Patients who had been transplanted or who had died during the data collection period and patients who did not answer the phone to take the post-test after five attempts during the follow-up day and the three days afterward were considered to have dropped out or been lost. It should be noted that the researcher responsible for playing the video was not blinded, since she applied the intervention and managed the data collection team. However, in order not to compromise the research results, the team responsible for administering the pre and post-test and the statistician responsible for analyzing the data were blinded.
The data was analyzed using the Statistical Package for the Social Sciences for Windows (SPSS) software, version 20.0. Initially, the relative and absolute frequencies of the qualitative variables and the statistical measures of minimum, maximum, mean, and standard deviation of the quantitative variables were calculated. The homogeneity between the representatives of the control and intervention groups at baseline and the assessment of differences between the knowledge, attitude, and practice of self-care between the groups were checked using the chi-square test for the qualitative variables and the Student’s t-test and Mann-Whitney test. The variables were quantitative. To check for normality, the Kolmogorov-Smirnov test was applied to the entire sample of patients (members of the control and intervention groups). The Friedman test was used to compare the scores for knowledge, attitude, and practice of self-care with AVF at baseline (D0), on the seventh (D7) and fourteenth (D14) days. In cases where the Friedman test showed statistical significance, a post-hoc analysis was carried out for multiple comparisons. The Shapiro-Wilk test was used to assess the normality of the scores in each of the groups. A 5% significance level was used for all conclusions.
This study was approved by the Research Ethics Committee (CEP) of the Federal University of Pernambuco, under opinion number 3.555.992. Data collection began only after approval from the REC and the signing of the Free and Informed Consent Term (FICT) by the study participants. It should also be noted that, as the study used headphones connected to the tablet to apply the intervention, certain precautions were taken to reduce discomfort caused by noise and the risk of infection caused by sharing the device. An intermediate volume was used for the sound during video playback, which could be changed according to the participant’s wishes. As for the risk of cross-infection, this was reduced by antisepsis with 70% alcohol before and after each use of the headphones.
The final sample was 28 patients in the intervention group and 27 patients in the control group, which was lower than planned for the knowledge outcome (47/47) and higher for the attitude (10/10) and practice (12/12) outcomes. Details of the recruitment and follow-up of participants in the control and intervention groups are shown in . The control and intervention groups were homogeneous in terms of gender, marital status, health insurance, occupation, presence of complications, current AVF, age, schooling, per capita income, and length of HD treatment with AVF and current AVF . There was also homogeneity between the knowledge, attitude, and practice scores at baseline (D0). Regarding the difference between the groups in knowledge, attitude, and practice of self-care with AVF on the seventh (D7) and fourteenth day (D14), statistical significance was only found between the practice scores on the seventh day of follow-up, as shown in . When comparing the scores for knowledge, attitude, and practice of self-care with AVF within each group during follow-up, there was statistical significance between the knowledge and practice of the intervention and control groups. Attitude was only significant in the intervention group, as shown in . Post-hoc analysis for multiple comparisons showed significant changes between patients’ knowledge at baseline and day 7 in the control group (p=0.001) and intervention group (0.048) and at baseline and day 14 for both groups (p= 0.001 in the control group and p=0.048 in the intervention group). No differences were observed between the seventh and fourteenth day in the two groups. Regarding the attitude toward self-care among patients in the intervention group, the post-hoc test for multiple comparisons did not confirm statistical significance after applying the video. On the other hand, the practice of self-care with the AVF showed a statistically significant difference between the scores measured at baseline and on the fourteenth day in the control (p<0.001) and intervention (p<0.001) groups. Differences were also found between practice on the seventh and fourteenth day in both groups (p=0.001 in the control group and p<0.001 in the intervention group). There was no significant difference between self-care practice at baseline and the seventh day of follow-up in the two groups analyzed.
Self-care is defined by Dorothea Orem in her theory as the actions taken by individuals for their own benefit in order to maintain life, health, and well-being. When these actions are carried out properly, they can help maintain the structural integrity and functioning of the human body . Facilitating factors or barriers can be found in the development of these actions by patients. ET should be explored by nurses in order to promote motivation and increase patients’ health literacy about the symptoms of chronic kidney disease, favoring independent self-management . It should be noted that health education actions aimed at patients about vascular access are recommended by the main guidelines available . In addition, the adoption of these actions, when they take a patient-centered approach, with an analysis of their knowledge needs, can make them the protagonists in clinical decision-making regarding their state of health . The evaluation of the effect of an educational intervention for self-care with AVF has been shown in two high-impact studies and in only one of them was the intervention of an educational video . The first study evaluated the effect of a structured action based on a multi-method approach with theoretical and practical stages and the use of writing, listening, and visual stimulation. The patients evaluated showed an improvement in self-care behaviors with AVF, both in terms of managing signs and symptoms and preventing complications with the fistula . The use of health education based on an educational video was found to generate a significant increase in scores for self-care behaviors with AVF after two and four weeks. However, a significant increase in self-care behaviors was also found in the control group, which received a face-to-face intervention with a verbal explanation of access care . A similar result was found in this study since the measure of self-care practice showed differences at baseline on the fourteenth day and on the seventh and fourteenth days after in the control and intervention groups. However, no differences were observed between the baseline and the first seven days after playing the video. With regard to knowledge, there was a statistical difference between patients in the intervention group at baseline and on the seventh and fourteenth days of follow-up, but there was no difference between knowledge measured on the seventh and fourteenth days. This result may be related to the fact that educational videos, when used in hospital environments, are more effective in improving short-term health literacy than in changing the patient’s behavior and lifestyle . On the other hand, it is considered that hemodialysis treatment allows a bond with the health team because it is a continuous outpatient therapy, so the educational video can be played repeatedly, especially in the idle moments of the therapy itself. With regard to changes in the practice of self-care, the mechanisms that are effective in modifying health behavior are still little known. This phenomenon needs to be better understood in order to clarify the relationship between health behaviors and the factors that motivate them to take place . To do this, it is necessary to measure such behaviors and analyze their predictors and theoretical explanations of co-occurrences. Psychological, socio-cognitive, environmental, and political variables and mechanisms can influence changes in health behaviors and should be analyzed in correlational and interventionist studies . The influence of these mechanisms may require greater adaptation from the patient, so it may take longer after the intervention has been applied for changes in self-care practices to be observed. This may justify the statistically significant difference between the practice measured at baseline and fourteen days after the educational video was applied. With regard to patient attitude, although no significant differences were confirmed between the scores measured before and after the educational video was used, it is important to note that the patients already had high levels of attitude toward self-care with the AVF even before the intervention was used. The fact that patients see the fistula as essential for the continuity and success of the treatment, as well as for their survival , may have contributed to the high attitude scores found. It is noteworthy that educational initiatives must involve the provision of information about the functioning, preservation and self-care behaviors of the AVF. It is necessary to consider the needs and doubts of patients and their families in order to promote improved communication and encouragement in maintaining self-care activities . Therefore, in addition to an educational component, training needs to include support interventions that promote the development of coping skills to deal with the demands of AVF and its negative impacts on the patient’s life . This understanding can help professionals conduct educational actions that promote critical thinking in patients, favoring the acquisition of appropriate self-care practices based on a positive attitude towards AVF. In addition to the positive results among patients in the intervention group, a significant increase in knowledge was also observed among patients in the control group, measured at baseline and after seven and fourteen days of follow-up. This may have occurred because measuring knowledge by reading the items on the ECAPA-FAV at baseline may have encouraged patients to seek appropriate information on self-care with vascular access. Ensuring that adults with chronic conditions have access to information can increase their confidence in carrying out self-care activities. The patient’s search for this information tends to help increase their confidence in reporting their concerns to the doctor and in understanding when health care should be sought from professionals in the area . In view of the above, this study may have contributed to the acquisition of information on self-care with AVF, both among the patients in the intervention group and those in the control group. It provided clarification for the participants in the first group through the educational video and may have encouraged the participants in the second group to seek knowledge on the subject. Furthermore, despite the positive effect of the isolated application of the video on the knowledge and practice of self-care with AVF among patients undergoing hemodialysis treatment, it is essential that it is also applied as part of more complex educational actions. These actions should consider motivational approaches, promoting active listening to the patient in order to identify and resolve existing doubts about self-care with vascular access, as well as identifying the factors that help and hinder its implementation. It can be seen that the losses during the study’s follow-up resulted in a final sample below that was planned for the knowledge outcome, equivalent to a test power of 56%, which limited the interpretation of the results to knowledge of self-care with AVF. In addition, the use of self-assessment to measure knowledge about self-care with AVF may have had an impact on the results achieved, since people who know little about a subject tend to overestimate their own knowledge, while those who are considered experts tend to underestimate it (cognitive bias). Using the video individually or as part of more complex interventions can help stimulate the patient’s interest in the subject, as well as prompt a broader discussion of the subject. In addition, the educational video can be used continuously during hemodialysis treatment, which is considered idle time for patients. It can be played in the hemodialysis room in order to reach a greater number of people at the same time since many services have TV sets that can play the proposed educational technology. It is suggested that new clinical studies be conducted with the application of other educational interventions, providing a longer follow-up period to assess their effect on knowledge, attitude, and self-care practice over longer periods. It is also pertinent to verify the effect of the technology on the functionality of the vascular access by evaluating the parameters of the adequacy of the AVF, such as satisfactory blood flow, venous pressure, and urea clearance, as well as evidence of complications such as stenosis, thrombosis, and infections.
Significant increases in the knowledge and practice of renal patients were identified during follow-up in the control and intervention groups. With regard to attitude towards self-care, no statistical significance was observed in either group. It is believed that the changes in knowledge and practice among participants in the control group may have been motivated by the fact that reading the items on the measurement scale used in this study encouraged patients to look for appropriate information on the subject. This study’s results may be useful for the care of renal patients on hemodialysis and the management of services, since it makes the educational video available as an evaluated product, which may contribute to the acquisition of self-care behaviors with AVF. In addition, the video as educational technology offers the advantage of allowing the content to be reproduced repeatedly in the dialysis room itself and without the need for a health professional to show it. In the field of research and teaching, this study contributes to the development of scientific knowledge on the promotion of self-care in chronic renal patients, as well as providing scientific evidence on the effectiveness of an educational strategy to promote self-care for chronic renal patients. Its results can be used to support further research on the subject, to improve the quality of life and adherence to treatment for patients with chronic renal failure, and to guide teaching practice.
|
Efficacy and safety of pediatric massage in the treatment of anorexia | 4f408329-bb47-439f-8159-36be8ec08472 | 8084013 | Pediatrics[mh] | Introduction Anorexia is a common and frequently-occurring disease in clinical pediatrics. It refers to a chronic digestive disorder syndrome in which appetite decreases or disappears and appetite is reduced. Epidemiological surveys abroad show that the incidence of aversion to eating in infants and preschool children is 12% to 34%. Children with anorexia in children will show gastrointestinal discomfort, abdominal pain, abdominal swelling, vomiting, constipation, and anorexia. The appearance of these symptoms can indicate to a certain extent that the child's digestive tract system has organic disease, Anorexia in children can lead to malnutrition. Long-term anorexia will affect the growth and development of children and reduce immunity, which will lead to a variety of diseases. In severe cases, children will develop infections of the central system or mental disorders. At present, the treatment of anorexia in children in modern medicine is mainly based on conventional drugs, such as the application of drugs that promote gastrointestinal motility, gastrointestinal biological agents, etc. These drugs can improve the symptoms of children to a certain extent, but a large number of clinical reports found that this type of drug has relatively large side effects, which is not conducive to the physical and mental health of patients and children. Pediatric tuina is an external Traditional Chinese Medicine treatment method. It does not require injections or medications. It acts on the meridians and acupoints of the human body to treat anorexia. It is safe, non-toxic, green, and effective. In view of more and more clinical reports, it is said that massage in children has a significant effect on the treatment of anorexia, but there is no relevant evidence-based medical evidence to confirm it. Therefore, this study will conduct a systematic review and meta-analysis of the effectiveness and safety of pediatric massage for the treatment of anorexia.
Methods 2.1 Study registration This protocol was registered with the International Platform of Registered Systematic Review and Meta-Analysis Protocols (INPLASY) on March 15, 2021 and was last updated on March 15, 2021 (registration number INPLASY202130050). 2.2 Inclusion criteria for study selection 2.2.1 Types of studies Clinical randomized controlled trials (RCT) containing pediatric massage for anorexia will be included, but do not limit language and publication status. 2.2.2 Types of participants There are clear and recognized diagnostic and curative criteria, and all patients are diagnosed with anorexia in children, regardless of sex, age, and source of cases. 2.2.3 Types of interventions 2.2.3.1 Experimental interventions Pediatric massage will include all different schools of pediatric massage techniques. Mixed therapy based on pediatric massage will also be included. 2.2.3.2 Control interventions The control group will receive one of the following treatments: routine pharmaceutical treatment, no treatment, and placebo. 2.2.4 Types of outcome measures 2.2.4.1 Primary outcome Clinical efficacy, including total effective rate or cure rate, clinical symptom score, will be considered as the primary outcome. 2.2.4.2 Secondary outcomes Body mass index (BMI), Traditional Chinese Medicine symptom score changes, and recurrence will be secondary outcomes. 2.3 Exclusion criteria Non-randomized controlled trials; no exact diagnostic scale or therapeutic scale; no moxibustion as the main treatment in the experimental group, and moxibustion therapy was found in the control group. Repeated literature; theory and review literature; animal experiments; nursing research. 2.4 The retrieval methods and strategies of this study 2.4.1 Electronic database retrieval We’ll retrieve 8 databases, the electronic databases, including the PubMed, Embase, Cochrane Library, Web Of Science, Chinese Biomedical Literature Database (CBM), China National Knowledge Infrastructure (CNKI), Wanfang Database (WF), China Science Journal Database (VIP), The retrieval date was established from the database to March 2021, without any language restriction. And will searching the relevant literature by combining subject words with free words, search terms consist (“children” or “pediatric” or “baby” or “infant” or “minors”) AND (“anorexia” or “cibophobia” or “Anorexia nervosa” or “piddle”) and intervention (“Massage” or “Tuina” or “manipulative therapy”) and research types (“randomized controlled trial” or “controlled clinical trial” or “random trials” or “RCT”). The PubMed search strategy is shown in Table . 2.4.2 Searching other resources We will combine manual retrieval of literature resource database to search relevant conference papers that meet the inclusion criteria. In addition, the grey literature, as well as ongoing and recently completed studies, will be searched on Clinicaltrials.gov. 2.5 Data extraction and management 2.5.1 Literature inclusion and data extraction Two reviewers will independently extract relevant data from the eligible RCTs, including the first author, participants’ baseline characteristics, sample size, intervention, intervention time, follow-up, results, and adverse events. Any discrepancies will be resolved through consultation with a third reviewer. If necessary, we will also contact the original author for more information. The inclusion process of this study will be carried out as shown in Fig. . 2.5.2 Methodological quality evaluation Two evaluators independently select the literature according to the inclusion and exclusion criteria and cross-check. In case of disagreement, a third evaluator will assist in the decision. The extracted data included the first author, year of publication, number of patients, age, sex, intervention measures, outcome indicators, etc. The Jadad scale to evaluate quality into literature, including: random sequence (right 2 points, 1 points not clear, inappropriate 0), distribution, hidden (right 2 points, 1 points not clear, inappropriate 0), blinded (right 2 points, 1 points not clear, inappropriate 0), lost to follow-up and exit (describe 1 points, not describe 0); 0–3 is classified as low quality and 4–7 as high quality. 2.6 Quantitative data synthesis and statistical methods 2.6.1 Quantitative data synthesis We will use RevMan V.5.3 software for statistical analysis. For continuous variables, when outcomes were measured by the same scale, the results were reported as standardized mean difference (MD) and 95% confidence interval (CI); when different scales were used, the results were reported as standardized mean difference (SMD) and 95% CI. Categorical data will be calculated with the risk ratio (RR) and 95% CI. 2.6.2 Assessment of heterogeneity We will use I 2 test and Chi-square test to evaluate the heterogeneity of the results. When I 2 ≤ 50% and P > .10, the results of the study will be considered as homogeneous, and fixed effect model will be used; otherwise, random effect model will be used. 2.6.3 Subgroup analysis If significant heterogeneity is detected in our meta-analysis, we will perform subgroup analysis based on different control groups. 2.6.4 Sensitivity analysis When there are sufficient RCTs, we will conduct sensitivity analysis based on methodological quality, sample size, and missing data to evaluate the robustness of the research results. 2.6.5 Assessment of reporting biases Publication bias will be analyzed through the funnel plot. If the funnel plot is asymmetric, there may be a publication bias in the research results.
Study registration This protocol was registered with the International Platform of Registered Systematic Review and Meta-Analysis Protocols (INPLASY) on March 15, 2021 and was last updated on March 15, 2021 (registration number INPLASY202130050).
Inclusion criteria for study selection 2.2.1 Types of studies Clinical randomized controlled trials (RCT) containing pediatric massage for anorexia will be included, but do not limit language and publication status. 2.2.2 Types of participants There are clear and recognized diagnostic and curative criteria, and all patients are diagnosed with anorexia in children, regardless of sex, age, and source of cases. 2.2.3 Types of interventions 2.2.3.1 Experimental interventions Pediatric massage will include all different schools of pediatric massage techniques. Mixed therapy based on pediatric massage will also be included. 2.2.3.2 Control interventions The control group will receive one of the following treatments: routine pharmaceutical treatment, no treatment, and placebo. 2.2.4 Types of outcome measures 2.2.4.1 Primary outcome Clinical efficacy, including total effective rate or cure rate, clinical symptom score, will be considered as the primary outcome. 2.2.4.2 Secondary outcomes Body mass index (BMI), Traditional Chinese Medicine symptom score changes, and recurrence will be secondary outcomes.
Types of studies Clinical randomized controlled trials (RCT) containing pediatric massage for anorexia will be included, but do not limit language and publication status.
Types of participants There are clear and recognized diagnostic and curative criteria, and all patients are diagnosed with anorexia in children, regardless of sex, age, and source of cases.
Types of interventions 2.2.3.1 Experimental interventions Pediatric massage will include all different schools of pediatric massage techniques. Mixed therapy based on pediatric massage will also be included. 2.2.3.2 Control interventions The control group will receive one of the following treatments: routine pharmaceutical treatment, no treatment, and placebo.
Experimental interventions Pediatric massage will include all different schools of pediatric massage techniques. Mixed therapy based on pediatric massage will also be included.
Control interventions The control group will receive one of the following treatments: routine pharmaceutical treatment, no treatment, and placebo.
Types of outcome measures 2.2.4.1 Primary outcome Clinical efficacy, including total effective rate or cure rate, clinical symptom score, will be considered as the primary outcome. 2.2.4.2 Secondary outcomes Body mass index (BMI), Traditional Chinese Medicine symptom score changes, and recurrence will be secondary outcomes.
Primary outcome Clinical efficacy, including total effective rate or cure rate, clinical symptom score, will be considered as the primary outcome.
Secondary outcomes Body mass index (BMI), Traditional Chinese Medicine symptom score changes, and recurrence will be secondary outcomes.
Exclusion criteria Non-randomized controlled trials; no exact diagnostic scale or therapeutic scale; no moxibustion as the main treatment in the experimental group, and moxibustion therapy was found in the control group. Repeated literature; theory and review literature; animal experiments; nursing research.
The retrieval methods and strategies of this study 2.4.1 Electronic database retrieval We’ll retrieve 8 databases, the electronic databases, including the PubMed, Embase, Cochrane Library, Web Of Science, Chinese Biomedical Literature Database (CBM), China National Knowledge Infrastructure (CNKI), Wanfang Database (WF), China Science Journal Database (VIP), The retrieval date was established from the database to March 2021, without any language restriction. And will searching the relevant literature by combining subject words with free words, search terms consist (“children” or “pediatric” or “baby” or “infant” or “minors”) AND (“anorexia” or “cibophobia” or “Anorexia nervosa” or “piddle”) and intervention (“Massage” or “Tuina” or “manipulative therapy”) and research types (“randomized controlled trial” or “controlled clinical trial” or “random trials” or “RCT”). The PubMed search strategy is shown in Table . 2.4.2 Searching other resources We will combine manual retrieval of literature resource database to search relevant conference papers that meet the inclusion criteria. In addition, the grey literature, as well as ongoing and recently completed studies, will be searched on Clinicaltrials.gov.
Electronic database retrieval We’ll retrieve 8 databases, the electronic databases, including the PubMed, Embase, Cochrane Library, Web Of Science, Chinese Biomedical Literature Database (CBM), China National Knowledge Infrastructure (CNKI), Wanfang Database (WF), China Science Journal Database (VIP), The retrieval date was established from the database to March 2021, without any language restriction. And will searching the relevant literature by combining subject words with free words, search terms consist (“children” or “pediatric” or “baby” or “infant” or “minors”) AND (“anorexia” or “cibophobia” or “Anorexia nervosa” or “piddle”) and intervention (“Massage” or “Tuina” or “manipulative therapy”) and research types (“randomized controlled trial” or “controlled clinical trial” or “random trials” or “RCT”). The PubMed search strategy is shown in Table .
Searching other resources We will combine manual retrieval of literature resource database to search relevant conference papers that meet the inclusion criteria. In addition, the grey literature, as well as ongoing and recently completed studies, will be searched on Clinicaltrials.gov.
Data extraction and management 2.5.1 Literature inclusion and data extraction Two reviewers will independently extract relevant data from the eligible RCTs, including the first author, participants’ baseline characteristics, sample size, intervention, intervention time, follow-up, results, and adverse events. Any discrepancies will be resolved through consultation with a third reviewer. If necessary, we will also contact the original author for more information. The inclusion process of this study will be carried out as shown in Fig. . 2.5.2 Methodological quality evaluation Two evaluators independently select the literature according to the inclusion and exclusion criteria and cross-check. In case of disagreement, a third evaluator will assist in the decision. The extracted data included the first author, year of publication, number of patients, age, sex, intervention measures, outcome indicators, etc. The Jadad scale to evaluate quality into literature, including: random sequence (right 2 points, 1 points not clear, inappropriate 0), distribution, hidden (right 2 points, 1 points not clear, inappropriate 0), blinded (right 2 points, 1 points not clear, inappropriate 0), lost to follow-up and exit (describe 1 points, not describe 0); 0–3 is classified as low quality and 4–7 as high quality.
Literature inclusion and data extraction Two reviewers will independently extract relevant data from the eligible RCTs, including the first author, participants’ baseline characteristics, sample size, intervention, intervention time, follow-up, results, and adverse events. Any discrepancies will be resolved through consultation with a third reviewer. If necessary, we will also contact the original author for more information. The inclusion process of this study will be carried out as shown in Fig. .
Methodological quality evaluation Two evaluators independently select the literature according to the inclusion and exclusion criteria and cross-check. In case of disagreement, a third evaluator will assist in the decision. The extracted data included the first author, year of publication, number of patients, age, sex, intervention measures, outcome indicators, etc. The Jadad scale to evaluate quality into literature, including: random sequence (right 2 points, 1 points not clear, inappropriate 0), distribution, hidden (right 2 points, 1 points not clear, inappropriate 0), blinded (right 2 points, 1 points not clear, inappropriate 0), lost to follow-up and exit (describe 1 points, not describe 0); 0–3 is classified as low quality and 4–7 as high quality.
Quantitative data synthesis and statistical methods 2.6.1 Quantitative data synthesis We will use RevMan V.5.3 software for statistical analysis. For continuous variables, when outcomes were measured by the same scale, the results were reported as standardized mean difference (MD) and 95% confidence interval (CI); when different scales were used, the results were reported as standardized mean difference (SMD) and 95% CI. Categorical data will be calculated with the risk ratio (RR) and 95% CI. 2.6.2 Assessment of heterogeneity We will use I 2 test and Chi-square test to evaluate the heterogeneity of the results. When I 2 ≤ 50% and P > .10, the results of the study will be considered as homogeneous, and fixed effect model will be used; otherwise, random effect model will be used. 2.6.3 Subgroup analysis If significant heterogeneity is detected in our meta-analysis, we will perform subgroup analysis based on different control groups. 2.6.4 Sensitivity analysis When there are sufficient RCTs, we will conduct sensitivity analysis based on methodological quality, sample size, and missing data to evaluate the robustness of the research results. 2.6.5 Assessment of reporting biases Publication bias will be analyzed through the funnel plot. If the funnel plot is asymmetric, there may be a publication bias in the research results.
Quantitative data synthesis We will use RevMan V.5.3 software for statistical analysis. For continuous variables, when outcomes were measured by the same scale, the results were reported as standardized mean difference (MD) and 95% confidence interval (CI); when different scales were used, the results were reported as standardized mean difference (SMD) and 95% CI. Categorical data will be calculated with the risk ratio (RR) and 95% CI.
Assessment of heterogeneity We will use I 2 test and Chi-square test to evaluate the heterogeneity of the results. When I 2 ≤ 50% and P > .10, the results of the study will be considered as homogeneous, and fixed effect model will be used; otherwise, random effect model will be used.
Subgroup analysis If significant heterogeneity is detected in our meta-analysis, we will perform subgroup analysis based on different control groups.
Sensitivity analysis When there are sufficient RCTs, we will conduct sensitivity analysis based on methodological quality, sample size, and missing data to evaluate the robustness of the research results.
Assessment of reporting biases Publication bias will be analyzed through the funnel plot. If the funnel plot is asymmetric, there may be a publication bias in the research results.
Discussion Anorexia in children is a common and frequently-occurring disease in children characterized by aversion to eating for a long time, poor appetite, or reduced appetite. The clinical manifestations are loss of appetite and food intake below the normal level by >60%. Some children may refuse to eat. If children's appetite is not improved for a long time, it will cause malnutrition, anemia, weight loss, affect growth and development, and cause severe impact on children's immunity and intelligence, as well as increase susceptibility to other system diseases. The pediatric massage of traditional Chinese medicine can reconcile qi and blood, adjust yin and yang, effectively improve the patient's spleen and stomach weakness, alleviate the patient's indigestion, and can completely cure the disease. At present, more and more studies believe that pediatric massage has a good clinical effect in the treatment of anorexia, and the clinical acceptance is high, but the author has not yet seen a systematic review of the effectiveness and safety of related pediatric massage in anorexia. Therefore, this article needs to systematically evaluate and meta-analyze the safety and effectiveness of Tuina therapy in children with anorexia, so as to provide evidence-based medical evidence for future clinical guidance.
Data curation: Baoxiu Yi, Wenguang Chen. Formal analysis: Baoxiu Yi, Gen Deng. Investigation: Baoxiu Yi, Wenguang Chen. Methodology: Wenguang Chen, Yiyi Wang. Project administration: Jinfeng Wang, Zhenhai Chi. Software: Gen Deng, Jinfeng Wang. Supervision: Baoxiu Yi, Zhenhai Chi. Validation: Gen Deng, Zhenhai Chi. Visualization: Yiyi Wang, Jinfeng Wang. Writing – original draft: Baoxiu Yi, Zhenhai Chi. Writing – review & editing: Baoxiu Yi, Zhenhai Chi.
|
Online Information Exchange and Anxiety Spread in the Early Stage of the Novel Coronavirus (COVID-19) Outbreak in South Korea: Structural Topic Model and Network Analysis | 91382efb-9d91-45f3-9b81-932937be32ac | 7268668 | Health Communication[mh] | The recent appearance of the novel coronavirus disease (COVID-19) has been devastating worldwide. In South Korea, hundreds of new cases have been diagnosed daily since late February 2020. The cumulative number of confirmed cases at the time of writing (April 1, 2020) exceeded 9000. Internationally, over 800,000 cases have been confirmed in more than 200 countries, areas, and territories , despite the World Health Organization’s request for global efforts to slow down the spread of the virus the previous month . Most countries have strongly recommended basic preventive methods such as quarantine and isolation of suspected cases, a macrolevel campaign on improving personal hygiene (eg, more frequent hand washing), or using masks in public sites. Additionally, some countries including Korea are now implementing more severe measures such as a social distancing that require the general population to refrain from congregating in public places. In the event of a population-wide infectious disease outbreak such as COVID-19 people’s online activities could significantly affect public concerns and health behaviors. Many studies have indicated people’s active use of online information in various crisis situations, including a public health crisis . Information and emotional exchanges between people on the internet form public opinions and concerns, which in turn affects people’s cognition and behavior. Although these opinions and information on the internet are sometimes useful, they are not always appropriate. There could be dissemination of misinformation, which may lead to inappropriate medical advice or unnecessary anxiety . Analyzing data on the internet that records how people voluntarily exchange opinions and information about COVID-19, such as relevant posts from social media services, provides a valuable opportunity to understand and monitor the public concerns over COVID-19 and the dissemination of related information on the internet. Considering the need to manage rumors and monitor public opinions and behaviors in the context of a mass infectious disease, and the importance of internet opinions in the event of a population-wide outbreak, the analysis of the online data has great implications for the formation of efficient and effective health policies and appropriate provision of information. These spontaneously written language materials contain a wealth of information about various topics on COVID-19 that cannot be thoroughly predicted by health policy makers and public health researchers, and thus cannot be measured by a traditional predetermined questionnaire. Hence, analyzing web-based data can supplement traditional surveys and contribute to making health policies for the general population . Web data analysis is particularly valuable in the early stage of an infectious disease outbreak. In the early stage of a new disease outbreak, health authorities may lack proper guidelines for the disease, and people may not find trustworthy information from other sources. Because of this situation, people might be more affected by uncertain information on the internet. Thus, monitoring web data in the early stage of an infectious disease outbreak is important to prevent inappropriate dissemination of misinformation or unnecessary anxiety that could occur in the early stage of an outbreak. This study primarily evaluated the public concerns over COVID-19 in the outbreak’s early stage using data from the online social questions and answers (Q&A) forum in Korea’s largest search engine, Naver.com , and analyzed the characteristics of item responses. The Naver Q&A forum (the service named “Jisik-In,” meaning “an intellectual” in Korean) resembles Quora.com, as it allows users to post questions and answers on any topic. We used 13,148 questions and threaded answers posted on Naver’s Q&A forum in the early stage of the COVID-19 outbreak to analyze the characteristics of the public’s online concerns and the appropriateness of information circulated there. In summary, our research questions include the following: What is the focus of the questions on COVID-19 in the Naver Q&A forum? How do the subjects observed in numerous questions change by time and main events? What are the main objects concerning anxiety and worry on COVID-19? How appropriate or significant is the information provided in the answers to the questions communicating anxiety and worry?
Data Collection This study used the question and answer data available on Naver’s Q&A forum. The forum is open to the public and allows individuals to post both questions and answers, anonymously or otherwise. There were two main reasons for selecting Naver’s Q&A forum among numerous online services available for public opinion exchange. First, Naver is a service provider with a dominant market position in Korea. It receives about 30 million visitors daily and is estimated to be used by about 76% of Korean Internet users as the main search portal website based on a recent survey . Moreover, Naver is the only search engine of massive users with a Q&A forum that allows users to freely access the information exchange. Although similar information exchanges could happen in other internet communities and social network services, in general, only their members could approach and see them, so their influence is limited. Therefore, Naver’s Q&A forum data better illustrates the Korean public’s concern generated through online postings. Second, Unlike data from other social media, language data from a Q&A forum includes a detailed context of the author’s interests and feelings. The question and answer form allows the user to post detailed information because its aim is to help others understand the full situation. A Twitter post, for example, often simply reveals an author’s feelings or anxieties; however, a question from Naver’s Q&A forum explains the issue’s background. Our data, thus, may allow more informative analysis on general public concerns over COVID-19 than other sources of web-based data. COVID-19–related questions posted between January 20, 2020, and March 2, 2020, and their respective answers were collected from Naver’s Q&A forum and used for analysis. January 20, 2020, was chosen as the starting point for data collection since it corresponds with the diagnosis of the first patient with COVID-19 in Korea, and questions before this date are scarce. Duration of the previously mentioned data was then confirmed by considering the frequency and characteristics of the data during preprocessing. The procedure for identifying the COVID-19–related posts comprised several steps. First, one question and its attached answers were considered as one document. We collected all the documents (questions and answers) containing the word “코로나” (in English, “corona”) from December 30, 2019, to March 2, 2020. This, however, also extracts questions and answers using the word “코로나” that refers to objects other than COVID-19. Second, the data were reselected from the results of the first step using additional search criteria to select COVID-19–related questions and answers more accurately. The search criteria were as follows: ([코로나 or corona or 우한 or COVID] and [바이러스 or 폐렴]) or (코로나19 or 코로나 19 or 신종코로나 or 신종코로나 or COVID19 or COVID 19 or COVID-19). English and Korean words were mixed in the search criteria because Koreans use both languages commonly. The English translations of the Korean words used here are as follows: 코로나=corona, 우한=Wuhan, 바이러스=virus, 폐렴=pneumonia, 신종코로나=novel corona. Therefore, we included biological words such as virus or pneumonia, or a relatively formal name referring to this disease, such as COVID19 or corona19, in the search criteria. These criteria were used to identify questions and answers that were related to COVID-19 and contained at least a fraction of biological perspective, as we assumed that the biological words and formal name reflect the perspective at least a little. Since COVID-19 is a controversial issue in domestic politics, we searched for posts that contained at least some biological perspective and excluded posts written purely from a political viewpoint. Third, we isolated the questions from the selected data and again selected the questions satisfying the following search criteria: (코로나 or corona or 우한 or COVID). Questions including irrelevant words (eg, pets, guitars) were removed. Since dogs or cats could be infected with a different type of coronavirus than COVID-19, there were questions about it. We excluded the questions related to pets. Questions related to “corona,” the guitar-producing company in Korea, were also excluded. The advanced criteria were applied only to the questions because some users supplied answers without considering the questions’ contents. In other words, although rare, there were cases in which answers related to COVID-19 were attached to questions unrelated to COVID-19. This ensured that both questions and answers were related to COVID-19. Fourth, the duplicate questions were deleted. Q&A forum users sometimes reposted the same question or posted a similar question with a slight change in the wording, which were unhelpful. The criteria used to find duplicate questions were whether the first 50 or the last 50 characters of the question, including blank spaces, were duplicated. Fifth, the few questions (n=14) posted before January 20, 2020, were deleted, as the data in this period cannot represent public concern properly. As a result, 13,148 questions and 29,040 answers, which were assumed to be related to COVID-19, were collected. presents a schematic summary of this process. Data Analysis Several text-mining techniques, including structural topic modeling and word network analysis, were used to analyze public concerns from 13,148 questions. Language data analyses have often used human interpretation capability . The theme in language materials results from synthesizing various information, more than what is simply and explicitly expressed. Therefore, it is convenient to capture themes by mobilizing human ability to interpret texts. However, there is a clear limitation on the amount of data that can be processed by a few researchers. This explains why several previous studies analyzing medical-related media or internet posts used a small amount of sample data . These studies were also susceptible to the unfavorable effects of human researchers’ subjectivity. Using text mining techniques to extract useful information from large volumes of data using computers, this study objectively estimated public concerns from large volumes of language data. Although text mining allows us to examine a huge amount of data, these techniques usually cannot capture delicate nuances. For example, using basic text mining techniques, determining whether the answer is correct information or a rumor is somewhat difficult. This poses a challenge to the researcher because rumors, especially convincing rumors, use similar words and connection of words with valid information. To compensate for this, we also used a method that allows medical doctors (family medicine specialists) to categorize answers to questions on a particular topic and then analyze the characteristics of those answers. In summary, three main methods were used for analysis: the structural topic model (STM), network analysis, and professional qualitative classification. Each method is described in the following sections. Structural Topic Model The STM, a type of topic modeling method, was selected to extract the overall themes or focus of 13,148 questions and examine how the theme or focus changed over time. Like most topic modeling methods developed after the latent Dirichlet allocation (LDA), the STM can extract multiple topics and the topics’ probability distribution in each document from a large number of documents. The extracted topics and their distribution are information that summarizes the given documents . The topic estimation process is based on several assumptions. Topic modeling methods assume that a document is a simple set of words and a topic is a probability distribution of words (eg, cat: 0.015, dog: 0.01, pet: 0.009, etc). Each document contains multiple topics with specific probability distributions (eg, the first document: topic 1=0.4, topic 2=0.2, and topic 3=0.4). It is then assumed that individual documents were randomly generated from the topics and their distribution per document, and were not directly written by humans; thus, the most probable topics and their distributions are estimated considering the given data . Naturally, the probability distribution of words itself does not have intuitive meaning; however, we can interpret the meaning of the topic from the probability distribution of words. When a topic is expressed qualitatively, it mainly appears as an unequal use of words. For example, suppose a topic of “cancer screening test” exists in the language materials. Certain words like “screen” or “mammography” would be used more frequently in this material than other words. Therefore, if we can deduce the word probability distributions, which are likely to produce the given documents, we can also deduce the topics’ meanings by noting the high probability words in the corresponding probability distributions. Distributions of topics in each document are also important information for interpreting the topics because they can be used to identify how each topic is realized into language material. We can also identify documents that have the highest proportion on each individual topic. For example, we can identify the top ten documents with the highest proportion of topic 2 among all documents. After reading the ten documents, we can understand the detailed context and intuitive meaning of topic 2 more accurately. In short, the topic modeling method estimates topics and their probability distributions per document that explain the given documents appropriately. Although the probability distribution itself does not provide intuitive meaning, researchers could interpret the topics’ meaning. Besides this general function of topic modeling, the STM estimates how much of the document’s meta information affects the topics’ proportion or content . Meta information refers to other information that exists in the document outside the document’s content (eg, when a document was written or the type of author). The STM estimates how the meta information affects the proportion and content of the extracted topics. Given that this study primarily aimed to analyze how the topics of questions changed over time, these attributes of the STM were deemed appropriate to achieve our research objectives. This study estimated how the time of posting the question affected the proportion of question topics. The application of the STM to the questions is described as follows. Of the 13,148 questions, 12 questions were additionally eliminated in the preprocessing for the STM. We only used the words that appeared in at least 2 questions and questions that contained at least 2 words, as a word that only appears in a single document or a question that contains only 1 word carries little information for topic modeling. We estimated 50 topics from the remaining 13,136 questions. The number of topics was set at 50 because there was no significant improvement in topic modeling performance after 50 as measured by the held-out likelihood when changing the number of topics from 10 to 80 in increments of 5 . The posting time of the questions, which was measured by the unit of 1 day, was used as a covariate in our model for estimating topics’ proportion change by time. All the topics were interpreted and labeled by considering three kinds of information: the words given high probability in each topic, the words with a high frequency and exclusivity (FREX) score in each topic, and the documents with a high proportion of each topic. The importance of the high probability words and the documents estimated to have a high proportion of each topic in interpreting topics has been explained previously. The words with a high FREX score supplement the high probability words by considering exclusivity and frequency together . That is, a high FREX score word of a topic is important, especially in the topic. All the authors collectively interpreted and labeled 50 topics, considering the top 15 probability words, the top 15 FREX score words, and the top 10 questions with a high proportion of each topic. We also used the STM to sample representative questions containing subjects related to anxiety and worry. To analyze the appropriateness of the answers to the questions related to the public’s anxiety and worry, most representative questions needed to be reselected from the entire batch of questions. Since it is impossible for a small group of human researchers to review 13,148 questions directly, the STM results were used to select questions that contained anxiety- and worry-related topics. We extracted questions that contained a high proportion of topics discussing physical symptoms and self-protection methods against COVID-19 because they were revealed as the main targets of anxiety and worry. Network Analysis Topic modeling is useful for identifying broad topics in a large volume of documents; however, a researcher cannot control the model to see results on specific subjects. Therefore, analyzing the relationship between particular words of interest is important in achieving our research goals. One of our research objectives was to identify the objects and contexts of anxiety and worry expressed by people. To identify the source of anxiety and worry, we extracted the top 20 words that appeared most frequently in the questions including the words “불안” or “걱정,” which are Korean words for “anxiety” and “worry,” respectively. We considered both words appearing in the same question as a linkage or an association between the words. Additionally, we created a word network using the top 50 words linked to the two Korean words referring to anxiety and worry to analyze the context of people’s anxiety and worry comprehensively. Observing individual words associated with anxiety and worry cannot accurately analyze the full context of anxiety and worry. A network of words linked to anxiety and worry allowed us to further analyze the context of anxiety and worry. Therefore, the top 50 words associated with the two Korean words referring to anxiety and worry were gathered, and the network was created. Connection criterion is a coappearance of words in the same question. This network is extremely dense because the network is made of words that are used in a similar context. Only the most prominent links were required to be selected to extract particularly prominent meanings. For this, only the links with the highest weight, (ie, 500 of the most frequent connections) were extracted to create a subnetwork. This subnetwork was assumed to contain the most prominent contexts related to people’s anxiety and worry. The network community detection algorithm was applied to this subnetwork to extract distinguished themes from the network. This algorithm identifies relatively more cohesive communities of nodes within a network . When applied to our word network, it worked to find a set of words that appeared more frequently with each other in the entire network. For example, the algorithm could judge that a “China,” “Wuhan,” “pneumonia,” “infection,” and “travel” word cluster appeared frequently in the entire word network. This was interpreted as a distinct theme that we found through network analysis. In other words, the network community detection algorithm identified the most prominent and distinct contexts that appeared when people expressed anxiety or worry. We identified cohesive communities of words, which contained at least 5 words because this amount was required to interpret the word community into a meaningful theme. Walktrap was selected from the various network community detection algorithms because of its excellent performance while overcoming the “resolution problem” . A resolution problem refers to a situation in which algorithms do not capture a community of a small number of nodes properly, a frequently faced problem in applying network community detection algorithms. Put simply, the Walktrap algorithm calculates the distance between nodes in the network using a random walk from each node and finds communities based on that distance. The analyst sets the step of the random walk for distance calculation, which we set to 2. The network analysis was implemented through the igraph package of R (R Foundation for Statistical Computing). The Walktrap algorithm was also used to identify cohesive communities of topics related to each other. The STM estimates the correlation between topics . A positive correlation between two topics means that the two topics are likely to appear together in the same document. We assumed this positive correlation to be a link between topics and formed a network of all topics. Next, using the network community detection algorithm, we analyzed whether there were relatively more cohesive communities of topics in the entire network of topics. In other words, we tried to identify sets of topics that were often expressed together. Consequently, we identified 6 topic communities that were judged to be cohesive. We labeled each topic community to express broader themes embracing topics belonging to the topic community, considering each topic’s interpretation made through the STM. That is, after estimating 50 topics from 13,136 questions, these topics were summarized into 6 topic communities using the Walktrap algorithm and then interpreted. This additional step was employed due to the fact that, although 50 topics is a great summary, it is still a lot of information for a person to grasp intuitively. Professionals’ Qualitative Coding Using the results of the aforementioned methods, we assessed the appropriateness of the answers to questions dealing with the main targets of anxiety and worry, and found that the main themes were physical symptoms and self-protection methods. We selected sample questions and answers that addressed these two subjects. Two medical doctors, who are family medicine specialists and the authors of this paper, classified the answers into 5 independent categories: appropriate answers, unrelated answers, wrong answers, advertisement, and other. If there was a disagreement on an answer’s category, they discussed until they reached an agreement and recorded the agreed result. The sample questions were selected by considering the proportion of topics in each question that resulted from using the STM. The themes of physical symptom and self-protection correlated with the fourth and fifth topic communities, respectively. Choosing high proportion questions from each community of topics could form good sample questions that appropriately represent each theme. Therefore, the proportion of the topic communities 4 and 5 was calculated for each question based on the topic proportion information per question. It was created by adding up the proportion of topics belonging to each topic community. The top 100 questions were selected for both topic communities 4 and 5, and the answers to each question were identified. The number of answers was 250 and 306, respectively. Morphological Analysis and Part of Speech Tagging For the previously mentioned methods, especially STM and network analysis, to function properly, morphemes need to be extracted from our language data. That is, we need information about what kinds of words appeared from where and at what frequency. The program used to determine this is called a morphological analyzer. From the many types of morphological analyzers applicable to Korean, Komoran was used because it is resilient to the spacing problem in Korean and is sufficiently qualified, as it won awards from the National Institute of Korean Language in 2016. We would like to mention in advance a caveat regarding the words extracted from the data by Komoran; Korean and English do not respond 1:1. For example, there are several Korean words that can be translated as fever (“열” and “발열”). When translating such words into English, we numbered them (eg, fever_1, fever_2).
This study used the question and answer data available on Naver’s Q&A forum. The forum is open to the public and allows individuals to post both questions and answers, anonymously or otherwise. There were two main reasons for selecting Naver’s Q&A forum among numerous online services available for public opinion exchange. First, Naver is a service provider with a dominant market position in Korea. It receives about 30 million visitors daily and is estimated to be used by about 76% of Korean Internet users as the main search portal website based on a recent survey . Moreover, Naver is the only search engine of massive users with a Q&A forum that allows users to freely access the information exchange. Although similar information exchanges could happen in other internet communities and social network services, in general, only their members could approach and see them, so their influence is limited. Therefore, Naver’s Q&A forum data better illustrates the Korean public’s concern generated through online postings. Second, Unlike data from other social media, language data from a Q&A forum includes a detailed context of the author’s interests and feelings. The question and answer form allows the user to post detailed information because its aim is to help others understand the full situation. A Twitter post, for example, often simply reveals an author’s feelings or anxieties; however, a question from Naver’s Q&A forum explains the issue’s background. Our data, thus, may allow more informative analysis on general public concerns over COVID-19 than other sources of web-based data. COVID-19–related questions posted between January 20, 2020, and March 2, 2020, and their respective answers were collected from Naver’s Q&A forum and used for analysis. January 20, 2020, was chosen as the starting point for data collection since it corresponds with the diagnosis of the first patient with COVID-19 in Korea, and questions before this date are scarce. Duration of the previously mentioned data was then confirmed by considering the frequency and characteristics of the data during preprocessing. The procedure for identifying the COVID-19–related posts comprised several steps. First, one question and its attached answers were considered as one document. We collected all the documents (questions and answers) containing the word “코로나” (in English, “corona”) from December 30, 2019, to March 2, 2020. This, however, also extracts questions and answers using the word “코로나” that refers to objects other than COVID-19. Second, the data were reselected from the results of the first step using additional search criteria to select COVID-19–related questions and answers more accurately. The search criteria were as follows: ([코로나 or corona or 우한 or COVID] and [바이러스 or 폐렴]) or (코로나19 or 코로나 19 or 신종코로나 or 신종코로나 or COVID19 or COVID 19 or COVID-19). English and Korean words were mixed in the search criteria because Koreans use both languages commonly. The English translations of the Korean words used here are as follows: 코로나=corona, 우한=Wuhan, 바이러스=virus, 폐렴=pneumonia, 신종코로나=novel corona. Therefore, we included biological words such as virus or pneumonia, or a relatively formal name referring to this disease, such as COVID19 or corona19, in the search criteria. These criteria were used to identify questions and answers that were related to COVID-19 and contained at least a fraction of biological perspective, as we assumed that the biological words and formal name reflect the perspective at least a little. Since COVID-19 is a controversial issue in domestic politics, we searched for posts that contained at least some biological perspective and excluded posts written purely from a political viewpoint. Third, we isolated the questions from the selected data and again selected the questions satisfying the following search criteria: (코로나 or corona or 우한 or COVID). Questions including irrelevant words (eg, pets, guitars) were removed. Since dogs or cats could be infected with a different type of coronavirus than COVID-19, there were questions about it. We excluded the questions related to pets. Questions related to “corona,” the guitar-producing company in Korea, were also excluded. The advanced criteria were applied only to the questions because some users supplied answers without considering the questions’ contents. In other words, although rare, there were cases in which answers related to COVID-19 were attached to questions unrelated to COVID-19. This ensured that both questions and answers were related to COVID-19. Fourth, the duplicate questions were deleted. Q&A forum users sometimes reposted the same question or posted a similar question with a slight change in the wording, which were unhelpful. The criteria used to find duplicate questions were whether the first 50 or the last 50 characters of the question, including blank spaces, were duplicated. Fifth, the few questions (n=14) posted before January 20, 2020, were deleted, as the data in this period cannot represent public concern properly. As a result, 13,148 questions and 29,040 answers, which were assumed to be related to COVID-19, were collected. presents a schematic summary of this process.
Several text-mining techniques, including structural topic modeling and word network analysis, were used to analyze public concerns from 13,148 questions. Language data analyses have often used human interpretation capability . The theme in language materials results from synthesizing various information, more than what is simply and explicitly expressed. Therefore, it is convenient to capture themes by mobilizing human ability to interpret texts. However, there is a clear limitation on the amount of data that can be processed by a few researchers. This explains why several previous studies analyzing medical-related media or internet posts used a small amount of sample data . These studies were also susceptible to the unfavorable effects of human researchers’ subjectivity. Using text mining techniques to extract useful information from large volumes of data using computers, this study objectively estimated public concerns from large volumes of language data. Although text mining allows us to examine a huge amount of data, these techniques usually cannot capture delicate nuances. For example, using basic text mining techniques, determining whether the answer is correct information or a rumor is somewhat difficult. This poses a challenge to the researcher because rumors, especially convincing rumors, use similar words and connection of words with valid information. To compensate for this, we also used a method that allows medical doctors (family medicine specialists) to categorize answers to questions on a particular topic and then analyze the characteristics of those answers. In summary, three main methods were used for analysis: the structural topic model (STM), network analysis, and professional qualitative classification. Each method is described in the following sections. Structural Topic Model The STM, a type of topic modeling method, was selected to extract the overall themes or focus of 13,148 questions and examine how the theme or focus changed over time. Like most topic modeling methods developed after the latent Dirichlet allocation (LDA), the STM can extract multiple topics and the topics’ probability distribution in each document from a large number of documents. The extracted topics and their distribution are information that summarizes the given documents . The topic estimation process is based on several assumptions. Topic modeling methods assume that a document is a simple set of words and a topic is a probability distribution of words (eg, cat: 0.015, dog: 0.01, pet: 0.009, etc). Each document contains multiple topics with specific probability distributions (eg, the first document: topic 1=0.4, topic 2=0.2, and topic 3=0.4). It is then assumed that individual documents were randomly generated from the topics and their distribution per document, and were not directly written by humans; thus, the most probable topics and their distributions are estimated considering the given data . Naturally, the probability distribution of words itself does not have intuitive meaning; however, we can interpret the meaning of the topic from the probability distribution of words. When a topic is expressed qualitatively, it mainly appears as an unequal use of words. For example, suppose a topic of “cancer screening test” exists in the language materials. Certain words like “screen” or “mammography” would be used more frequently in this material than other words. Therefore, if we can deduce the word probability distributions, which are likely to produce the given documents, we can also deduce the topics’ meanings by noting the high probability words in the corresponding probability distributions. Distributions of topics in each document are also important information for interpreting the topics because they can be used to identify how each topic is realized into language material. We can also identify documents that have the highest proportion on each individual topic. For example, we can identify the top ten documents with the highest proportion of topic 2 among all documents. After reading the ten documents, we can understand the detailed context and intuitive meaning of topic 2 more accurately. In short, the topic modeling method estimates topics and their probability distributions per document that explain the given documents appropriately. Although the probability distribution itself does not provide intuitive meaning, researchers could interpret the topics’ meaning. Besides this general function of topic modeling, the STM estimates how much of the document’s meta information affects the topics’ proportion or content . Meta information refers to other information that exists in the document outside the document’s content (eg, when a document was written or the type of author). The STM estimates how the meta information affects the proportion and content of the extracted topics. Given that this study primarily aimed to analyze how the topics of questions changed over time, these attributes of the STM were deemed appropriate to achieve our research objectives. This study estimated how the time of posting the question affected the proportion of question topics. The application of the STM to the questions is described as follows. Of the 13,148 questions, 12 questions were additionally eliminated in the preprocessing for the STM. We only used the words that appeared in at least 2 questions and questions that contained at least 2 words, as a word that only appears in a single document or a question that contains only 1 word carries little information for topic modeling. We estimated 50 topics from the remaining 13,136 questions. The number of topics was set at 50 because there was no significant improvement in topic modeling performance after 50 as measured by the held-out likelihood when changing the number of topics from 10 to 80 in increments of 5 . The posting time of the questions, which was measured by the unit of 1 day, was used as a covariate in our model for estimating topics’ proportion change by time. All the topics were interpreted and labeled by considering three kinds of information: the words given high probability in each topic, the words with a high frequency and exclusivity (FREX) score in each topic, and the documents with a high proportion of each topic. The importance of the high probability words and the documents estimated to have a high proportion of each topic in interpreting topics has been explained previously. The words with a high FREX score supplement the high probability words by considering exclusivity and frequency together . That is, a high FREX score word of a topic is important, especially in the topic. All the authors collectively interpreted and labeled 50 topics, considering the top 15 probability words, the top 15 FREX score words, and the top 10 questions with a high proportion of each topic. We also used the STM to sample representative questions containing subjects related to anxiety and worry. To analyze the appropriateness of the answers to the questions related to the public’s anxiety and worry, most representative questions needed to be reselected from the entire batch of questions. Since it is impossible for a small group of human researchers to review 13,148 questions directly, the STM results were used to select questions that contained anxiety- and worry-related topics. We extracted questions that contained a high proportion of topics discussing physical symptoms and self-protection methods against COVID-19 because they were revealed as the main targets of anxiety and worry. Network Analysis Topic modeling is useful for identifying broad topics in a large volume of documents; however, a researcher cannot control the model to see results on specific subjects. Therefore, analyzing the relationship between particular words of interest is important in achieving our research goals. One of our research objectives was to identify the objects and contexts of anxiety and worry expressed by people. To identify the source of anxiety and worry, we extracted the top 20 words that appeared most frequently in the questions including the words “불안” or “걱정,” which are Korean words for “anxiety” and “worry,” respectively. We considered both words appearing in the same question as a linkage or an association between the words. Additionally, we created a word network using the top 50 words linked to the two Korean words referring to anxiety and worry to analyze the context of people’s anxiety and worry comprehensively. Observing individual words associated with anxiety and worry cannot accurately analyze the full context of anxiety and worry. A network of words linked to anxiety and worry allowed us to further analyze the context of anxiety and worry. Therefore, the top 50 words associated with the two Korean words referring to anxiety and worry were gathered, and the network was created. Connection criterion is a coappearance of words in the same question. This network is extremely dense because the network is made of words that are used in a similar context. Only the most prominent links were required to be selected to extract particularly prominent meanings. For this, only the links with the highest weight, (ie, 500 of the most frequent connections) were extracted to create a subnetwork. This subnetwork was assumed to contain the most prominent contexts related to people’s anxiety and worry. The network community detection algorithm was applied to this subnetwork to extract distinguished themes from the network. This algorithm identifies relatively more cohesive communities of nodes within a network . When applied to our word network, it worked to find a set of words that appeared more frequently with each other in the entire network. For example, the algorithm could judge that a “China,” “Wuhan,” “pneumonia,” “infection,” and “travel” word cluster appeared frequently in the entire word network. This was interpreted as a distinct theme that we found through network analysis. In other words, the network community detection algorithm identified the most prominent and distinct contexts that appeared when people expressed anxiety or worry. We identified cohesive communities of words, which contained at least 5 words because this amount was required to interpret the word community into a meaningful theme. Walktrap was selected from the various network community detection algorithms because of its excellent performance while overcoming the “resolution problem” . A resolution problem refers to a situation in which algorithms do not capture a community of a small number of nodes properly, a frequently faced problem in applying network community detection algorithms. Put simply, the Walktrap algorithm calculates the distance between nodes in the network using a random walk from each node and finds communities based on that distance. The analyst sets the step of the random walk for distance calculation, which we set to 2. The network analysis was implemented through the igraph package of R (R Foundation for Statistical Computing). The Walktrap algorithm was also used to identify cohesive communities of topics related to each other. The STM estimates the correlation between topics . A positive correlation between two topics means that the two topics are likely to appear together in the same document. We assumed this positive correlation to be a link between topics and formed a network of all topics. Next, using the network community detection algorithm, we analyzed whether there were relatively more cohesive communities of topics in the entire network of topics. In other words, we tried to identify sets of topics that were often expressed together. Consequently, we identified 6 topic communities that were judged to be cohesive. We labeled each topic community to express broader themes embracing topics belonging to the topic community, considering each topic’s interpretation made through the STM. That is, after estimating 50 topics from 13,136 questions, these topics were summarized into 6 topic communities using the Walktrap algorithm and then interpreted. This additional step was employed due to the fact that, although 50 topics is a great summary, it is still a lot of information for a person to grasp intuitively. Professionals’ Qualitative Coding Using the results of the aforementioned methods, we assessed the appropriateness of the answers to questions dealing with the main targets of anxiety and worry, and found that the main themes were physical symptoms and self-protection methods. We selected sample questions and answers that addressed these two subjects. Two medical doctors, who are family medicine specialists and the authors of this paper, classified the answers into 5 independent categories: appropriate answers, unrelated answers, wrong answers, advertisement, and other. If there was a disagreement on an answer’s category, they discussed until they reached an agreement and recorded the agreed result. The sample questions were selected by considering the proportion of topics in each question that resulted from using the STM. The themes of physical symptom and self-protection correlated with the fourth and fifth topic communities, respectively. Choosing high proportion questions from each community of topics could form good sample questions that appropriately represent each theme. Therefore, the proportion of the topic communities 4 and 5 was calculated for each question based on the topic proportion information per question. It was created by adding up the proportion of topics belonging to each topic community. The top 100 questions were selected for both topic communities 4 and 5, and the answers to each question were identified. The number of answers was 250 and 306, respectively. Morphological Analysis and Part of Speech Tagging For the previously mentioned methods, especially STM and network analysis, to function properly, morphemes need to be extracted from our language data. That is, we need information about what kinds of words appeared from where and at what frequency. The program used to determine this is called a morphological analyzer. From the many types of morphological analyzers applicable to Korean, Komoran was used because it is resilient to the spacing problem in Korean and is sufficiently qualified, as it won awards from the National Institute of Korean Language in 2016. We would like to mention in advance a caveat regarding the words extracted from the data by Komoran; Korean and English do not respond 1:1. For example, there are several Korean words that can be translated as fever (“열” and “발열”). When translating such words into English, we numbered them (eg, fever_1, fever_2).
The STM, a type of topic modeling method, was selected to extract the overall themes or focus of 13,148 questions and examine how the theme or focus changed over time. Like most topic modeling methods developed after the latent Dirichlet allocation (LDA), the STM can extract multiple topics and the topics’ probability distribution in each document from a large number of documents. The extracted topics and their distribution are information that summarizes the given documents . The topic estimation process is based on several assumptions. Topic modeling methods assume that a document is a simple set of words and a topic is a probability distribution of words (eg, cat: 0.015, dog: 0.01, pet: 0.009, etc). Each document contains multiple topics with specific probability distributions (eg, the first document: topic 1=0.4, topic 2=0.2, and topic 3=0.4). It is then assumed that individual documents were randomly generated from the topics and their distribution per document, and were not directly written by humans; thus, the most probable topics and their distributions are estimated considering the given data . Naturally, the probability distribution of words itself does not have intuitive meaning; however, we can interpret the meaning of the topic from the probability distribution of words. When a topic is expressed qualitatively, it mainly appears as an unequal use of words. For example, suppose a topic of “cancer screening test” exists in the language materials. Certain words like “screen” or “mammography” would be used more frequently in this material than other words. Therefore, if we can deduce the word probability distributions, which are likely to produce the given documents, we can also deduce the topics’ meanings by noting the high probability words in the corresponding probability distributions. Distributions of topics in each document are also important information for interpreting the topics because they can be used to identify how each topic is realized into language material. We can also identify documents that have the highest proportion on each individual topic. For example, we can identify the top ten documents with the highest proportion of topic 2 among all documents. After reading the ten documents, we can understand the detailed context and intuitive meaning of topic 2 more accurately. In short, the topic modeling method estimates topics and their probability distributions per document that explain the given documents appropriately. Although the probability distribution itself does not provide intuitive meaning, researchers could interpret the topics’ meaning. Besides this general function of topic modeling, the STM estimates how much of the document’s meta information affects the topics’ proportion or content . Meta information refers to other information that exists in the document outside the document’s content (eg, when a document was written or the type of author). The STM estimates how the meta information affects the proportion and content of the extracted topics. Given that this study primarily aimed to analyze how the topics of questions changed over time, these attributes of the STM were deemed appropriate to achieve our research objectives. This study estimated how the time of posting the question affected the proportion of question topics. The application of the STM to the questions is described as follows. Of the 13,148 questions, 12 questions were additionally eliminated in the preprocessing for the STM. We only used the words that appeared in at least 2 questions and questions that contained at least 2 words, as a word that only appears in a single document or a question that contains only 1 word carries little information for topic modeling. We estimated 50 topics from the remaining 13,136 questions. The number of topics was set at 50 because there was no significant improvement in topic modeling performance after 50 as measured by the held-out likelihood when changing the number of topics from 10 to 80 in increments of 5 . The posting time of the questions, which was measured by the unit of 1 day, was used as a covariate in our model for estimating topics’ proportion change by time. All the topics were interpreted and labeled by considering three kinds of information: the words given high probability in each topic, the words with a high frequency and exclusivity (FREX) score in each topic, and the documents with a high proportion of each topic. The importance of the high probability words and the documents estimated to have a high proportion of each topic in interpreting topics has been explained previously. The words with a high FREX score supplement the high probability words by considering exclusivity and frequency together . That is, a high FREX score word of a topic is important, especially in the topic. All the authors collectively interpreted and labeled 50 topics, considering the top 15 probability words, the top 15 FREX score words, and the top 10 questions with a high proportion of each topic. We also used the STM to sample representative questions containing subjects related to anxiety and worry. To analyze the appropriateness of the answers to the questions related to the public’s anxiety and worry, most representative questions needed to be reselected from the entire batch of questions. Since it is impossible for a small group of human researchers to review 13,148 questions directly, the STM results were used to select questions that contained anxiety- and worry-related topics. We extracted questions that contained a high proportion of topics discussing physical symptoms and self-protection methods against COVID-19 because they were revealed as the main targets of anxiety and worry.
Topic modeling is useful for identifying broad topics in a large volume of documents; however, a researcher cannot control the model to see results on specific subjects. Therefore, analyzing the relationship between particular words of interest is important in achieving our research goals. One of our research objectives was to identify the objects and contexts of anxiety and worry expressed by people. To identify the source of anxiety and worry, we extracted the top 20 words that appeared most frequently in the questions including the words “불안” or “걱정,” which are Korean words for “anxiety” and “worry,” respectively. We considered both words appearing in the same question as a linkage or an association between the words. Additionally, we created a word network using the top 50 words linked to the two Korean words referring to anxiety and worry to analyze the context of people’s anxiety and worry comprehensively. Observing individual words associated with anxiety and worry cannot accurately analyze the full context of anxiety and worry. A network of words linked to anxiety and worry allowed us to further analyze the context of anxiety and worry. Therefore, the top 50 words associated with the two Korean words referring to anxiety and worry were gathered, and the network was created. Connection criterion is a coappearance of words in the same question. This network is extremely dense because the network is made of words that are used in a similar context. Only the most prominent links were required to be selected to extract particularly prominent meanings. For this, only the links with the highest weight, (ie, 500 of the most frequent connections) were extracted to create a subnetwork. This subnetwork was assumed to contain the most prominent contexts related to people’s anxiety and worry. The network community detection algorithm was applied to this subnetwork to extract distinguished themes from the network. This algorithm identifies relatively more cohesive communities of nodes within a network . When applied to our word network, it worked to find a set of words that appeared more frequently with each other in the entire network. For example, the algorithm could judge that a “China,” “Wuhan,” “pneumonia,” “infection,” and “travel” word cluster appeared frequently in the entire word network. This was interpreted as a distinct theme that we found through network analysis. In other words, the network community detection algorithm identified the most prominent and distinct contexts that appeared when people expressed anxiety or worry. We identified cohesive communities of words, which contained at least 5 words because this amount was required to interpret the word community into a meaningful theme. Walktrap was selected from the various network community detection algorithms because of its excellent performance while overcoming the “resolution problem” . A resolution problem refers to a situation in which algorithms do not capture a community of a small number of nodes properly, a frequently faced problem in applying network community detection algorithms. Put simply, the Walktrap algorithm calculates the distance between nodes in the network using a random walk from each node and finds communities based on that distance. The analyst sets the step of the random walk for distance calculation, which we set to 2. The network analysis was implemented through the igraph package of R (R Foundation for Statistical Computing). The Walktrap algorithm was also used to identify cohesive communities of topics related to each other. The STM estimates the correlation between topics . A positive correlation between two topics means that the two topics are likely to appear together in the same document. We assumed this positive correlation to be a link between topics and formed a network of all topics. Next, using the network community detection algorithm, we analyzed whether there were relatively more cohesive communities of topics in the entire network of topics. In other words, we tried to identify sets of topics that were often expressed together. Consequently, we identified 6 topic communities that were judged to be cohesive. We labeled each topic community to express broader themes embracing topics belonging to the topic community, considering each topic’s interpretation made through the STM. That is, after estimating 50 topics from 13,136 questions, these topics were summarized into 6 topic communities using the Walktrap algorithm and then interpreted. This additional step was employed due to the fact that, although 50 topics is a great summary, it is still a lot of information for a person to grasp intuitively.
Using the results of the aforementioned methods, we assessed the appropriateness of the answers to questions dealing with the main targets of anxiety and worry, and found that the main themes were physical symptoms and self-protection methods. We selected sample questions and answers that addressed these two subjects. Two medical doctors, who are family medicine specialists and the authors of this paper, classified the answers into 5 independent categories: appropriate answers, unrelated answers, wrong answers, advertisement, and other. If there was a disagreement on an answer’s category, they discussed until they reached an agreement and recorded the agreed result. The sample questions were selected by considering the proportion of topics in each question that resulted from using the STM. The themes of physical symptom and self-protection correlated with the fourth and fifth topic communities, respectively. Choosing high proportion questions from each community of topics could form good sample questions that appropriately represent each theme. Therefore, the proportion of the topic communities 4 and 5 was calculated for each question based on the topic proportion information per question. It was created by adding up the proportion of topics belonging to each topic community. The top 100 questions were selected for both topic communities 4 and 5, and the answers to each question were identified. The number of answers was 250 and 306, respectively.
For the previously mentioned methods, especially STM and network analysis, to function properly, morphemes need to be extracted from our language data. That is, we need information about what kinds of words appeared from where and at what frequency. The program used to determine this is called a morphological analyzer. From the many types of morphological analyzers applicable to Korean, Komoran was used because it is resilient to the spacing problem in Korean and is sufficiently qualified, as it won awards from the National Institute of Korean Language in 2016. We would like to mention in advance a caveat regarding the words extracted from the data by Komoran; Korean and English do not respond 1:1. For example, there are several Korean words that can be translated as fever (“열” and “발열”). When translating such words into English, we numbered them (eg, fever_1, fever_2).
Frequency of Documents and Words Our data included a total of 13,148 questions. presents the number of questions sorted by date. Of the words that appeared in the questions, presents the top 30 frequency words. Those occupying the top 5 positions include “cough,” “symptom,” “throat,” “mask,” and “confirmed diagnosis.” Structural Topic Model About 50 topics were estimated from 13,136 questions using the STM and were interpreted by the authors. presents the results of the interpretation. The first column from the left is the topic number, the second column is the topic’s interpretation, and the third column is the topic community number that each topic belongs to. Topic numbers and topic community numbers are nominal numbers for distinction. Most topics had an apparent subject or consistent contents that allowed a precise interpretation. Some topics, however, were extracted because of expressions that appeared repeatedly in various questions, regardless of the content of the question. This is because topic modeling captures the coappearance patterns of multiple words observed in documents without considering the meanings of words. For example, if many questions on different subjects contain similar expressions, like “please answer my question or you could be cursed,” the topic model would capture that pattern and estimate the topic based on the pattern. In our model, several topics were extracted because of unique Korean language usages. In this case, we unified the interpretation of the topic as “Questions involving particular Korean language expressions without a common subject.” Additionally, we presented the most prominent Korean expression in these topics in Korean in parentheses and translated its meaning into English. The third column in resulted from applying Walktrap to the topics’ correlation network for grouping topics into several cohesive communities. A total of 50 topics were categorized into 6 topic communities, each given a number (the rightmost column). The topics in were sorted according to the topic community numbers to help readers easily identify topics belonging to each topic community. presents a visualization of the entire topics network and communities of topics identified through Walktrap. A node is a topic, the number below a node is the topic number, a link is a positive correlation between topics, and the color of the node indicates the topic grouping. Therefore, topics of the same color belong to the same community. presents the result of the authors’ interpretations of topic communities. The STM calculates the proportions of topics on a per-document basis. This allowed us to calculate the proportion of each topic in the entire document, which could also be used to calculate the proportion of the topic communities in the entire document. We aggregated proportions of topics belonging to the same community. Additionally, the STM calculated the proportion variation of all topics over time because the time variable was set as a covariate in our STM. The estimates of topics’ proportion changes over time could also be aggregated to produce the proportion change of the topic communities over time. and present the results. Topic community 4 (questions suspecting possible COVID-19 infection after developing a particular symptom) occupied the largest proportion of all the questions. It increased sharply in late February when the number of infections had begun to increase earnestly in Korea. In , the yellow line (topic community 4) and the dashed line (the number of confirmed patients) are almost parallel after the second increase for topic community 4, the yellow line mimicking the increase in the number of confirmed patients. Moreover, it is noticeable that the portion of topic community 2 (concerns over working conditions caused by COVID-19) increased slightly as the COVID-19 situation became prolonged. Network Analysis We extracted 20 words that were most frequently connected with words referring to anxiety and worry (in Korean, “불안” and “걱정,” respectively) from the questions posted from January 20 to March 1, 2020. presents the results. In , the top 5 words include “cough,” “symptom,” “throat,” “mask,” and “cold.” This allowed us to infer that people’s anxiety was centered on physical symptoms and key self-protection methods, such as wearing masks. We also checked whether the list of words associated with anxiety and worry would vary with time. The period (January 20-March 1, 2020) was divided into 6 weeks, and the top words associated with anxiety and worry was extracted from the data of the 6 subperiods . During the week of January 20-26, when the first confirmed COVID-19 cases were reported, the primary subject of anxiety was “China and traveling.” This was natural, considering that COVID-19 was limited to the Chinese Mainland at that time. However, as the number of confirmed COVID-19 cases increased in Korea, words related to physical symptoms appeared as the top words. The period after February 17 was when the number of confirmed cases increased sharply in South Korea by 2 or 3 digits. Henceforth, Koreans needed to be careful about contact with confirmed patients, and the word “confirmed diagnosis” emerged as the main word associated with anxiety. The word “mask” was consistently included in the top 10 words in all periods. In other words, the anxiety and worry about self-protection have been prevalent consistently throughout the entire period. We formed a word network using the top 50 words linked to anxiety and worry and extracted its subnetwork based on the most prominent 500 links. By applying the Walktrap algorithm to the subnetwork, we could extract three cohesive word communities or three distinct themes. presents a visualization of the three sets of words comprising at least 5 words. One is related to physical symptoms, another is related to self-protection, and the last is related to China. In other words, people’s anxiety was expressed in three main themes. This result was consistent with our reasoning based on the types of words linked to anxiety and worry. Professionals’ Qualitative Coding shows the results of two medical doctors’ categorization of the answers to sample questions dealing with the main targets of anxiety and worry into five categories. The answers to questions about physical symptoms were often appropriate and relatively less distorted. On the other hand, there were many advertising answers to questions related to self-protection measures.
Our data included a total of 13,148 questions. presents the number of questions sorted by date. Of the words that appeared in the questions, presents the top 30 frequency words. Those occupying the top 5 positions include “cough,” “symptom,” “throat,” “mask,” and “confirmed diagnosis.”
About 50 topics were estimated from 13,136 questions using the STM and were interpreted by the authors. presents the results of the interpretation. The first column from the left is the topic number, the second column is the topic’s interpretation, and the third column is the topic community number that each topic belongs to. Topic numbers and topic community numbers are nominal numbers for distinction. Most topics had an apparent subject or consistent contents that allowed a precise interpretation. Some topics, however, were extracted because of expressions that appeared repeatedly in various questions, regardless of the content of the question. This is because topic modeling captures the coappearance patterns of multiple words observed in documents without considering the meanings of words. For example, if many questions on different subjects contain similar expressions, like “please answer my question or you could be cursed,” the topic model would capture that pattern and estimate the topic based on the pattern. In our model, several topics were extracted because of unique Korean language usages. In this case, we unified the interpretation of the topic as “Questions involving particular Korean language expressions without a common subject.” Additionally, we presented the most prominent Korean expression in these topics in Korean in parentheses and translated its meaning into English. The third column in resulted from applying Walktrap to the topics’ correlation network for grouping topics into several cohesive communities. A total of 50 topics were categorized into 6 topic communities, each given a number (the rightmost column). The topics in were sorted according to the topic community numbers to help readers easily identify topics belonging to each topic community. presents a visualization of the entire topics network and communities of topics identified through Walktrap. A node is a topic, the number below a node is the topic number, a link is a positive correlation between topics, and the color of the node indicates the topic grouping. Therefore, topics of the same color belong to the same community. presents the result of the authors’ interpretations of topic communities. The STM calculates the proportions of topics on a per-document basis. This allowed us to calculate the proportion of each topic in the entire document, which could also be used to calculate the proportion of the topic communities in the entire document. We aggregated proportions of topics belonging to the same community. Additionally, the STM calculated the proportion variation of all topics over time because the time variable was set as a covariate in our STM. The estimates of topics’ proportion changes over time could also be aggregated to produce the proportion change of the topic communities over time. and present the results. Topic community 4 (questions suspecting possible COVID-19 infection after developing a particular symptom) occupied the largest proportion of all the questions. It increased sharply in late February when the number of infections had begun to increase earnestly in Korea. In , the yellow line (topic community 4) and the dashed line (the number of confirmed patients) are almost parallel after the second increase for topic community 4, the yellow line mimicking the increase in the number of confirmed patients. Moreover, it is noticeable that the portion of topic community 2 (concerns over working conditions caused by COVID-19) increased slightly as the COVID-19 situation became prolonged.
We extracted 20 words that were most frequently connected with words referring to anxiety and worry (in Korean, “불안” and “걱정,” respectively) from the questions posted from January 20 to March 1, 2020. presents the results. In , the top 5 words include “cough,” “symptom,” “throat,” “mask,” and “cold.” This allowed us to infer that people’s anxiety was centered on physical symptoms and key self-protection methods, such as wearing masks. We also checked whether the list of words associated with anxiety and worry would vary with time. The period (January 20-March 1, 2020) was divided into 6 weeks, and the top words associated with anxiety and worry was extracted from the data of the 6 subperiods . During the week of January 20-26, when the first confirmed COVID-19 cases were reported, the primary subject of anxiety was “China and traveling.” This was natural, considering that COVID-19 was limited to the Chinese Mainland at that time. However, as the number of confirmed COVID-19 cases increased in Korea, words related to physical symptoms appeared as the top words. The period after February 17 was when the number of confirmed cases increased sharply in South Korea by 2 or 3 digits. Henceforth, Koreans needed to be careful about contact with confirmed patients, and the word “confirmed diagnosis” emerged as the main word associated with anxiety. The word “mask” was consistently included in the top 10 words in all periods. In other words, the anxiety and worry about self-protection have been prevalent consistently throughout the entire period. We formed a word network using the top 50 words linked to anxiety and worry and extracted its subnetwork based on the most prominent 500 links. By applying the Walktrap algorithm to the subnetwork, we could extract three cohesive word communities or three distinct themes. presents a visualization of the three sets of words comprising at least 5 words. One is related to physical symptoms, another is related to self-protection, and the last is related to China. In other words, people’s anxiety was expressed in three main themes. This result was consistent with our reasoning based on the types of words linked to anxiety and worry.
shows the results of two medical doctors’ categorization of the answers to sample questions dealing with the main targets of anxiety and worry into five categories. The answers to questions about physical symptoms were often appropriate and relatively less distorted. On the other hand, there were many advertising answers to questions related to self-protection measures.
In the event of a novel infectious disease outbreak, the general population cannot easily assess the accuracy of the information regarding the disease, and there is increased reliance on online information. Obtaining appropriate and accurate information is extremely difficult, especially in the early stage of an outbreak, due to the uncertainty about the disease. There may be a delay before the governance body, such as health authorities, announces official statements regarding the disease, including symptoms, treatment, or preventive measures. However, it is unlikely that the public would refrain from seeking information and patiently wait until the accurate information becomes publicly available. In this scenario, it is more probable that the members of the populace would seek to acquire the much-needed information from other sources such as the internet. Consistent with previous studies, our data showed that people sought information regarding COVID-19 on the web. The analysis of 13,136 questions revealed that the largest proportion of topics was regarding anxiety and worries about possible COVID-19 infection after developing a particular symptom. The proportion of topics regarding concerns over working conditions also slightly increased as the COVID-19 outbreak became prolonged. Physical symptoms such as cough, throat pain, and sputum as well as self-protective measures such as wearing a mask were some top key words that simultaneously appeared with the words anxiety and worry in the word network analysis. This implies that the people were mainly concerned about whether developing a particular physical symptom was relevant to COVID-19 and ways to protect them from COVID-19. We also analyzed the appropriateness of the answers that replied to the questions. About 63% of the total answers to questions on suspecting possible COVID-19 infection after developing particular symptoms were evaluated as appropriate, while 15.6% of the answers were incorrect, implying a potential dissemination of misinformation. For questions regarding self-protective measures, such as questions asking how to wear masks properly, as much as 66.3% of the answers were advertisements. Thus, it can be assumed that the general population may have difficulty in obtaining appropriate information on self-protective measures. This study contributes to the establishment of early health communication about public concerns and anxiety observed at the early stage of COVID-19. The initial stage of the epidemic is when neither the health authorities’ policies nor people’s understanding of the epidemic has stabilized. Under these circumstances, people’s online exchange of information and emotion can have a great influence. Governments should implement proper measures to establish online health communication in the early stage of an outbreak to provide appropriate and accurate information. Considering that web data-based studies related to COVID-19 are rare at present, our research’s policy and academic value is more pronounced. Although there are studies that use web data to analyze various characteristics of public psychology for other infectious diseases, there are few studies related to COVID-19. Given that numerous countries have been affected recently and are in the initial stage of the COVID-19 outbreak, our study on South Korea, which experienced COVID-19 relatively early, could be a reference point for policy making in other countries. This study also contributes to devising methods of measuring public psychology using language data in the circumstance of an infectious disease outbreak. Understanding public psychology and culture is essential when dealing with infectious diseases because public psychology greatly impacts the management of infectious diseases . Public anger toward the infected, for example, is common in infectious disease outbreaks, which can induce the infected to hide from quarantine efforts such as screening tests. This is because the infected people will try to avoid the intense social anger directed at them. To prevent this, it is necessary to promptly measure public sentiment in detail and organize appropriate responses such as creating social support for the infected. However, as previously indicated, traditional surveys are relatively difficult to implement quickly and are not free from limitations (eg, effect of researchers’ frame on answers, relative difficulty of collecting real time data) . Recently, analysis of language data using computers and statistical models has been introduced, and several scholars have suggested its usefulness . Our paper provides examples of its use concerning an infectious disease outbreak. Beyond the Q&A forum, various language materials are available on the internet, which can be used to actively supplement existing methods of investigation and create diverse methods to approximate public anxiety. In short, this study shows the potential for “online data-based health policy decision making.” Additionally, this study is differentiated from other studies in how we used text mining techniques. Previously, papers using health-related text data often used frequency as the main information, such as the number of Twitter mentions , or applied LDA , which is the most commonly used topic modeling technique . By using the STM, this study more systematically analyzed how the proportions of COVID-19–related topics varied over time while maintaining the advantages of topic modeling methods. Moreover, this study does not merely apply existing text mining techniques to health-related data but also contributes methodologically on how to use topic modeling methods. Various topic modeling methods have been used as the LDA in many studies , most of which draw conclusions from the estimated topics. However, solely focusing on topics has one downside: if the amount of data is substantial, the number of topics will also increase, making it difficult to identify overall patterns that appear in the entire data set. Too many topics, for example 300 topics, represent vast information, which would likely pose a formidable barrier to human researchers. Our research has introduced a way to find sets of topics connected to each other by forming a network of topics and finding cohesive topic communities in them through a network community detection algorithm. In our results, 6 topic communities were found and each community contained content-related topics. It is surprising that the topics belonging to each topic community are meaningfully interrelated, as the topic communities are derived from the community detection algorithm and a correlation matrix deduced by the STM results, not through human researchers’ categorization based on the topics’ contents. In other words, this study proposes a data-driven method of making topic clusters, which could be used for detecting broader themes from numerous topics that were estimated from a vast number of documents. Various advanced analyses are possible using the data and results of this study. One of the notable results in our study is that the proportion of topic community 4 (questions suspecting possible COVID-19 infection after developing a particular symptom) seems to be linked to the actual number of confirmed patients. Considering that the symptoms observed in topic community 4 (eg, cough, throat pain, and sputum) were reported as common symptoms of COVID-19 in other clinical reports, this linkage might not be a coincidence. This suggests that information extracted from web data may help identify and even predict the actual trend of infectious diseases. Future research could use sophisticated time series analyses to scrutinize this possibility. Additionally, it is noteworthy that a certain kind of question could have a high proportion of answers written from commercial motivations . This implies that there are many attempts to commercially exploit the infectious disease crisis, and the social effects related to these attempts could be explored using this study’s results. Finally, if adequate data of a longer period is available, analyzing how the importance of online information and communication changes over time would be a valuable research project. These further studies will enhance the efficient use of online data for public health. This study has limitations concerning the range of data. Although Naver is the most popular portal website in Korea, people are not just using this one service. For further comprehensive analysis of information and emotion exchange in online spaces, various sources of web data, including various social network services, need to be incorporated. Furthermore, since the internet user population does not appropriately represent the entire population, it is necessary to consider using the data produced through traditional methods such as surveys and the internet natural language data together. Nevertheless, this study showed how health information exchanged based on the disease transmission related to people’s anxiety and commercial interest in new infectious disease outbreak via a novel approach using online data analysis and topic modeling.
|
Mechanism of herbal medicine on hypertensive nephropathy | f534f38f-7b1c-4e98-baaa-3d5999604c3e | 7893801 | Pharmacology[mh] | Introduction The association between high blood pressure and the kidney is high . The kidney participates in the formation of blood pressure through the secretion of renin and the regulation of body fluids . The imbalance of such regulation leads to hypertension. In addition, the kidney is one of the important target organs affected by hypertension-associated damage . The renal tubules are sensitive to ischemia, and dysfunction of distal tubules concentration often occurs first, including increased nocturia, decreased urine-specific gravity and decreased urine osmotic pressure . Over time, proteinuria, mostly mild, may occur after glomerular ischemia, while moderate proteinuria may occur in certain patients with high blood pressure . The decrease in the glomerular filtration rate eventually leads to end-stage renal disease (ESRD). In terms of renal pathology, the kidney size is normal at the early stage, while the kidney volume is decreased at the late stage, and its surface is fine and granular. This is caused by arteriolosclerosis, which leads to renal parenchymal damage and ischemic sclerosis in certain glomeruli . Ischemic lesions are also found in the renal interstitium and tubules. In terms of treatment, the current therapeutic aim of modern medicine is only to protect the residual nephrons and to delay the progression of renal damage using antihypertensive drugs . Herbal medicine has been historically used in the treatment of hypertensive nephropathy, and its clinical effect is remarkable. Therefore, it is of great clinical and economical importance to explore its mechanism of action. The present study aims to review the recent basic research on the mechanism of herbal medicine for the treatment of hypertensive nephropathy.
Pathogenesis of hypertensive nephropathy Hypertensive nephropathy is more common in elderly patients >50 years of age with a long history of chronic hypertension, and is more prevalent in men compared with women . The clinical manifestations appear later than the pathological changes . Hypertension usually lasts for >10 years before the gradual emergence of nocturia, mild proteinuria and other associated clinical symptoms . This disease is usually preceded by distal tubular dysfunction, followed by glomerular dysfunction . The patient's renal pathology starts with renal arteriopathy, followed by ischemic renal parenchymal damage . Renal arteriolosclerosis mainly affects the arterioles before the glomerulus, including hyaline degeneration of the glomerular arterioles, and thickening of the medium membrane of the interlobular and arcuate arteries . The reason for this is vascular endothelial damage and increased vascular cavity pressure, which results in the subcutaneous accumulation of plasma components . Hypertrophy and hyperplasia of membranous smooth muscle cells in the interlobular and arcuate arteries are accompanied by different degrees of intimal fibrosis . Both lesions lead to hardening and thickening of the arteriole walls, narrowing of the lumen and decreased renal blood supply, followed by ischemic renal parenchymal damage . As arteriopathy progresses, the glomerulus first undergo ischemic shrinkage, namely capillary basement membrane wrinkling, while the lumen remains open. Next, ischemic sclerosis occurs, where the basement membrane is highly wrinkled and the capillary lumen collapses . The renal tubules and interstitium are also ischemic, including tubules atrophy, thickening of the basement membrane, interstitial fibrosis and limited mononuclear cell infiltration . When parts of the glomeruli are damaged, the remaining glomeruli compensatively enhance the discharge of metabolic waste substances, eventually leading to glomerulosclerosis. The mechanisms of hypertensive nephropathy is mainly renal hemodynamic changes and vascular remodeling caused by hypertension . When hypertension occurs, changes in renal hemodynamics will lead to changes in the function and structure of renal arterioles, which is known as vascular remodeling . During hypertension, the change in arteriole function is mainly manifested as increased responsiveness to vasoconstrictive substances , which results in increased vascular resistance and decreased renal plasma flow. However, in previous studies, the renal function remained normal due to increased glomerular filtration fraction. If hypertension persists, it can lead to structural changes in renal arterioles, particularly in the interlobular and arcuate arteries, and to hypertrophy and proliferation of smooth muscle cells . The mechanism is complex, and is the result of various active substances in the circulation, as well as an imbalance of vascular endothelial synthesis and secretion. For example, endothelin-1 (ET-1) is increased, while nitric oxide (NO) is decreased . Finally, the renal arteriole wall thickens, the lumen narrows, the vascular compliance decreases, the renal plasma flow further decreases and the renal function is damaged. However, not all the renal arterioles undergo hypertrophic remodeling, resulting in hypoperfusion and ischemic renal parenchymal damage. In fact, the arterioles in the other part of the kidney do not show hypertrophic remodeling, but rather compensatory hyperperfusion . The glomeruli, which are supplied by these arteries, also change from exhibiting hypertrophy to exhibiting focal segmental sclerosis . Although the purpose of hypertensive nephropathy treatment is to protect the residual nephron and delay the progression of renal damage, the key aim of the treatment is to effectively control blood pressure . Therefore, the majority of research on the treatment of hypertensive nephropathy focuses on hypertension. Moreover, relevant studies on the control and treatment of hypertensive nephropathy using traditional Chinese medicine via a variety of mechanisms also investigate hypertension . The basic studies on herbal medicine prescribed for hypertensive nephropathy are shown in .
Mechanism of herbal medicine in the treatment of hypertensive nephropathy Suppression of the renin-angiotensin system (RAS) The RAS plays an important role locally in the kidney. Angiotensin II (AngII) can directly bind to angiotensin receptors on renal arteriolar smooth muscle cells and stimulate vascular smooth muscle contraction . AngII also stimulates the sympathetic nerve to promote vascular smooth muscle resistance, thus leading to increased renal vascular resistance. In addition, AngII can increase sodium reabsorption through the aldosterone action on distal renal tubules, thus increasing the blood volume and leading to increased blood pressure . Numerous basic studies have shown that herbal medicine can play a crucial role in the treatment of hypertensive nephropathy by inhibiting RAS. Genipin, as one of the main components of Gardenia , can protect the renal function of spontaneously hypertensive rat (SHR) via the AngII-TLR/MyD88/mitogen activated protein kinase (MAPK) pathway . Qian Yang Yu Yin granules can suppress AngII in multiple manners. The mechanism includes alleviation of SHR and inhibition of 293T cells' effort induced by AngII through the epigenetic pathway associated with nicotinamide N-methyltransferase expression . The Jiangya Tongluo formula can regulate the protective effect of adrenomedullin and angiotensin in rats with hypertensive nephrosis . The heart-protecting musk pill can decrease the partial levels of AngII in SHR kidney, thus treating hypertensive nephropathy . In addition to studies on hypertensive nephropathy, numerous studies have demonstrated that herbal medicine can show efficacy in the treatment of chronic kidney disease or hypertension by inhibiting the RAS. For example, Chrysanthemum acts as an antihypertensive by acting on the RAS . The water extracts of kidney bean sprouts have been demonstrated to inhibit angiotensin converting enzyme, thus exhibiting potential for lowering blood pressure . Alisol B 23 acetate, as one of the main ingredients of Rhizoma alismatis , can suppress the expression of constituents of the RAS, and can inhibit the epithelial-to-mesenchymal transition (EMT) in nephrectomised rats, thus lowering blood pressure, decreasing serum creatinine and preventing proteinuria . In addition, Alisol B 23 acetate can block the RAS/Wnt/β-catenin axis to improve podocyte injury and the EMT of HK-2 cells . Ergone, one of the main ingredients of Polyporus umbellatus , and pachymic acid B, one of the main ingredients of Poria cocos , have the same effect . In addition, poricoic acid ZA, ZF, ZG and ZH, which are important components of Poria cocos , inhibit the effect of the RAS to protect podocytes and renal tubular epithelial cells, but affect the RAS and the transforming growth factor-β1 (TGF-β1)/Smad axis . Previous studies have shown that poricoic acid ZC, ZD and ZE in Poria Cocos protect renal interstitial fibrosis due to unilateral ureteral obstruction in mice via TGFβ/Smad pathway . It has been reported that 25-O-methylalisol F, the main component of Alisma , protects EMT of rat renal proximal tubular epithelial cell lines through this pathway . The therapeutic effect of Radix Scrophulariae on SHR can be attributed to the suppression of the RAS through the inhibition of the extracellular regulated protein kinase 1/2, c-Jun N-terminal kinase and p38 MAPK pathways . Xin-Ji-Er-Kang can inhibit oxidative stress by affecting the RAS, and can improve renal injury after myocardial infarction in rats . In addition, all herbal medicines that contain flavonoids, terpenoids, saponins and alkaloids are able to inhibit the RAS . Among them, common herbs containing flavonoids are Scutellaria baicalensis, Flos cmysanthemi, Sambucus adnata wall, bud of Chinese Scholar tree, Equisetum spp, Chrysanthemum indicum L., Chamaecyparis obtusa, Orthosiphon stamineus and Tropaeolum Majus L. . Common herbs containing terpenoids are the surface layer and sclerotium of Poria cocos, Alismatis rhizome and Polyporus umbellatus . A common herb containing saponins is the ginseng root . Common herbs containing alkaloids are Gambirplant, leonurus, Ophora flavescens, S. subprostrata, S. alopecuroides and Uncaria rhynchophylla . Inhibition of sympathetic excitation In patients with hypertension, the sympathetic adrenaline system is hyperactive from the central to the arterial walls. The synthesis and release of the neurotransmitter catecholamine increases, thus leading to renal arteriole contraction, and renal vascular resistance increases, thus affecting vascular remodeling . In addition, the catecholamine released by sympathetic nerves can directly act on proximal renal tubules, and increase sodium reabsorption, blood volume and blood pressure . Although it has not been demonstrated yet that the mechanism of herbal medicine in the treatment of hypertensive nephropathy involves the regulation of the sympathetic nervous system, numerous herbal medicines have been reported to be able to play a role in the inhibition of sympathetic nervous system in previous basic experiments and clinical trials. Chrysanthemum plays a role in decreasing blood pressure by inhibiting the sympathetic nerve . Radix scrophulariae , by inhibiting sympathetic excitement, suppresses SHR, and ventricular remodeling occurs . Astragaloside IV can decrease norepinephrine levels in the blood of high-fat diet-induced obese rats and in kidney tissues, which indirectly demonstrates that Astragaloside IV has the effect of inhibiting sympathetic nerves . Guizhi decoction can inhibit the cholinergic transdifferentiation of sympathetic nerves, and improve the anatomical and functional denervation of sympathetic nerves . In addition, acupuncture, electroacupuncture and moxibustion can also regulate the sympathetic nervous system, although their mechanism of action is complex . Antioxidant stress and anti-inflammatory responses Oxidative stress is caused by the imbalance of reactive oxygen species (ROS) and the antioxidant mechanism in the body. In hypertensive nephropathy, inflammatory damage is caused by the interaction of various cells such as macrophages and T lymphocytes, or inflammatory mediators or chemokines . These inflammatory cells secrete cytokines that can lead to endothelial dysfunction, which can aggravate and even lead to hypertension. Inflammatory reactions and oxidative stress play a common role and cause each other in hypertension-associated renal damage . Therefore, anti-oxidative stress and anti-inflammation can play a role in alleviating hypertensive nephropathy. As the most widely used herb in cardiovascular diseases, Salvia miltiorrhiza can significantly improve SHR blood pressure, decrease ROS production and improve vascular remodeling . The flower of Coreopsis tinctoria Nutt. is widely used in the treatment of hypertension, diabetes, obesity and other diseases. It exerts anti-inflammatory effects through its antioxidant stress properties and its ability to inhibit tumor necrosis factor-α (TNF-α), interleukin-6 (IL-6) and nuclear factor-κB (NF-κB) . Brazilian red propolis could alleviate hypertension and kidney injury in 5/6 renal ablation model rats through antioxidant stress . As an extract of Apocynum venetum , its polyphenols can improve the renal index of D-galactose-induced oxidative stress in mouse models . Tulbaghia violacea can improve NF-κB and TGF-β expression in Dahl salt-sensitive rat kidneys, and plays a role in lowering blood pressure and protecting renal function . Resveratrol, the main component of Veratrum nigrum L., has been demonstrated to have anti-ROS effects, and has the potential to lower blood pressure . As the main component of celery seeds, 3-n-butylphthalide plays a protective role in renal tubules through decreased stress, as well as the expression of pro-inflammatory cytokines and TGF-β1 in kidney tissues . Resveratrol, the main component of Veratrum nigrum L., has been demonstrated to have anti-ROS effects, and has the potential to lower blood pressure . Paeonol can effectively improve the blood pressure of spontaneously hypertensive rats, and its mechanism may be associated with reduction of blood viscosity, antioxidant stress and improvement of antioxidant capacity . Galangin, as the main extract of Alpinia officinarum hance , inhibits ROS as well as the mRNA expression of prostaglandin-endoperoxide synthase 2, TNF-α, IL-1β and IL-18, thus exerting a protective effect in rat renal epithelial cells . Icariin, as the main component of Epimedium brevicornu , can decrease the production of ROS by inhibiting the activity of NADPH oxidase, thus reducing the vasoconstriction effect of AngII-induced hypertension in rats . Natural antioxidants derived from food and herbal extracts such as tea polyphenols, curcumin and lycopene, have been widely used as complementary therapies to slow the progression of ESRD. Jiang Ya Yi Shen granules exert their protective role by inhibiting NF-κB signaling-mediated micro-inflammatory cytokines, including IL-6, TNF-α and intercellular cell adhesion molecule-1 (ICAM-1), on SHR nephropathy . Tongxinluo can inhibit the effects of oxidative stress and improve SHR glomerular sclerosis . Ban Xia Bai Zhu Tian Ma decoction could inhibit IL-1, IL-6, TNF-α and inducible nitric oxide synthase (NOS) in SHR to improve the heart damage caused by hypertension . Qian Yang Yu Yin granules can also inhibit the influence of AngII on the NADPH oxidase 4-dependent pathway, thereby inhibiting the proliferation of human mesangial cells, and lowering the production of ROS and anti-inflammatory response . The heart-protecting musk pill, also called Shexiang Baoxin pill, can decrease TGF-β and ICAM-1, thus exerting an anti-inflammatory effect, and can be used to treat SHR nephropathy and to improve vascular remodeling . Xin-Ji-Er-Kang-induced NOS in high-salt induced hypertensive mice can improve the activity and oxidative stress, and alleviate vascular remodeling . Qingxuan Jiangya decoction can affect the TGF-β1/Smad signaling pathway to play a crucial role in improving renal interstitial fibrosis in SHR . In a previous study, Shenkang improved renal injury in mice with unilateral ureteral occlusion by acting on the TGF-β/Smad3, Sirtuin/forkhead box protein O and B-cell lymphoma-2-associated X protein pathways . Regulation of vasoactive substances and other mechanisms of endothelial cell protection Hypertension can promote the synthesis of endothelial cells and the secretion of a variety of vasoactive substances. These substances maintain vascular tension and permeability, but can lead to vascular smooth muscle hypertrophy and hyperplasia. The sustained effect of blood pressure on vascular endothelial cells will result in endothelial cell damage . In addition, the increase in endogenous plasma NO synthase inhibitors in patients affects the decrease in NO synthesis by endothelial cells . The increase in ET-1, which can lead to vasoconstriction, eventually leads to enhanced vasoconstriction response and increased renal vascular resistance, and promotes the occurrence of vascular remodeling . Therefore, the best indicator of endothelial cell function is observation of the dynamic changes in vasoactive factors such as NO and ET-1. Cirsium japonicum improves the cardiac effects of renal hypertension in 2-kidney 1-clip rats by increasing serum NO levels . Morinda citrifolia can significantly decrease blood pressure and 24-h urinary NO metabolite in SHR, and its juice extract can increase the phosphorylation of endothelial NOS in human umbilical vein endothelial cells, and promote the endothelial vasodilation of the aortic ring and NO products in rats . Zingiber officinale var. rubrum exerts a significant vascular relaxation effect in SHR. Its possible mechanism of vasodilatation includes the release of NO or transmembrane calcium channels . Curcumin can protect the renal kidney function of cadmium-induced renal damage in rats, and can play a protective role on renal injury caused by hyperuricemia or high-fructose intake, and one of the mechanisms is to increase the production of NO . Morin (also known as 3,5,7,2′,4′-pentahydroxyflavone) is widely present in fruits and vegetables such as almond, old fustic, Indian guava and Osage orange. This compound may play a strong role in vascular widening by NO, muscarinic receptors, β2-adrenegic receptors and calcium channels . Hydroxysafflor yellow A, the principal component of Carthamus tinctorius L., induces angiogenesis in rat mesenteric arteries by transient receptor potential vanilloid 4 (TRPV4) -dependent calcium influx in endothelial cells . A large number of clinical experiments showed that sodium tanshinone IIA sulfonate combined with angiotensin receptor blockers (ARBs) had a stronger effect on improving renal function in patients with primary hypertensive nephropathy compared with ARB monotherapy . Sodium tanshinone IIA sulfonate, the main ingredient in the herb Salvia miltiorrhiza , has been shown to protect vascular endothelial cells. In addition, the combination of caffeic acid and ferulic acid can dilate blood vessels and resist ET-1, while exerting a hypotensive effect through ester bonds and telmisartan . Qingxuan Jiangya decoction can prevent hypertension and improve vascular remodeling in SHR by lowering the serum ET-1 level and inhibiting the TGF-β1/Smad pathway . 17-Methoxyl-7-hydroxy-furanchalcone, as an active ingredient of Fordia cauliflora , was capable of improving cardiac reconstruction from hypertension in rats by regulating the eNOS-NO signaling pathway . The combination of Astragalus membranaceus and Salvia miltiorrhiza can improve IL-1β levels in SHR urine and eNOS levels in AngII-damaged human renal glomerular endothelial cells superfluid . San Cao decoction in network pharmacologic analysis may play a role in lowering blood pressure by regulating the PI3K-Akt-eNOS pathway . Improvement of obesity-associated factors Metabolic disorder is also an important cause of hypertensive nephropathy . Obesity plays a greater role than blood pressure in the progression of hypertensive kidney disease . Obesity itself is a risk factor for high blood pressure. And in obese patients, renal dysfunction and associated increased sodium reabsorption in renal tubules can lead to hypertension . The compression of perirenal fat on the kidneys results in the activation of RAS . Chronic obesity may gradually amplify hypertension, leading to resistance to antihypertensive treatment. . Insulin resistance leads to the constriction of the extruded arterioles, thus leading to high glomerular pressure, hyperperfusion and hyperfiltration . These studies have demonstrated that obesity is closely associated with the incidence of hypertensive nephropathy. Herbal medicine has unique advantages in improving obesity. Astragaloside IV, as one of the main ingredients of Astragalus , is used to treat hypertension in high-fat diet-induced obese rats due to its anti-inflammatory effect and its ability to improve leptin resistance . Citrus paradisi and Ocimum sanctum infusions can decrease blood pressure and protect kidney function in obese rats . A number of studies have shown that Chinese herbs can improve the effects of obesity on the kidneys of patients. Coptidis rhizoma can lower the blood lipid level and renal weight of fat-prone rats, and can improve urinary protein creatinine ratio and creatinine clearance rate in rats . The mechanism may be associated with the inhibition of the NLRP3 inflammasome . Through treatment of obesity-associated glomerulopathy in model rats with Tribulus terrestris L., it was found that the herb could decrease the body weight, blood pressure, serum cystatin C levels and migration of rats, as well as improve human endothelial cells migration, thus protecting renal function . Curcumin, as one of the most important components of turmeric, can improve body weight, abdominal fat index, urinary protein excretion and average glomerular diameter in mice, and can protect podocytes from leptin damage by blocking the Wnt/β-catenin pathway . At the formula level, Mai Tong Fang inhibits fat generation and triglyceride accumulation in 3T3-L1 adipocytes .
The RAS plays an important role locally in the kidney. Angiotensin II (AngII) can directly bind to angiotensin receptors on renal arteriolar smooth muscle cells and stimulate vascular smooth muscle contraction . AngII also stimulates the sympathetic nerve to promote vascular smooth muscle resistance, thus leading to increased renal vascular resistance. In addition, AngII can increase sodium reabsorption through the aldosterone action on distal renal tubules, thus increasing the blood volume and leading to increased blood pressure . Numerous basic studies have shown that herbal medicine can play a crucial role in the treatment of hypertensive nephropathy by inhibiting RAS. Genipin, as one of the main components of Gardenia , can protect the renal function of spontaneously hypertensive rat (SHR) via the AngII-TLR/MyD88/mitogen activated protein kinase (MAPK) pathway . Qian Yang Yu Yin granules can suppress AngII in multiple manners. The mechanism includes alleviation of SHR and inhibition of 293T cells' effort induced by AngII through the epigenetic pathway associated with nicotinamide N-methyltransferase expression . The Jiangya Tongluo formula can regulate the protective effect of adrenomedullin and angiotensin in rats with hypertensive nephrosis . The heart-protecting musk pill can decrease the partial levels of AngII in SHR kidney, thus treating hypertensive nephropathy . In addition to studies on hypertensive nephropathy, numerous studies have demonstrated that herbal medicine can show efficacy in the treatment of chronic kidney disease or hypertension by inhibiting the RAS. For example, Chrysanthemum acts as an antihypertensive by acting on the RAS . The water extracts of kidney bean sprouts have been demonstrated to inhibit angiotensin converting enzyme, thus exhibiting potential for lowering blood pressure . Alisol B 23 acetate, as one of the main ingredients of Rhizoma alismatis , can suppress the expression of constituents of the RAS, and can inhibit the epithelial-to-mesenchymal transition (EMT) in nephrectomised rats, thus lowering blood pressure, decreasing serum creatinine and preventing proteinuria . In addition, Alisol B 23 acetate can block the RAS/Wnt/β-catenin axis to improve podocyte injury and the EMT of HK-2 cells . Ergone, one of the main ingredients of Polyporus umbellatus , and pachymic acid B, one of the main ingredients of Poria cocos , have the same effect . In addition, poricoic acid ZA, ZF, ZG and ZH, which are important components of Poria cocos , inhibit the effect of the RAS to protect podocytes and renal tubular epithelial cells, but affect the RAS and the transforming growth factor-β1 (TGF-β1)/Smad axis . Previous studies have shown that poricoic acid ZC, ZD and ZE in Poria Cocos protect renal interstitial fibrosis due to unilateral ureteral obstruction in mice via TGFβ/Smad pathway . It has been reported that 25-O-methylalisol F, the main component of Alisma , protects EMT of rat renal proximal tubular epithelial cell lines through this pathway . The therapeutic effect of Radix Scrophulariae on SHR can be attributed to the suppression of the RAS through the inhibition of the extracellular regulated protein kinase 1/2, c-Jun N-terminal kinase and p38 MAPK pathways . Xin-Ji-Er-Kang can inhibit oxidative stress by affecting the RAS, and can improve renal injury after myocardial infarction in rats . In addition, all herbal medicines that contain flavonoids, terpenoids, saponins and alkaloids are able to inhibit the RAS . Among them, common herbs containing flavonoids are Scutellaria baicalensis, Flos cmysanthemi, Sambucus adnata wall, bud of Chinese Scholar tree, Equisetum spp, Chrysanthemum indicum L., Chamaecyparis obtusa, Orthosiphon stamineus and Tropaeolum Majus L. . Common herbs containing terpenoids are the surface layer and sclerotium of Poria cocos, Alismatis rhizome and Polyporus umbellatus . A common herb containing saponins is the ginseng root . Common herbs containing alkaloids are Gambirplant, leonurus, Ophora flavescens, S. subprostrata, S. alopecuroides and Uncaria rhynchophylla . Inhibition of sympathetic excitation In patients with hypertension, the sympathetic adrenaline system is hyperactive from the central to the arterial walls. The synthesis and release of the neurotransmitter catecholamine increases, thus leading to renal arteriole contraction, and renal vascular resistance increases, thus affecting vascular remodeling . In addition, the catecholamine released by sympathetic nerves can directly act on proximal renal tubules, and increase sodium reabsorption, blood volume and blood pressure . Although it has not been demonstrated yet that the mechanism of herbal medicine in the treatment of hypertensive nephropathy involves the regulation of the sympathetic nervous system, numerous herbal medicines have been reported to be able to play a role in the inhibition of sympathetic nervous system in previous basic experiments and clinical trials. Chrysanthemum plays a role in decreasing blood pressure by inhibiting the sympathetic nerve . Radix scrophulariae , by inhibiting sympathetic excitement, suppresses SHR, and ventricular remodeling occurs . Astragaloside IV can decrease norepinephrine levels in the blood of high-fat diet-induced obese rats and in kidney tissues, which indirectly demonstrates that Astragaloside IV has the effect of inhibiting sympathetic nerves . Guizhi decoction can inhibit the cholinergic transdifferentiation of sympathetic nerves, and improve the anatomical and functional denervation of sympathetic nerves . In addition, acupuncture, electroacupuncture and moxibustion can also regulate the sympathetic nervous system, although their mechanism of action is complex . Antioxidant stress and anti-inflammatory responses Oxidative stress is caused by the imbalance of reactive oxygen species (ROS) and the antioxidant mechanism in the body. In hypertensive nephropathy, inflammatory damage is caused by the interaction of various cells such as macrophages and T lymphocytes, or inflammatory mediators or chemokines . These inflammatory cells secrete cytokines that can lead to endothelial dysfunction, which can aggravate and even lead to hypertension. Inflammatory reactions and oxidative stress play a common role and cause each other in hypertension-associated renal damage . Therefore, anti-oxidative stress and anti-inflammation can play a role in alleviating hypertensive nephropathy. As the most widely used herb in cardiovascular diseases, Salvia miltiorrhiza can significantly improve SHR blood pressure, decrease ROS production and improve vascular remodeling . The flower of Coreopsis tinctoria Nutt. is widely used in the treatment of hypertension, diabetes, obesity and other diseases. It exerts anti-inflammatory effects through its antioxidant stress properties and its ability to inhibit tumor necrosis factor-α (TNF-α), interleukin-6 (IL-6) and nuclear factor-κB (NF-κB) . Brazilian red propolis could alleviate hypertension and kidney injury in 5/6 renal ablation model rats through antioxidant stress . As an extract of Apocynum venetum , its polyphenols can improve the renal index of D-galactose-induced oxidative stress in mouse models . Tulbaghia violacea can improve NF-κB and TGF-β expression in Dahl salt-sensitive rat kidneys, and plays a role in lowering blood pressure and protecting renal function . Resveratrol, the main component of Veratrum nigrum L., has been demonstrated to have anti-ROS effects, and has the potential to lower blood pressure . As the main component of celery seeds, 3-n-butylphthalide plays a protective role in renal tubules through decreased stress, as well as the expression of pro-inflammatory cytokines and TGF-β1 in kidney tissues . Resveratrol, the main component of Veratrum nigrum L., has been demonstrated to have anti-ROS effects, and has the potential to lower blood pressure . Paeonol can effectively improve the blood pressure of spontaneously hypertensive rats, and its mechanism may be associated with reduction of blood viscosity, antioxidant stress and improvement of antioxidant capacity . Galangin, as the main extract of Alpinia officinarum hance , inhibits ROS as well as the mRNA expression of prostaglandin-endoperoxide synthase 2, TNF-α, IL-1β and IL-18, thus exerting a protective effect in rat renal epithelial cells . Icariin, as the main component of Epimedium brevicornu , can decrease the production of ROS by inhibiting the activity of NADPH oxidase, thus reducing the vasoconstriction effect of AngII-induced hypertension in rats . Natural antioxidants derived from food and herbal extracts such as tea polyphenols, curcumin and lycopene, have been widely used as complementary therapies to slow the progression of ESRD. Jiang Ya Yi Shen granules exert their protective role by inhibiting NF-κB signaling-mediated micro-inflammatory cytokines, including IL-6, TNF-α and intercellular cell adhesion molecule-1 (ICAM-1), on SHR nephropathy . Tongxinluo can inhibit the effects of oxidative stress and improve SHR glomerular sclerosis . Ban Xia Bai Zhu Tian Ma decoction could inhibit IL-1, IL-6, TNF-α and inducible nitric oxide synthase (NOS) in SHR to improve the heart damage caused by hypertension . Qian Yang Yu Yin granules can also inhibit the influence of AngII on the NADPH oxidase 4-dependent pathway, thereby inhibiting the proliferation of human mesangial cells, and lowering the production of ROS and anti-inflammatory response . The heart-protecting musk pill, also called Shexiang Baoxin pill, can decrease TGF-β and ICAM-1, thus exerting an anti-inflammatory effect, and can be used to treat SHR nephropathy and to improve vascular remodeling . Xin-Ji-Er-Kang-induced NOS in high-salt induced hypertensive mice can improve the activity and oxidative stress, and alleviate vascular remodeling . Qingxuan Jiangya decoction can affect the TGF-β1/Smad signaling pathway to play a crucial role in improving renal interstitial fibrosis in SHR . In a previous study, Shenkang improved renal injury in mice with unilateral ureteral occlusion by acting on the TGF-β/Smad3, Sirtuin/forkhead box protein O and B-cell lymphoma-2-associated X protein pathways . Regulation of vasoactive substances and other mechanisms of endothelial cell protection Hypertension can promote the synthesis of endothelial cells and the secretion of a variety of vasoactive substances. These substances maintain vascular tension and permeability, but can lead to vascular smooth muscle hypertrophy and hyperplasia. The sustained effect of blood pressure on vascular endothelial cells will result in endothelial cell damage . In addition, the increase in endogenous plasma NO synthase inhibitors in patients affects the decrease in NO synthesis by endothelial cells . The increase in ET-1, which can lead to vasoconstriction, eventually leads to enhanced vasoconstriction response and increased renal vascular resistance, and promotes the occurrence of vascular remodeling . Therefore, the best indicator of endothelial cell function is observation of the dynamic changes in vasoactive factors such as NO and ET-1. Cirsium japonicum improves the cardiac effects of renal hypertension in 2-kidney 1-clip rats by increasing serum NO levels . Morinda citrifolia can significantly decrease blood pressure and 24-h urinary NO metabolite in SHR, and its juice extract can increase the phosphorylation of endothelial NOS in human umbilical vein endothelial cells, and promote the endothelial vasodilation of the aortic ring and NO products in rats . Zingiber officinale var. rubrum exerts a significant vascular relaxation effect in SHR. Its possible mechanism of vasodilatation includes the release of NO or transmembrane calcium channels . Curcumin can protect the renal kidney function of cadmium-induced renal damage in rats, and can play a protective role on renal injury caused by hyperuricemia or high-fructose intake, and one of the mechanisms is to increase the production of NO . Morin (also known as 3,5,7,2′,4′-pentahydroxyflavone) is widely present in fruits and vegetables such as almond, old fustic, Indian guava and Osage orange. This compound may play a strong role in vascular widening by NO, muscarinic receptors, β2-adrenegic receptors and calcium channels . Hydroxysafflor yellow A, the principal component of Carthamus tinctorius L., induces angiogenesis in rat mesenteric arteries by transient receptor potential vanilloid 4 (TRPV4) -dependent calcium influx in endothelial cells . A large number of clinical experiments showed that sodium tanshinone IIA sulfonate combined with angiotensin receptor blockers (ARBs) had a stronger effect on improving renal function in patients with primary hypertensive nephropathy compared with ARB monotherapy . Sodium tanshinone IIA sulfonate, the main ingredient in the herb Salvia miltiorrhiza , has been shown to protect vascular endothelial cells. In addition, the combination of caffeic acid and ferulic acid can dilate blood vessels and resist ET-1, while exerting a hypotensive effect through ester bonds and telmisartan . Qingxuan Jiangya decoction can prevent hypertension and improve vascular remodeling in SHR by lowering the serum ET-1 level and inhibiting the TGF-β1/Smad pathway . 17-Methoxyl-7-hydroxy-furanchalcone, as an active ingredient of Fordia cauliflora , was capable of improving cardiac reconstruction from hypertension in rats by regulating the eNOS-NO signaling pathway . The combination of Astragalus membranaceus and Salvia miltiorrhiza can improve IL-1β levels in SHR urine and eNOS levels in AngII-damaged human renal glomerular endothelial cells superfluid . San Cao decoction in network pharmacologic analysis may play a role in lowering blood pressure by regulating the PI3K-Akt-eNOS pathway . Improvement of obesity-associated factors Metabolic disorder is also an important cause of hypertensive nephropathy . Obesity plays a greater role than blood pressure in the progression of hypertensive kidney disease . Obesity itself is a risk factor for high blood pressure. And in obese patients, renal dysfunction and associated increased sodium reabsorption in renal tubules can lead to hypertension . The compression of perirenal fat on the kidneys results in the activation of RAS . Chronic obesity may gradually amplify hypertension, leading to resistance to antihypertensive treatment. . Insulin resistance leads to the constriction of the extruded arterioles, thus leading to high glomerular pressure, hyperperfusion and hyperfiltration . These studies have demonstrated that obesity is closely associated with the incidence of hypertensive nephropathy. Herbal medicine has unique advantages in improving obesity. Astragaloside IV, as one of the main ingredients of Astragalus , is used to treat hypertension in high-fat diet-induced obese rats due to its anti-inflammatory effect and its ability to improve leptin resistance . Citrus paradisi and Ocimum sanctum infusions can decrease blood pressure and protect kidney function in obese rats . A number of studies have shown that Chinese herbs can improve the effects of obesity on the kidneys of patients. Coptidis rhizoma can lower the blood lipid level and renal weight of fat-prone rats, and can improve urinary protein creatinine ratio and creatinine clearance rate in rats . The mechanism may be associated with the inhibition of the NLRP3 inflammasome . Through treatment of obesity-associated glomerulopathy in model rats with Tribulus terrestris L., it was found that the herb could decrease the body weight, blood pressure, serum cystatin C levels and migration of rats, as well as improve human endothelial cells migration, thus protecting renal function . Curcumin, as one of the most important components of turmeric, can improve body weight, abdominal fat index, urinary protein excretion and average glomerular diameter in mice, and can protect podocytes from leptin damage by blocking the Wnt/β-catenin pathway . At the formula level, Mai Tong Fang inhibits fat generation and triglyceride accumulation in 3T3-L1 adipocytes .
The RAS plays an important role locally in the kidney. Angiotensin II (AngII) can directly bind to angiotensin receptors on renal arteriolar smooth muscle cells and stimulate vascular smooth muscle contraction . AngII also stimulates the sympathetic nerve to promote vascular smooth muscle resistance, thus leading to increased renal vascular resistance. In addition, AngII can increase sodium reabsorption through the aldosterone action on distal renal tubules, thus increasing the blood volume and leading to increased blood pressure . Numerous basic studies have shown that herbal medicine can play a crucial role in the treatment of hypertensive nephropathy by inhibiting RAS. Genipin, as one of the main components of Gardenia , can protect the renal function of spontaneously hypertensive rat (SHR) via the AngII-TLR/MyD88/mitogen activated protein kinase (MAPK) pathway . Qian Yang Yu Yin granules can suppress AngII in multiple manners. The mechanism includes alleviation of SHR and inhibition of 293T cells' effort induced by AngII through the epigenetic pathway associated with nicotinamide N-methyltransferase expression . The Jiangya Tongluo formula can regulate the protective effect of adrenomedullin and angiotensin in rats with hypertensive nephrosis . The heart-protecting musk pill can decrease the partial levels of AngII in SHR kidney, thus treating hypertensive nephropathy . In addition to studies on hypertensive nephropathy, numerous studies have demonstrated that herbal medicine can show efficacy in the treatment of chronic kidney disease or hypertension by inhibiting the RAS. For example, Chrysanthemum acts as an antihypertensive by acting on the RAS . The water extracts of kidney bean sprouts have been demonstrated to inhibit angiotensin converting enzyme, thus exhibiting potential for lowering blood pressure . Alisol B 23 acetate, as one of the main ingredients of Rhizoma alismatis , can suppress the expression of constituents of the RAS, and can inhibit the epithelial-to-mesenchymal transition (EMT) in nephrectomised rats, thus lowering blood pressure, decreasing serum creatinine and preventing proteinuria . In addition, Alisol B 23 acetate can block the RAS/Wnt/β-catenin axis to improve podocyte injury and the EMT of HK-2 cells . Ergone, one of the main ingredients of Polyporus umbellatus , and pachymic acid B, one of the main ingredients of Poria cocos , have the same effect . In addition, poricoic acid ZA, ZF, ZG and ZH, which are important components of Poria cocos , inhibit the effect of the RAS to protect podocytes and renal tubular epithelial cells, but affect the RAS and the transforming growth factor-β1 (TGF-β1)/Smad axis . Previous studies have shown that poricoic acid ZC, ZD and ZE in Poria Cocos protect renal interstitial fibrosis due to unilateral ureteral obstruction in mice via TGFβ/Smad pathway . It has been reported that 25-O-methylalisol F, the main component of Alisma , protects EMT of rat renal proximal tubular epithelial cell lines through this pathway . The therapeutic effect of Radix Scrophulariae on SHR can be attributed to the suppression of the RAS through the inhibition of the extracellular regulated protein kinase 1/2, c-Jun N-terminal kinase and p38 MAPK pathways . Xin-Ji-Er-Kang can inhibit oxidative stress by affecting the RAS, and can improve renal injury after myocardial infarction in rats . In addition, all herbal medicines that contain flavonoids, terpenoids, saponins and alkaloids are able to inhibit the RAS . Among them, common herbs containing flavonoids are Scutellaria baicalensis, Flos cmysanthemi, Sambucus adnata wall, bud of Chinese Scholar tree, Equisetum spp, Chrysanthemum indicum L., Chamaecyparis obtusa, Orthosiphon stamineus and Tropaeolum Majus L. . Common herbs containing terpenoids are the surface layer and sclerotium of Poria cocos, Alismatis rhizome and Polyporus umbellatus . A common herb containing saponins is the ginseng root . Common herbs containing alkaloids are Gambirplant, leonurus, Ophora flavescens, S. subprostrata, S. alopecuroides and Uncaria rhynchophylla .
In patients with hypertension, the sympathetic adrenaline system is hyperactive from the central to the arterial walls. The synthesis and release of the neurotransmitter catecholamine increases, thus leading to renal arteriole contraction, and renal vascular resistance increases, thus affecting vascular remodeling . In addition, the catecholamine released by sympathetic nerves can directly act on proximal renal tubules, and increase sodium reabsorption, blood volume and blood pressure . Although it has not been demonstrated yet that the mechanism of herbal medicine in the treatment of hypertensive nephropathy involves the regulation of the sympathetic nervous system, numerous herbal medicines have been reported to be able to play a role in the inhibition of sympathetic nervous system in previous basic experiments and clinical trials. Chrysanthemum plays a role in decreasing blood pressure by inhibiting the sympathetic nerve . Radix scrophulariae , by inhibiting sympathetic excitement, suppresses SHR, and ventricular remodeling occurs . Astragaloside IV can decrease norepinephrine levels in the blood of high-fat diet-induced obese rats and in kidney tissues, which indirectly demonstrates that Astragaloside IV has the effect of inhibiting sympathetic nerves . Guizhi decoction can inhibit the cholinergic transdifferentiation of sympathetic nerves, and improve the anatomical and functional denervation of sympathetic nerves . In addition, acupuncture, electroacupuncture and moxibustion can also regulate the sympathetic nervous system, although their mechanism of action is complex .
Oxidative stress is caused by the imbalance of reactive oxygen species (ROS) and the antioxidant mechanism in the body. In hypertensive nephropathy, inflammatory damage is caused by the interaction of various cells such as macrophages and T lymphocytes, or inflammatory mediators or chemokines . These inflammatory cells secrete cytokines that can lead to endothelial dysfunction, which can aggravate and even lead to hypertension. Inflammatory reactions and oxidative stress play a common role and cause each other in hypertension-associated renal damage . Therefore, anti-oxidative stress and anti-inflammation can play a role in alleviating hypertensive nephropathy. As the most widely used herb in cardiovascular diseases, Salvia miltiorrhiza can significantly improve SHR blood pressure, decrease ROS production and improve vascular remodeling . The flower of Coreopsis tinctoria Nutt. is widely used in the treatment of hypertension, diabetes, obesity and other diseases. It exerts anti-inflammatory effects through its antioxidant stress properties and its ability to inhibit tumor necrosis factor-α (TNF-α), interleukin-6 (IL-6) and nuclear factor-κB (NF-κB) . Brazilian red propolis could alleviate hypertension and kidney injury in 5/6 renal ablation model rats through antioxidant stress . As an extract of Apocynum venetum , its polyphenols can improve the renal index of D-galactose-induced oxidative stress in mouse models . Tulbaghia violacea can improve NF-κB and TGF-β expression in Dahl salt-sensitive rat kidneys, and plays a role in lowering blood pressure and protecting renal function . Resveratrol, the main component of Veratrum nigrum L., has been demonstrated to have anti-ROS effects, and has the potential to lower blood pressure . As the main component of celery seeds, 3-n-butylphthalide plays a protective role in renal tubules through decreased stress, as well as the expression of pro-inflammatory cytokines and TGF-β1 in kidney tissues . Resveratrol, the main component of Veratrum nigrum L., has been demonstrated to have anti-ROS effects, and has the potential to lower blood pressure . Paeonol can effectively improve the blood pressure of spontaneously hypertensive rats, and its mechanism may be associated with reduction of blood viscosity, antioxidant stress and improvement of antioxidant capacity . Galangin, as the main extract of Alpinia officinarum hance , inhibits ROS as well as the mRNA expression of prostaglandin-endoperoxide synthase 2, TNF-α, IL-1β and IL-18, thus exerting a protective effect in rat renal epithelial cells . Icariin, as the main component of Epimedium brevicornu , can decrease the production of ROS by inhibiting the activity of NADPH oxidase, thus reducing the vasoconstriction effect of AngII-induced hypertension in rats . Natural antioxidants derived from food and herbal extracts such as tea polyphenols, curcumin and lycopene, have been widely used as complementary therapies to slow the progression of ESRD. Jiang Ya Yi Shen granules exert their protective role by inhibiting NF-κB signaling-mediated micro-inflammatory cytokines, including IL-6, TNF-α and intercellular cell adhesion molecule-1 (ICAM-1), on SHR nephropathy . Tongxinluo can inhibit the effects of oxidative stress and improve SHR glomerular sclerosis . Ban Xia Bai Zhu Tian Ma decoction could inhibit IL-1, IL-6, TNF-α and inducible nitric oxide synthase (NOS) in SHR to improve the heart damage caused by hypertension . Qian Yang Yu Yin granules can also inhibit the influence of AngII on the NADPH oxidase 4-dependent pathway, thereby inhibiting the proliferation of human mesangial cells, and lowering the production of ROS and anti-inflammatory response . The heart-protecting musk pill, also called Shexiang Baoxin pill, can decrease TGF-β and ICAM-1, thus exerting an anti-inflammatory effect, and can be used to treat SHR nephropathy and to improve vascular remodeling . Xin-Ji-Er-Kang-induced NOS in high-salt induced hypertensive mice can improve the activity and oxidative stress, and alleviate vascular remodeling . Qingxuan Jiangya decoction can affect the TGF-β1/Smad signaling pathway to play a crucial role in improving renal interstitial fibrosis in SHR . In a previous study, Shenkang improved renal injury in mice with unilateral ureteral occlusion by acting on the TGF-β/Smad3, Sirtuin/forkhead box protein O and B-cell lymphoma-2-associated X protein pathways .
Hypertension can promote the synthesis of endothelial cells and the secretion of a variety of vasoactive substances. These substances maintain vascular tension and permeability, but can lead to vascular smooth muscle hypertrophy and hyperplasia. The sustained effect of blood pressure on vascular endothelial cells will result in endothelial cell damage . In addition, the increase in endogenous plasma NO synthase inhibitors in patients affects the decrease in NO synthesis by endothelial cells . The increase in ET-1, which can lead to vasoconstriction, eventually leads to enhanced vasoconstriction response and increased renal vascular resistance, and promotes the occurrence of vascular remodeling . Therefore, the best indicator of endothelial cell function is observation of the dynamic changes in vasoactive factors such as NO and ET-1. Cirsium japonicum improves the cardiac effects of renal hypertension in 2-kidney 1-clip rats by increasing serum NO levels . Morinda citrifolia can significantly decrease blood pressure and 24-h urinary NO metabolite in SHR, and its juice extract can increase the phosphorylation of endothelial NOS in human umbilical vein endothelial cells, and promote the endothelial vasodilation of the aortic ring and NO products in rats . Zingiber officinale var. rubrum exerts a significant vascular relaxation effect in SHR. Its possible mechanism of vasodilatation includes the release of NO or transmembrane calcium channels . Curcumin can protect the renal kidney function of cadmium-induced renal damage in rats, and can play a protective role on renal injury caused by hyperuricemia or high-fructose intake, and one of the mechanisms is to increase the production of NO . Morin (also known as 3,5,7,2′,4′-pentahydroxyflavone) is widely present in fruits and vegetables such as almond, old fustic, Indian guava and Osage orange. This compound may play a strong role in vascular widening by NO, muscarinic receptors, β2-adrenegic receptors and calcium channels . Hydroxysafflor yellow A, the principal component of Carthamus tinctorius L., induces angiogenesis in rat mesenteric arteries by transient receptor potential vanilloid 4 (TRPV4) -dependent calcium influx in endothelial cells . A large number of clinical experiments showed that sodium tanshinone IIA sulfonate combined with angiotensin receptor blockers (ARBs) had a stronger effect on improving renal function in patients with primary hypertensive nephropathy compared with ARB monotherapy . Sodium tanshinone IIA sulfonate, the main ingredient in the herb Salvia miltiorrhiza , has been shown to protect vascular endothelial cells. In addition, the combination of caffeic acid and ferulic acid can dilate blood vessels and resist ET-1, while exerting a hypotensive effect through ester bonds and telmisartan . Qingxuan Jiangya decoction can prevent hypertension and improve vascular remodeling in SHR by lowering the serum ET-1 level and inhibiting the TGF-β1/Smad pathway . 17-Methoxyl-7-hydroxy-furanchalcone, as an active ingredient of Fordia cauliflora , was capable of improving cardiac reconstruction from hypertension in rats by regulating the eNOS-NO signaling pathway . The combination of Astragalus membranaceus and Salvia miltiorrhiza can improve IL-1β levels in SHR urine and eNOS levels in AngII-damaged human renal glomerular endothelial cells superfluid . San Cao decoction in network pharmacologic analysis may play a role in lowering blood pressure by regulating the PI3K-Akt-eNOS pathway .
Metabolic disorder is also an important cause of hypertensive nephropathy . Obesity plays a greater role than blood pressure in the progression of hypertensive kidney disease . Obesity itself is a risk factor for high blood pressure. And in obese patients, renal dysfunction and associated increased sodium reabsorption in renal tubules can lead to hypertension . The compression of perirenal fat on the kidneys results in the activation of RAS . Chronic obesity may gradually amplify hypertension, leading to resistance to antihypertensive treatment. . Insulin resistance leads to the constriction of the extruded arterioles, thus leading to high glomerular pressure, hyperperfusion and hyperfiltration . These studies have demonstrated that obesity is closely associated with the incidence of hypertensive nephropathy. Herbal medicine has unique advantages in improving obesity. Astragaloside IV, as one of the main ingredients of Astragalus , is used to treat hypertension in high-fat diet-induced obese rats due to its anti-inflammatory effect and its ability to improve leptin resistance . Citrus paradisi and Ocimum sanctum infusions can decrease blood pressure and protect kidney function in obese rats . A number of studies have shown that Chinese herbs can improve the effects of obesity on the kidneys of patients. Coptidis rhizoma can lower the blood lipid level and renal weight of fat-prone rats, and can improve urinary protein creatinine ratio and creatinine clearance rate in rats . The mechanism may be associated with the inhibition of the NLRP3 inflammasome . Through treatment of obesity-associated glomerulopathy in model rats with Tribulus terrestris L., it was found that the herb could decrease the body weight, blood pressure, serum cystatin C levels and migration of rats, as well as improve human endothelial cells migration, thus protecting renal function . Curcumin, as one of the most important components of turmeric, can improve body weight, abdominal fat index, urinary protein excretion and average glomerular diameter in mice, and can protect podocytes from leptin damage by blocking the Wnt/β-catenin pathway . At the formula level, Mai Tong Fang inhibits fat generation and triglyceride accumulation in 3T3-L1 adipocytes .
Model of hypertensive nephropathy in herbal medicine research In hypertensive nephropathy, herbal medicine has significant clinical efficacy in relieving proteinuria and controlling the progression of renal injury. However, the number of studies on the treatment of hypertensive nephropathy with Chinese herbal medicine is limited . As aforementioned, certain compounds can control and alleviate diseases from multiple perspectives, and the Chinese herbs that contain such compounds have been listed. Subsequently, the present review tried to analyze the similarities of these plant medicines based on the theory of traditional Chinese medicine in an attempt to reach a conclusion. However, there are only few studies on this topic. Thus, the present review can only briefly discuss the summary of the application of herbal medicine in the basic research of hypertensive nephropathy. In clinical research, the disease is often treated as one of the complications of hypertension, which has not received considerable attention. This is understandable, since the most effective way to control hypertensive nephropathy is to control blood pressure . Therefore, animal experimental models of hypertensive nephropathy are often used directly in hypertension models. Since hypertensive nephropathy is nephropathy caused by hypertension, modeling should ensure the presence of proteinuria without directly damaging the kidney. Therefore, the genetic hypertension model is the most common in such studies, while the renal hypertension model is the least desirable. Among these models, SHR was produced by inbreeding in Wistar rats with the highest blood pressure, and may progress to myocardial hypertrophy, heart failure, renal insufficiency and endothelium-dependent diastolic function impairment . Dahl salt-sensitive rats are SD rats on a high-salt diet . These rats showed myocardial hypertrophy, severe heart failure, hypertensive nephropathy, impaired endothelium-dependent diastolic function and other impairments . These animal models can reflect the pathogenesis of hypertensive nephropathy. Regarding cell models, previous studies have focused on endothelial cell injury, glomerular sclerosis and renal interstitial fibrosis . Therefore, the current common cell model involves the use of Ang II to interfere with endothelial cells and observe whether their function is abnormal, or to interfere with glomerular epithelial cells and renal tubular epithelial cells and observe whether they undergo EMT . The pathogenesis of hypertensive nephropathy is not only caused by an abnormal RAS, but is the result of multiple mechanisms. The best model would be extracting the serum of hypertensive animals or patients to incubate cells . However, no such model has been reported in the studies on herbal medicine for hypertensive nephropathy thus far. In future experimental cell research, such a model should be developed, so as to better reconstruct the patients' disease.
Discussion This review summarized three points. Firstly, the pathogenesis of hypertensive nephropathy was summarized. Secondly, herbal medicine studies based on these mechanisms were listed. Thirdly, the shortcomings of the current basic research on hypertensive nephropathy models and areas for improvement were discussed. Hypertension nephropathy is a relatively complex mechanism of nephropathy. The basic pathogenesis of this disease includes renal hemodynamic changes and vascular remodeling, which are caused by various etiologies. At present, the treatment of hypertensive nephropathy in modern medicine is concentrated at a single site or approach, but the curative effect is not ideal. Different components of herbal medicine have obvious advantages in the treatment of hypertensive nephropathy. Previous studies on the efficacy and mechanism of herbal medicine in treating hypertensive nephropathy have suggested that herbal medicine plays an important role in improving renal perfusion, controlling vascular remodeling and delaying renal function progression. An interesting finding in these basic studies was that some herbs can act on two or three mechanisms at once. These experiments also provide evidence for the advantages of Chinese herbal medicine in the treatment of hypertensive nephropathy. However, the clinical studies of herbal medicine on patients with hypertensive nephropathy are relatively scarce. Although basic research is essential in terms of the explanation of the mechanism, it is only used to observe the changes in a certain organ or even a certain type of cell, which makes the basic research itself somewhat static and one-sided, and it cannot observe the changes of patients dynamically and comprehensively as with clinical research. Therefore, basic research can only provide clues for the direction of clinical medicine, and cannot replace clinical research. Due to the lack of clinical data in this field, the content of this review has some limitations. In order to better promote traditional Chinese medicine, identify the efficacy of these herbs and explore their potential mechanisms, more clinical studies related to Chinese herbs are required in the future, as well as more well-designed, large-sample, long-term, randomized and controlled clinical trials to verify the efficacy and safety. With the in-depth study of herbal medicine, modern medicine will not only be able to treat hypertensive nephropathy, but also can make great progress in other disciplines.
Supporting Data
|
Predictors of occult lymph node metastasis in clinical T1 lung adenocarcinoma: a retrospective dual-center study | 9b66b5bd-2346-4401-9638-42f537ec2f12 | 11871705 | Surgical Procedures, Operative[mh] | According to the latest estimates from the International Agency for Research on Cancer, lung cancer remains the most prevalent cancer globally, accounting for 12.4% of all cancer cases, and it is also the leading cause of cancer-related mortality . With the increased use of computed tomography (CT) imaging, more lung cancers are now being detected at an early stage. Consequently, the incidence of advanced-stage lung cancer has declined, while the incidence of localized-stage lung cancer is increasing each year . Surgical resection remains the primary treatment modality for early-stage lung cancer . Adenocarcinoma, the most prevalent histological subtype of non-small cell lung cancer (NSCLC), has a prognosis that is primarily determined by the stage of the tumor. However, accurate preoperative lymph node staging remains challenging, as CT and positron emission tomography/computed tomography (PET/CT) scans show limited sensitivity in detecting mediastinal lymph node metastasis . Thus, obtaining precise preoperative N staging is crucial for guiding treatment decisions and predicting outcomes in patients with lung adenocarcinoma. Several clinical factors have been linked to lymph node metastasis in lung adenocarcinoma, including age , consolidation-to-tumor ratio , depth ratio , smoking status , gender , tumor size , tumor location , and preoperative carcinoembryonic antigen (CEA) level . Additionally, Cytokeratin fragment 19 (CYFRA21-1) may help predict mediastinal lymph node metastasis in lung cancer , and carcinoma antigen 125 (CA125) has been suggested as a potential marker for occult lymph node metastasis (OLNM) in NSCLC . Pathological features, such as micropapillary and solid patterns, have been associated with occult N2 lymph node metastasis . Moreover, pleural invasion and lymphovascular invasion correlate with lobar lymph node metastasis in non-primary tumor-bearing lobes of NSCLC , radiological signs like pleural indention and the nonvascular penetration sign may predict lymph node metastasis in T1 lung adenocarcinoma . Anaplastic lymphoma kinase ( ALK ) rearrangement has also been associated with OLNM in lung adenocarcinoma and is considered a risk factor for postoperative recurrence in early-stage disease . However, few studies have integrated ALK gene expression with CT features and clinicopathological characteristics to predict OLNM in T1 lung adenocarcinoma. Furthermore, the need for lymph node dissection in early-stage lung adenocarcinoma remains controversial, particularly in patients considered to have a low risk of metastasis. Accurately identifying patients at risk of OLNM could help refine treatment strategies and improve patient outcomes. In this retrospective study, we investigated the incidence and predictors of OLNM in clinical T1 lung adenocarcinoma, aiming to provide data that may assist clinicians in risk stratification and therapeutic decision-making.
Patients This study was approved by the institutional review board of our institution, with a waiver of informed consent due to its retrospective nature. We retrospectively analyzed clinical data from patients at Guangxi Medical University Cancer Hospital and the Affiliated Hospital Youjiang Medical College for Nationalities, who underwent surgical resection combined with lymph node dissection and were pathologically diagnosed with lung adenocarcinoma from January 2016 to December 2023. The inclusion criteria were as follows: (1) The clinical N stage before the operation should be N0 and the maximum tumor diameter on the CT image should be 3 cm or less; (2) The pathology was confirmed as lung adenocarcinoma; (3) lobectomy, wedge resection, or segmentectomy combined with lymph node dissection was performed. Exclusion criteria included: (1) patients who received neoadjuvant therapy; (2) carcinoma in situ; (3) incomplete clinical data; (4) lack of systematic lymph node dissection or sampling during surgery. Clinical stage N0 was defined based on CT scans, indicating the absence of hilar or mediastinal lymph node enlargement, with lymph nodes measuring less than 1 cm in their shortest axis diameter. Systematic lymph node dissection was defined as the removal of at least three mediastinal lymph node stations, including the subcarinal station when applicable, along with all hilar and intrapulmonary lymph nodes on the ipsilateral side. Systematic lymph node sampling, by contrast, involved the targeted removal of specific lymph node stations as predetermined by the surgeon . A total of 1138 patients were included in the study. Clinical data collected included age, gender, smoking history, ALK (D5F3 Ventana) rearrangement status, preoperative levels of CYFRA21-1, CEA, and CA125, as well as the type of surgery performed. TNM staging was determined according to the 8th edition guidelines of the Inter-National Association for the Study of Lung Cancer. All surgical specimens were examined by pathologists whose observations were recorded. Each pathology report was reviewed for lymph node status, emphysema, visceral pleural invasion, perineural invasion, lymphovascular invasion, and spread through air spaces (STAS), and pathological subtypes of adenocarcinoma (lepidic, acinar, papillary, micropapillary, solid, complex glandular, cribriform). The IASLC/ATS/ERS lung adenocarcinoma classification system was applied. The percentage of each histological component was recorded in 5% increments, and tumors were classified according to the predominant pattern. The pattern was considered present if ≥ 5% of the histological pattern was present in the tumor. The CT features of the tumor were interpreted by two radiologists (with 20 and 9 years of experience in thoracic imaging diagnosis), who were blinded to the related clinicopathological data. In case of any differences, decisions were reached by consensus. The following CT features of tumors were measured and analyzed: tumor size (maximum diameter), density (solid, part solid, ground-glass opacity [GGO]), tumor lobe (right upper lobe, right middle lobe, right lower lobe, left upper lobe, left lower lobe), location (central, peripheral), lobulation, spiculation, pleural indentation, air bronchogram, vascular cluster, vacuole, and emphysema. Statistical analysis Continuous variables with a normal distribution were presented as the mean ± standard deviation. Skewed data were reported as the median with interquartile ranges (median [Q1, Q3]). Categorical variables were expressed as frequencies and percentages. To investigate the association between variables and lymph node metastasis, univariate binary logistic regression analyses were first conducted. Variables that were statistically significant in the univariate analyses were subsequently included in multivariate binary logistic regression analyses. The diagnostic performance of the model was evaluated by calculating the area under the receiver operating characteristic curve (AUC), along with the sensitivity and specificity. All data were performed with the SPSS software (version 22; IBM, USA). A two-sided P < 0.05 was considered statistically significant.
This study was approved by the institutional review board of our institution, with a waiver of informed consent due to its retrospective nature. We retrospectively analyzed clinical data from patients at Guangxi Medical University Cancer Hospital and the Affiliated Hospital Youjiang Medical College for Nationalities, who underwent surgical resection combined with lymph node dissection and were pathologically diagnosed with lung adenocarcinoma from January 2016 to December 2023. The inclusion criteria were as follows: (1) The clinical N stage before the operation should be N0 and the maximum tumor diameter on the CT image should be 3 cm or less; (2) The pathology was confirmed as lung adenocarcinoma; (3) lobectomy, wedge resection, or segmentectomy combined with lymph node dissection was performed. Exclusion criteria included: (1) patients who received neoadjuvant therapy; (2) carcinoma in situ; (3) incomplete clinical data; (4) lack of systematic lymph node dissection or sampling during surgery. Clinical stage N0 was defined based on CT scans, indicating the absence of hilar or mediastinal lymph node enlargement, with lymph nodes measuring less than 1 cm in their shortest axis diameter. Systematic lymph node dissection was defined as the removal of at least three mediastinal lymph node stations, including the subcarinal station when applicable, along with all hilar and intrapulmonary lymph nodes on the ipsilateral side. Systematic lymph node sampling, by contrast, involved the targeted removal of specific lymph node stations as predetermined by the surgeon . A total of 1138 patients were included in the study. Clinical data collected included age, gender, smoking history, ALK (D5F3 Ventana) rearrangement status, preoperative levels of CYFRA21-1, CEA, and CA125, as well as the type of surgery performed. TNM staging was determined according to the 8th edition guidelines of the Inter-National Association for the Study of Lung Cancer. All surgical specimens were examined by pathologists whose observations were recorded. Each pathology report was reviewed for lymph node status, emphysema, visceral pleural invasion, perineural invasion, lymphovascular invasion, and spread through air spaces (STAS), and pathological subtypes of adenocarcinoma (lepidic, acinar, papillary, micropapillary, solid, complex glandular, cribriform). The IASLC/ATS/ERS lung adenocarcinoma classification system was applied. The percentage of each histological component was recorded in 5% increments, and tumors were classified according to the predominant pattern. The pattern was considered present if ≥ 5% of the histological pattern was present in the tumor. The CT features of the tumor were interpreted by two radiologists (with 20 and 9 years of experience in thoracic imaging diagnosis), who were blinded to the related clinicopathological data. In case of any differences, decisions were reached by consensus. The following CT features of tumors were measured and analyzed: tumor size (maximum diameter), density (solid, part solid, ground-glass opacity [GGO]), tumor lobe (right upper lobe, right middle lobe, right lower lobe, left upper lobe, left lower lobe), location (central, peripheral), lobulation, spiculation, pleural indentation, air bronchogram, vascular cluster, vacuole, and emphysema.
Continuous variables with a normal distribution were presented as the mean ± standard deviation. Skewed data were reported as the median with interquartile ranges (median [Q1, Q3]). Categorical variables were expressed as frequencies and percentages. To investigate the association between variables and lymph node metastasis, univariate binary logistic regression analyses were first conducted. Variables that were statistically significant in the univariate analyses were subsequently included in multivariate binary logistic regression analyses. The diagnostic performance of the model was evaluated by calculating the area under the receiver operating characteristic curve (AUC), along with the sensitivity and specificity. All data were performed with the SPSS software (version 22; IBM, USA). A two-sided P < 0.05 was considered statistically significant.
In the study, we included 460 (40.4%) men and 678 (59.6%) women. The median age was 58 years (interquartile range [IQR]: 51, 65). 1136 (99.8%) patients had tumors located at the periphery. 229 (20.1%) patients were smokers. The majority of tumors were located in the right upper lobe (369 patients, 32.4%), while the fewest were in the right middle lobe (111 patients, 9.8%). OLNM was observed in 167 (14.7%) patients, with 55 (32.9%) having pathological N1 status and 112 (67.1%) having pathological N2 status. No lymph node metastasis was detected in the GGO group. The median tumor maximum diameter was 16.5 mm (IQR: 12, 22). Among all patients, 638 (56.0%) underwent lobectomy, 375 (33.0%) underwent wedge resection and 125 (11.0%) underwent segmentectomy. The detailed variables of the patients are listed in Table . Predictors of occult lymph node metastasis The results of the univariate analysis of OLNM in patients with clinical T1 lung adenocarcinoma are presented in Table . Patient age, gender, location of the tumor (central and peripheral), lung lobes, air bronchogram, complex glandular pattern, emphysema, and CYFRA21-1 did not differ significantly between the pathological N0 and pathological N1 + N2 groups. Factors significantly associated with OLNM included smoking history ((odds ratio [OR] = 1.891, p < 0.001), tumor maximum diameter (OR = 1.129, p < 0.001), density (OR = 0.116, p < 0.001), lobulation (OR = 47.179, p < 0.001), spiculation (OR = 6.762, p < 0.001), pleural indentation (OR = 4.605, p < 0.001), vascular cluster sign (OR = 2.791, p < 0.001), vacuole (OR = 2.190, p < 0.001), lepidic pattern (OR = 0.214, p < 0.001), acinar pattern (OR = 2.516, p < 0.001), papillary pattern (OR = 1.483, p = 0.020), micropapillary pattern (OR = 4.895, p < 0.001), solid pattern (OR = 6.885, p < 0.001), cribriform pattern (OR = 5.266, p = 0.002), predominant pattern (OR = 1.843, p < 0.001), visceral pleural invasion (OR = 2.835, p < 0.001), perineural invasion (OR = 9.023, p < 0.001), lymphovascular invasion (OR = 9.668, p < 0.001), STAS (OR = 7.320, p < 0.001), ALK (OR = 4.221, p < 0.001), CEA (OR = 3.443, p < 0.001), CA125 (OR = 2.930, p = 0.002). Multivariate logistic regression analysis revealed that lobulation (OR = 11.083, p < 0.001), spiculation (OR = 1.638, p = 0.045), lymphovascular invasion (OR = 2.969, p < 0.001), STAS (OR = 2.280, p < 0.001), micropapillary pattern (OR = 2.222, p = 0.001), solid pattern (OR = 2.681, p < 0.001), CEA (OR = 1.702, p = 0.029) were identified as independent predictors associated with an increased risk of OLNM. Conversely, part solid density (OR = 0.379, p = 0.004) and lepidic pattern (OR = 0.578, p = 0.031) were independent predictors associated with a decreased risk of OLNM. Specifically, the association of part solid density indicated that solid density was linked to a higher risk of OLNM. The results of multivariate logistic regression analyses of OLNM are shown in Table ; Fig. . The ability of each factor to predict OLNM on its own was not conclusive. However, integrating nine independent predictive factors (lobulation, speculation, density, lymphovascular invasion, STAS, lepidic pattern, micropapillary pattern, solid pattern, and CEA) into the multivariate logistic regression equation yielded more inclusive probability values. Then, we generated a receiver operating characteristic curve (ROC) to predict OLNM (Fig. ). The AUC was 0.916 ( p < 0.001, 95% CI 0.898–0.934), indicating a high diagnostic value. Predictors of occult N2 lymph node metastasis The results of the univariate analysis of occult N2 lymph node metastasis in patients with clinical T1 lung adenocarcinoma are shown in Table . Patient age, gender, location of the tumor (central and peripheral), lung lobes, air bronchogram, papillary pattern, complex glandular pattern, emphysema, and CYFRA21-1 did not differ significantly between the pathological N0 and pathological N2 groups. Occult N2 lymph node metastasis was significantly associated with smoking history (OR = 1.693, p = 0.020), tumor maximum diameter (OR = 1.117, p < 0.001), density (OR = 0.146, p < 0.001), lobulation (OR = 42.066, p < 0.001), spiculation (OR = 6.177, p < 0.001), pleural indentation (OR = 4.377, p < 0.001), vascular cluster sign (OR = 2.376, p = 0.003), vacuole (OR = 2.367, p < 0.001), lepidic pattern (OR = 0.185, p < 0.001), acinar pattern (OR = 3.322, p < 0.001), micropapillary pattern (OR = 4.892, p < 0.001), solid pattern (OR = 8.850, p < 0.001), cribriform pattern (OR = 5.625, p = 0.003, predominant pattern (OR = 1.887, p < 0.001), visceral pleural invasion (OR = 2.542, p < 0.001), perineural invasion (OR = 7.126, p < 0.001), lymphovascular invasion (OR = 9.111, p < 0.001), STAS(OR = 6.652, p < 0.001), ALK (OR = 6.379, p < 0.001), CEA (OR = 4.046, p < 0.001), CA125 (OR = 4.121, p < 0.001). Multivariate logistic regression analysis showed that lobulation (OR = 11.533, p < 0.001), lymphovascular invasion (OR = 2.752, p < 0.001), STAS (OR = 1.850, p = 0.027), micropapillary pattern (OR = 2.181, p = 0.004), solid pattern (OR = 3.579, p < 0.001), CEA (OR = 2.051, p = 0.009), ALK (OR = 2.748, p = 0.017) were identified as independent predictor associated with an increased risk of occult N2 lymph node metastasis. Conversely, the lepidic pattern (OR = 0.512, p = 0.027) was an independent predictor associated with a decreased risk of occult N2 lymph node metastasis (Table ; Fig. ).
The results of the univariate analysis of OLNM in patients with clinical T1 lung adenocarcinoma are presented in Table . Patient age, gender, location of the tumor (central and peripheral), lung lobes, air bronchogram, complex glandular pattern, emphysema, and CYFRA21-1 did not differ significantly between the pathological N0 and pathological N1 + N2 groups. Factors significantly associated with OLNM included smoking history ((odds ratio [OR] = 1.891, p < 0.001), tumor maximum diameter (OR = 1.129, p < 0.001), density (OR = 0.116, p < 0.001), lobulation (OR = 47.179, p < 0.001), spiculation (OR = 6.762, p < 0.001), pleural indentation (OR = 4.605, p < 0.001), vascular cluster sign (OR = 2.791, p < 0.001), vacuole (OR = 2.190, p < 0.001), lepidic pattern (OR = 0.214, p < 0.001), acinar pattern (OR = 2.516, p < 0.001), papillary pattern (OR = 1.483, p = 0.020), micropapillary pattern (OR = 4.895, p < 0.001), solid pattern (OR = 6.885, p < 0.001), cribriform pattern (OR = 5.266, p = 0.002), predominant pattern (OR = 1.843, p < 0.001), visceral pleural invasion (OR = 2.835, p < 0.001), perineural invasion (OR = 9.023, p < 0.001), lymphovascular invasion (OR = 9.668, p < 0.001), STAS (OR = 7.320, p < 0.001), ALK (OR = 4.221, p < 0.001), CEA (OR = 3.443, p < 0.001), CA125 (OR = 2.930, p = 0.002). Multivariate logistic regression analysis revealed that lobulation (OR = 11.083, p < 0.001), spiculation (OR = 1.638, p = 0.045), lymphovascular invasion (OR = 2.969, p < 0.001), STAS (OR = 2.280, p < 0.001), micropapillary pattern (OR = 2.222, p = 0.001), solid pattern (OR = 2.681, p < 0.001), CEA (OR = 1.702, p = 0.029) were identified as independent predictors associated with an increased risk of OLNM. Conversely, part solid density (OR = 0.379, p = 0.004) and lepidic pattern (OR = 0.578, p = 0.031) were independent predictors associated with a decreased risk of OLNM. Specifically, the association of part solid density indicated that solid density was linked to a higher risk of OLNM. The results of multivariate logistic regression analyses of OLNM are shown in Table ; Fig. . The ability of each factor to predict OLNM on its own was not conclusive. However, integrating nine independent predictive factors (lobulation, speculation, density, lymphovascular invasion, STAS, lepidic pattern, micropapillary pattern, solid pattern, and CEA) into the multivariate logistic regression equation yielded more inclusive probability values. Then, we generated a receiver operating characteristic curve (ROC) to predict OLNM (Fig. ). The AUC was 0.916 ( p < 0.001, 95% CI 0.898–0.934), indicating a high diagnostic value.
The results of the univariate analysis of occult N2 lymph node metastasis in patients with clinical T1 lung adenocarcinoma are shown in Table . Patient age, gender, location of the tumor (central and peripheral), lung lobes, air bronchogram, papillary pattern, complex glandular pattern, emphysema, and CYFRA21-1 did not differ significantly between the pathological N0 and pathological N2 groups. Occult N2 lymph node metastasis was significantly associated with smoking history (OR = 1.693, p = 0.020), tumor maximum diameter (OR = 1.117, p < 0.001), density (OR = 0.146, p < 0.001), lobulation (OR = 42.066, p < 0.001), spiculation (OR = 6.177, p < 0.001), pleural indentation (OR = 4.377, p < 0.001), vascular cluster sign (OR = 2.376, p = 0.003), vacuole (OR = 2.367, p < 0.001), lepidic pattern (OR = 0.185, p < 0.001), acinar pattern (OR = 3.322, p < 0.001), micropapillary pattern (OR = 4.892, p < 0.001), solid pattern (OR = 8.850, p < 0.001), cribriform pattern (OR = 5.625, p = 0.003, predominant pattern (OR = 1.887, p < 0.001), visceral pleural invasion (OR = 2.542, p < 0.001), perineural invasion (OR = 7.126, p < 0.001), lymphovascular invasion (OR = 9.111, p < 0.001), STAS(OR = 6.652, p < 0.001), ALK (OR = 6.379, p < 0.001), CEA (OR = 4.046, p < 0.001), CA125 (OR = 4.121, p < 0.001). Multivariate logistic regression analysis showed that lobulation (OR = 11.533, p < 0.001), lymphovascular invasion (OR = 2.752, p < 0.001), STAS (OR = 1.850, p = 0.027), micropapillary pattern (OR = 2.181, p = 0.004), solid pattern (OR = 3.579, p < 0.001), CEA (OR = 2.051, p = 0.009), ALK (OR = 2.748, p = 0.017) were identified as independent predictor associated with an increased risk of occult N2 lymph node metastasis. Conversely, the lepidic pattern (OR = 0.512, p = 0.027) was an independent predictor associated with a decreased risk of occult N2 lymph node metastasis (Table ; Fig. ).
Lymph node metastasis is an early event in the progression of lung cancer and is not uncommon even in tumors with a low clinical stage. According to the literature, the incidence of lymph node upstaging in patients with clinical stage I NSCLC who have undergone both radical surgery and lymph node dissection ranges from 10–35% . In this study, 167 (14.7%) patients with T1 lung adenocarcinoma were upstaged to pN1 or pN2 after anatomic resection. Despite the widespread use of conventional CT and PET/CT, accurate lymph node staging remains challenging. Therefore, identifying OLNM is critical to support treatment strategies for patients with clinical T1 lung adenocarcinoma, ensuring timely treatment and avoiding delays and overtreatment. Thus, in this retrospective study, we conducted a comprehensive analysis of the clinicopathological characteristics and CT features of all patients. First, our findings confirmed that CEA was a risk factor for lymph node metastasis, which is consistent with previous studies . This underscores the importance of CEA in assessing the likelihood of lymph node involvement in lung adenocarcinoma patients. Consequently, surgeons should exercise increased caution when performing lymph node dissection in lung adenocarcinoma patients with elevated serum CEA levels. Second, our study demonstrated that solid nodules are more prone to OLNM than part-solid nodules. Consistent with our findings, Xiao et al. reviewed 196 patients diagnosed with cT1N0M0 stage lung adenocarcinoma and identified solid nodules as a significant risk factor for lymph node metastasis. Furthermore, additional studies have corroborated these findings, indicating that a higher proportion of the solid component in part-solid adenocarcinomas is significantly associated with an increased likelihood of lymph node metastasis . Additionally, we observed that clinical T1 lung adenocarcinoma presenting as GGO on CT scans seldom exhibited OLNM, aligning with the findings of Wang et al. . Specifically, solids nodules exhibit a higher propensity for lymph node metastasis, potentially due to their histological characteristics. As the solid component of the tumor increases, the proportion of lepidic components decreases, with acinar, micropapillary, and solid structures becoming more prominent. This shift is associated with increased local invasiveness and a higher likelihood of lymphatic invasion . Third, our analysis revealed that tumor size does not have a significant effect on lymph node metastasis in lung adenocarcinoma patients, which aligns with the findings of several other studies . However, contrasting evidence exists, as some studies have reported that the incidence of lymph node metastasis increased with tumor size . Moreover, some studies have suggested that the risk of lymph node metastasis correlates more strongly with the size of the solid portion of the tumor rather than the overall tumor size . Given that our study included GGO nodules, which generally exhibit a lower metastatic potential, this could explain the discrepancy in results observed between our study and others. Regarding CT features, our study found that lobulation and spiculation were significantly associated with a higher risk of OLNM, and that lobulation was significantly associated with occult N2 lymph node metastasis. In 2023, Zhao et al. similarly identified lobulation as an independent predictor of OLNM in clinical IA-IIA lung adenocarcinoma. Likewise, Li et al. reported that spiculation served as an independent risk factor for predicting lymph node metastasis in patients with peripheral lung adenocarcinoma. In contrast, He et al. did not observe any significant association between lobulation, spiculation, and OLNM in peripheral non-small cell carcinoma. Furthermore, previous studies identified Type II pleural involvement — defined as a single linear or cord-like pleural tag, or a tumor abutting the pleura with a broad base (visible on both lung and mediastinal window images) — as an independent predictive factor for OLNM . This finding does not align with our results, which did not demonstrate an association between pleural indentation and OLNM. Such discrepancies may be attributed to differences in inclusion and exclusion criteria among studies. For instance, our study population may have included a higher proportion of early-stage lung adenocarcinoma cases, in which pleural involvement is less pronounced, thereby reducing its predictive value for OLNM. Additionally, we found no association between the presence of an air bronchogram and OLNM in lung adenocarcinoma, consistent with the findings of Yoshino et al. . Lobulated margins reflect heterogeneous growth rates within different regions of the tumor . This heterogeneity typically corresponds to enhanced cellular proliferation and increased angiogenesis, thereby promoting elevated invasiveness and metastatic potential . As a result, the tumor more readily breaches its primary boundaries, invades surrounding tissues, and disseminates via lymphatic channels to regional lymph nodes, increasing the likelihood of lymphatic spread. Pathological investigations have shown that spiculation can be attributed to thickened interlobular septa, fibrosis arising from the obstruction of peripheral vessels, or lymphatic channels filled with tumor cells . Consequently, the infiltration of tumor cells into lymphatic channels not only contributes to the formation of spiculated contours but also facilitates nodal dissemination. These insights may prove valuable in predicting OLNM in clinical T1 lung adenocarcinoma. Molecular testing has been widely adopted in clinical practice, serving not only to guide targeted drug therapy but also to aid in prognosis prediction and patient characterization. Patients with positive ALK expression tend to be diagnosed at a relatively advanced stage, often presenting with nodal and distant metastasis . However, the incidence of ALK alterations in early-stage NSCLC ranges from 2–7% . In this study, 53 (4.7%) patients with T1 lung adenocarcinoma exhibited positive ALK expression. We found a significant correlation between ALK positivity and occult N2 lymph node metastasis. Similarly, Seto et al. demonstrated that ALK rearrangements were significantly associated with a higher incidence of OLNM compared to ALK -negative adenocarcinomas . Furthermore, studies have shown that lung adenocarcinoma patients with ALK rearrangement have a poorer prognosis compared to those with wild-type ALK . Additionally, ALK rearrangements are more frequently observed in patients with poorly differentiated lung adenocarcinoma . Therefore, ALK positivity may serve as an important prognostic indicator for predicting OLNM in clinical T1 lung adenocarcinoma. Based on these findings, we hypothesize that performing radical resection in T1 lung adenocarcinomas with ALK positivity may provide a curative opportunity. In terms of pathology, lymphovascular invasion often indicates early lymph node metastasis and suggests a poor prognosis . Our study confirmed that lymphovascular invasion significantly increases the risk of OLNM. The relationship between visceral pleural infiltration and lymph node metastasis remains controversial , and it was not associated with OLNM in our study. Moreover, the literature suggested that lymph node metastasis in lung adenocarcinoma patients with tumor ≤ 3 cm varies by histological subtype . Specifically, solid and micropapillary patterns are significantly more likely to involve lymph nodes than other subtypes . Consistent with these findings, our study demonstrated that micropapillary and solid patterns are significantly associated with both OLNM and occult N2 lymph node metastasis. Furthermore, STAS has been recognized as a novel invasion mechanism, crucial for pathologists’ understanding of tumor behavior. In our study, STAS significantly increased the risk of OLNM in clinical T1 lung adenocarcinoma, consistent with the finding of Vaghjiani et al. . There is considerable evidence that STAS is associated with lower survival rates and acts as an independent prognostic factor, regardless of tumor stage . Travis et al. have suggested introducing STAS as a histological descriptor in the 9th edition of the TNM classification for lung cancer. Accurate preoperative identification of STAS by pathologists could serve as a crucial reference for selecting the optimal surgical approach and improving patient prognosis. Additionally, our results demonstrated that the lepidic pattern was an independent protective factor against OLNM, with a significantly lower risk of metastasis (OR = 0.578, 95% CI:0.351–0.950). Zhang et al. through a multicenter prospective clinical trial, found that the lepidic subtype of lung adenocarcinoma has a 94.0% accuracy rate in predicting negative lymph node status. The presence of the lepidic component could enhance prognosis prediction in patients with T1 lung adenocarcinoma . This study has several limitations that should be acknowledged. First, its retrospective design may introduce selection bias, potentially affecting the generalizability of the findings. While the inclusion of a two-centered sample enhances the representativeness of the results, the inherent constraints of retrospective studies remain a limitation. Second, the absence of PET imaging as a staging modality may have led to an overestimation of cT1N0 cases, potentially influencing the accuracy of preoperative staging. Third, the study specifically focused on patients with T1 lung adenocarcinoma and lacked long-term follow-up data, which limited the ability to evaluate the relationship between OLNM and survival outcomes. Future studies incorporating survival analysis are essential for a more comprehensive assessment of prognosis. Additionally, the relatively small number of OLNM cases in the sample reduced the statistical power of the findings. Larger-scale studies with prospective designs are necessary to validate and generalize these results, providing stronger evidence to guide clinical practice.
In conclusion, our study identified several independent predictors for OLNM and occult N2 lymph node metastasis in clinical T1 lung adenocarcinoma, including lobulation, spiculation, density, lymphovascular invasion, STAS, lepidic pattern, micropapillary pattern, solid pattern, CEA, and ALK . These predictive factors may assist clinicians in assessing the risk of OLNM in clinical T1 lung adenocarcinoma, potentially informing more targeted intervention strategies.
|
Pediatric Emergency Medicine Didactics and Simulation (PEMDAS): Pediatric Sedation Complications | 7cfe18ec-c2e5-49e4-87d1-0af6c9832b86 | 10861802 | Pediatrics[mh] | By the end of this session, participants will be able to: 1. List both indications and contraindications of emergency department (ED) procedural sedation with propofol and ketamine. 2. Know American Society of Anesthesiologists classifications, including patients who are appropriate for sedation in the ED. 3. Complete the informed verbal consent and obtain a presedation history and physical exam. 4. Demonstrate preparation for ED sedation, including obtaining vascular access and selecting and setting up an appropriately sized bag valve mask, capnography, pulse oximetry, and cardiac monitoring. 5. Know the correct dosing/redosing of sedation medications. 6. Manage complications of propofol. 7. Manage complications of ketamine. Procedural sedations in the pediatric emergency department (PED) are commonly performed using either propofol or ketamine. Sedation medications are provider dependent based on desired analgesic effect, availability of the medication in a provider's institution, and the provider's comfort level with the medication. Identifying who is safe for sedation in the emergency department (ED), appropriately setting up the room for sedation, obtaining informed verbal consent, and being familiar with the side effects of ketamine and propofol and how to manage these side effects are all important aspects of providing the best patient care in the ED. Ketamine is a dissociative anesthetic that also provides analgesia and is commonly used for procedures such as fracture reductions that require both properties. , Propofol is a short-acting, soy-based, IV, nonanalgesic sedative. It acts on neuronal lipid membranes as a GABA agonist, producing rapid and brief sedation in under 10 minutes. , Each institution may differ in the dosing of these sedatives. For children 3 years and younger, 2.0 mg/kg/dose of propofol can be used as an initial dose. For children 4 years and older, one can use 1.5 mg/kg/dose. , , Titration bolus doses range from 0.5 mg/kg to 1.0 mg/kg every 1–3 minutes as needed. , With each subsequent propofol bolus dose, the body tissues become saturated, thus delaying both the distribution of serum propofol and its clearance rate. Propofol is often administered after 0.5 mg/kg, maximum of 50.0 mg, of IV lidocaine has been administered due to the initial pain with propofol injection. Similar to propofol, there is a range of ketamine dosing that is provider dependent. IV ketamine is dosed at 1.0–1.5 mg/kg, with the initial maximum dose of 100.0 mg. Repeat doses of 0.5 mg/kg to 1.0 mg/kg 10 minutes after the initial dose can be administered. , Common side effects of ketamine include hypersalivation, hypertension, nystagmus, emesis, and emergence delirium. , Propofol-associated hypotension is also a well-documented side effect , , and is more commonly observed in patients with depleted intravascular volume. Hypotension is usually transient and self-resolving or responds to IV fluids. Although rare, ketamine-induced laryngospasms and propofol-induced hypoxia and apnea do occur. All are life-threatening and need to be acted on promptly. To manage laryngospasms, positive pressure ventilation with a jaw thrust and laryngospasm notch pressure can be performed. Ultimately a paralytic, preferably succinylcholine 1.0 mg/kg IV or 4.0 mg/kg IM, can be administered if the above interventions do not work. Corrective steps for propofol-induced apnea include stimulation, airway repositioning, supplemental oxygen, and bag valve mask ventilation. Supplemental preoxygenation is recommended for propofol sedations to improve periods of normal oxygenation during respiratory depression or apnea. , Interventions for these low-frequency but higher-acuity situations can be adequately practiced via simulation. These simulation cases were designed by pediatric emergency medicine (PEM) providers to augment the learner's foundation of sedation knowledge. Through participation in each simulation, learners demonstrated how to take a presedation history, perform a presedation exam, assess American Society of Anesthesiologists (ASA) classification, prepare the room for sedation, perform a time-out, and appropriately dose and redose two common sedation medications used in the ED. Learners were able to recognize common and life-threatening side effects of propofol and ketamine and demonstrated appropriate management of each. The simulations and debrief required around 2 hours to complete; each case ran around 30 minutes, and the debrief sessions required around 30 minutes. – were used in conjunction with . detailed three different patients—patients A, B, and C—each with differing histories, physical exams, and ASA classifications. This appendix was created for simulation case 2 to help learners triage appropriate patients for moderate sedation in the ED. A critical actions checklist modeled off emergency medicine (EM) procedural sedation standards of care at our three institutions was written for each simulation to ensure the educational objectives were met. Supplemental images of a shoulder dislocation and a forearm fracture for cases 1 and 2 were provided in . Debriefing was conducted immediately following both simulations in a safe environment using the Promoting Excellence and Reflective Learning in Simulation (PEARLS) debriefing framework. contained sedation simulation debriefing materials. We asked all participants across the three institutions to complete an evaluation for feedback. A didactic, PowerPoint-based presentation on propofol and ketamine was created to distribute before or after simulations. We implemented this simulation during PEM teaching sessions for a variety of PEM providers including scheduled weekly and monthly education days and orientation for newly starting trainees. The groups were made up of various levels of training, including fellows (PGY 4-PGY 6), residents (PGY 1-PGY 3), and advanced practice providers (APPs). Equipment/Environment Each simulation used a high-fidelity adolescent mannequin and was conducted either in a simulation lab meant to embody a PEM department room or in situ in the actual ED. Appropriate equipment, simulated medications, and personal protective equipment were provided (checklist in ). Each scenario started with a patient without IV access who required sedation for a procedure. The ED room was not set up for a sedation, and so, the provider had to adequately prepare it. Both cases were discussed and reviewed amongst the PEM physicians who ran the simulation prior to the simulations in order to help standardize the scenarios across the three institutions. Personnel These simulations were designed to accommodate up to 10 learners and targeted medical personnel working in a PED, including PEM fellows, APPs, PEM faculty, and pediatric and EM residents. When the simulations had enough learners to cover key roles, extra personnel functioned as bedside nurse, patient's parent, and proceduralist. One to two simulation instructors ran the simulation and the debrief afterward. Implementation Most learners had some familiarity with procedural sedation in the PED, although there was no expected specific knowledge requirement. The first scenario began with a 12-year-old male requiring a propofol sedation for an anterior shoulder dislocation. The team planned for reduction by orthopedic surgery under propofol sedation in the ED. Presedation history, focused presedation physical exam, and informed consent/assent were obtained by a sedationist/lead. The team ensured that the room was set up and the patient appropriately prepared for sedation. The team lead performed a time-out prior to the start of the sedation. The patient required two doses of propofol for adequate sedation. After the second dose of propofol, the patient developed apnea, which required a bag valve mask, and progressively worsening hypotension, which required a normal saline bolus or lactated ringers. The procedure was completed successfully, and the patient was discharged home. The team then moved to the next patient sedation (case 2 in ). The team was required to select from a list of patients which one was most appropriate for ED procedural sedation based on history and exam findings, coming up with an ASA level. The team selected which patient was appropriate for sedation under ketamine in the ED. Similar to the prior simulation, initial interventions included obtaining a presedation history, performing a focused presedation physical exam, obtaining informed consent/assent from the patient's legal guardian/patient, ensuring the room and patient were appropriately prepared for sedation, and performing a time-out prior to the start of the sedation. Shortly after ketamine was given, the patient developed laryngospasms and increased salivation with hypoxia to 85% and end-tidal carbon dioxide to zero. The team lead performed a jaw thrust with laryngospasm notch pressure, hyperextension of the neck, and bag valve mask ventilation with appropriate seal. The end-tidal continued to read zero, with worsening oxygenation saturations to 45%, perioral cyanosis, and a downtrend in the heart rate. The team activated a code response and administered succinylcholine while continuing airway interventions. Bag valve mask ventilations were continued until the patient was maintaining airway and oxygen saturations. The mother returned to the room, and the team lead explained the events to her after which the patient awakened and was discharged home. Debriefing Debriefing with participants took place immediately after simulations. Allotted time for debriefing of each simulation was around 40 minutes. The materials in were used to aid in discussion and were modeled after the PEARLS approach. Through PEARLS’ three broad educational strategies (fostering learner self-assessment, facilitation of a focused discussion, and information provided via directive feedback) and its blended approach, debriefing was interactive and collaborative. was provided to learners either prior to or after the simulation for further self-driven learning. Assessment Critical actions based on educational learning objectives were tracked throughout both simulations. Prompts were included in the simulations so that these critical actions would be completed. Both sedation simulations and debriefing were instructed and led by PEM providers with experience in simulation. Following sedation simulations, learners voluntarily provided feedback via a sedation simulation evaluation form . The form contained questions with answer options ranging from strongly disagree to strongly agree, as well as three open-ended questions meant to acquire qualitative feedback to improve the simulations and learners’ overall experiences. Each institution's feedback allowed us to edit the scenarios and create improved learning tools for pediatric emergency medicine providers. Each simulation used a high-fidelity adolescent mannequin and was conducted either in a simulation lab meant to embody a PEM department room or in situ in the actual ED. Appropriate equipment, simulated medications, and personal protective equipment were provided (checklist in ). Each scenario started with a patient without IV access who required sedation for a procedure. The ED room was not set up for a sedation, and so, the provider had to adequately prepare it. Both cases were discussed and reviewed amongst the PEM physicians who ran the simulation prior to the simulations in order to help standardize the scenarios across the three institutions. These simulations were designed to accommodate up to 10 learners and targeted medical personnel working in a PED, including PEM fellows, APPs, PEM faculty, and pediatric and EM residents. When the simulations had enough learners to cover key roles, extra personnel functioned as bedside nurse, patient's parent, and proceduralist. One to two simulation instructors ran the simulation and the debrief afterward. Most learners had some familiarity with procedural sedation in the PED, although there was no expected specific knowledge requirement. The first scenario began with a 12-year-old male requiring a propofol sedation for an anterior shoulder dislocation. The team planned for reduction by orthopedic surgery under propofol sedation in the ED. Presedation history, focused presedation physical exam, and informed consent/assent were obtained by a sedationist/lead. The team ensured that the room was set up and the patient appropriately prepared for sedation. The team lead performed a time-out prior to the start of the sedation. The patient required two doses of propofol for adequate sedation. After the second dose of propofol, the patient developed apnea, which required a bag valve mask, and progressively worsening hypotension, which required a normal saline bolus or lactated ringers. The procedure was completed successfully, and the patient was discharged home. The team then moved to the next patient sedation (case 2 in ). The team was required to select from a list of patients which one was most appropriate for ED procedural sedation based on history and exam findings, coming up with an ASA level. The team selected which patient was appropriate for sedation under ketamine in the ED. Similar to the prior simulation, initial interventions included obtaining a presedation history, performing a focused presedation physical exam, obtaining informed consent/assent from the patient's legal guardian/patient, ensuring the room and patient were appropriately prepared for sedation, and performing a time-out prior to the start of the sedation. Shortly after ketamine was given, the patient developed laryngospasms and increased salivation with hypoxia to 85% and end-tidal carbon dioxide to zero. The team lead performed a jaw thrust with laryngospasm notch pressure, hyperextension of the neck, and bag valve mask ventilation with appropriate seal. The end-tidal continued to read zero, with worsening oxygenation saturations to 45%, perioral cyanosis, and a downtrend in the heart rate. The team activated a code response and administered succinylcholine while continuing airway interventions. Bag valve mask ventilations were continued until the patient was maintaining airway and oxygen saturations. The mother returned to the room, and the team lead explained the events to her after which the patient awakened and was discharged home. Debriefing with participants took place immediately after simulations. Allotted time for debriefing of each simulation was around 40 minutes. The materials in were used to aid in discussion and were modeled after the PEARLS approach. Through PEARLS’ three broad educational strategies (fostering learner self-assessment, facilitation of a focused discussion, and information provided via directive feedback) and its blended approach, debriefing was interactive and collaborative. was provided to learners either prior to or after the simulation for further self-driven learning. Critical actions based on educational learning objectives were tracked throughout both simulations. Prompts were included in the simulations so that these critical actions would be completed. Both sedation simulations and debriefing were instructed and led by PEM providers with experience in simulation. Following sedation simulations, learners voluntarily provided feedback via a sedation simulation evaluation form . The form contained questions with answer options ranging from strongly disagree to strongly agree, as well as three open-ended questions meant to acquire qualitative feedback to improve the simulations and learners’ overall experiences. Each institution's feedback allowed us to edit the scenarios and create improved learning tools for pediatric emergency medicine providers. Both sedation simulations were implemented at three institutions with 58 participants, including PEM fellows and pediatric and EM residents. All facilitators running the simulations were either current PEM fellows or faculty, with some faculty having additional simulation-based medical education training or experience. The simulations were geared towards both expanding learners’ sedation knowledge base and reviewing prior knowledge. During the debrief in a safe environment, learners were able to give feedback on the cases and ask pertinent learning questions. At that time, learning objectives were reviewed, and any objectives that had required a prompt or had been missed were discussed in detail. The educational PowerPoint was used to augment any missed learning objectives. Immediately following the debrief sessions, 53 of the 58 participants (18 Phoenix Children's fellows, seven Medical College of Wisconsin fellows, 22 EM Vanderbilt residents, and six Vanderbilt fellows) filled out an evaluation that included five quantitative and three qualitative questions. The five quantitative questions were scored on a 5-point Likert scale (1 = strongly disagree, 2 = disagree, 3 = neither agree or disagree, 4 = agree, 5 = strongly agree ). Learners rated the scenario as clinically relevant ( M = 4.37) and effective at improving their comfort level in caring for critically ill patients ( M = 4.36). Learners felt the debrief provided valuable learning ( M = 4.40) and was a safe learning environment ( M = 4.50). Qualitative feedback included the following: cases “increased comfort level with propofol as we primarily use ketamine,” “great case,” and “this was very helpful.” Takeaway learning points included appropriate setup for airway compromise and hemodynamic instability when using propofol for sedation, review of dosing for medications, when to use succinylcholine, and the importance of a consent/assent discussion prior to sedation. These sedation simulation cases were written to improve providers’ knowledge of and comfort level in providing procedural sedation in the ED. Ketamine is a common medication used in ED procedural sedations due to its analgesic and dissociative properties. Propofol, although generally less commonly used for pediatric procedural sedations, is another advantageous sedation medication. Given the high volume of procedural sedations performed in the PED, providers should be comfortable with obtaining consent/assent, preparing the room for sedation, dosing/redosing of ketamine and propofol, these medications’ side effects, and the ability to manage side effects. Through the use of a high-fidelity mannequin in a realistic simulation scenario, providers can work on this skill set. This simulation was written so that instructors would have access to all potential resources needed to carry out the session in their own setting. Cues throughout are intended to help guide learners through the cases if they get stuck. A debriefing guide and an educational PowerPoint aid in postsimulation discussion and further learning. Learner feedback included the need for further discussion on how to obtain proper sedation consent and assent from the patient's guardian and the patient, respectively. Future iterations of these simulations will feature a more in-depth discussion of obtaining consent during the debrief session and inclusion of this information on the educational PowerPoint. The need for further education around propofol was evident from the learner feedback as well. These simulations can be used for new trainees at the beginning of the academic year for institutions where ketamine and propofol are commonly used and as a review for trainees at institutions where propofol is not as commonly utilized. Limitations include the inability to portray typical medication side effects, such as nystagmus and hypersalivation, with a high-fidelity mannequin. Secondly, we did not directly assess changes in comfort level and medical knowledge around pediatric procedural sedation medicine beyond the debriefing session. Given the high frequency with which PEM providers perform pediatric sedations, our hope is that providers can take the knowledge they absorb and immediately apply it clinically to maintain this skill set. The learning scenarios can be used either at the beginning of fellowship/residency to improve first-year fellow/resident comfort level with pediatric procedural sedation or throughout fellowship/residency as a means to review this material. Pediatric ketamine and propofol sedations are commonly performed in PEDs. Providers should be comfortable obtaining consent, setting up for a procedural sedation, dosing and redosing sedatives, knowing side effects of these medications, and acting on these side effects. Although life-threatening side effects are rare, providers should be able to promptly act on them. These simulations are effective ways to practice high-acuity events so that, when they occur in real time, the provider is ready. Sedation Simulation Cases.docx Sedation Simulation Patients.docx Critical Actions Checklist.docx Sedation Simulation Equipment.docx Sedation Simulation X-Ray Images.docx Sedation Simulation Debriefing Materials.docx Sedation Simulation Evaluation.docx Propofol and Ketamine.pptx All appendices are peer reviewed as integral parts of the Original Publication. |
Comparison of Screws with Different Diameters in Subperiosteal Implant Application with Finite Element Analysis | a9b16578-2352-4433-b4e0-93b44535350f | 11492873 | Dentistry[mh] | Eliminating tooth deficiencies, restoring function and aesthetics are the leading issues in dentistry practice. One of the most frequently preferred methods to overcome these problems encountered worldwide is intraosseous implant treatments - . Intraosseous implants developed by Brånemark need sufficient bone volume to be completely surrounded for proper function , . Therefore, a certain height and width of the alveolar bones are required for intraosseous implant treatment. In patients who do not have satisfactory alveolar bone volume for implant placement due to various reasons, the chosen bone area need to be prepared before implant placement. Many grafting methods have been proposed in the literature to prepare the surgical area for implant surgery - . Additionally, methods such as zygoma implants or angled implant placement techniques have been suggested in the literature to use the patient's residual bone . However, these treatment options also have their own complications and limitations . Prolongation of the treatment process due to the healing phase after grafting, creating an additional surgical area, sinus complications can be counted among these disadvantages , . In order to eliminate these restrictions and quickly restore the patient's function and aesthetics, personalized subperiosteal implants have been accepted as an increasingly used method in recent years - . The subperiosteal implant technique was first proposed in the literature in the 1940s. However, due to the limitations of production techniques, problems in bone implant compatibility and stabilization problems due to the lack of screw fixation, the use of such implants were extremely limited . Digital advances in computer-aided design and manufacturing software may reduce these problems and increase the indications of subperiosteal implants , . Moreover, as this type of implant can be applied without the need for additional surgery in Cawood IV, V and VI patients with severe bone resorption its popularity has increased in recent years . Subperiosteal implants enable elders with such extreme bone deficiencies to regain aesthetics and function more quickly , . Although its application requires high intra- and post-operative technical skills and knowledge, it is less traumatic for the patients compared to other alternatives - . Today, with the laser sintering method, which can be used clinically in the production of biomedical materials, it is possible to make personalized titanium implants that are highly compatible with the patient's bone under direction of tomography . Current subperiosteal implants can be fixed to the bone with mini screws in appropriate places of the jaws, and thus fixed prosthetic treatments can be performed in patients with severe bone deficiency without the need for extra grafting operations . Another advantage in subperiosteal implant surgery is that there is no need to wait for the osteointegration process, unlike the intraosseous implant option. The long-term success of treatments using this technique depends on the immobility of the implants in the bone and hermetic closure by surrounding soft tissues. The most important elements to keep implant stability are three-dimensional (3D) implant design to receive ideal support from the bone surfaces during lateral forces and the bone resistance provided by the help of screws. To gain maximum strength with bone screws, appropriate diameter for screws should be prefered to provide optimum stability and durability. However, there are not enough studies in the literature regarding the ideal screw diameter that can withstand occlusal forces in subperiosteal implants. We aimed to determine the most suitable mini screw diameter by evaluating the behavior of mini screws of two different diameters (1.5 and 2 mm) under occlusal force using the finite element analysis method.
In our finite element analysis study, the bone model and its quality were critical components. Between 2018 and 2021, we evaluated 49 patients seeking implant treatment at our clinic who, based on clinical and radiographic examinations, were found to have insufficient bone tissue for conventional implants. Of these, 33 patients were unsuitable for custom subperiosteal implant treatment due to factors such as uncontrolled comorbidities, bisphosphonate use, cleft lip and palate history, or smoking. All patients were over 60 years old. Despite meeting the criteria, four patients opted out of the treatment. Subperiosteal implants were ultimately administered to 12 patients. For 11 of these patients, pre-operative and post-operative CT scans were captured as volumetric binary files (VBF) and grouped into a file cluster. Using the Model to Model Distance Module in 3DSlicer, we computed a distance map between the 11 models, generating point-to-point distance tables from anatomically selected points. A principal component analysis module then determined a mean value for the group, which was used to create a template model with the Shape Variation Analyzer module. The Shape Population module visualized this template, resulting in a 3D model that informed all subsequent subperiosteal implant designs. This generated model incorporated mean values from the 11 patients' data, ensuring all anatomically relevant points were included. The maxillofacial models to be used (M1, M2) were provided by the medical design company BioTechnica (Turkey). The first step of the study was to convert these models into Stereolithography (STL) format. Later, computer-aided design (CAD) software was preferred to process the STL format. The reverse engineering module of the CAD software was used to convert the 3D models taken as point clouds into solid models. Thus, the 3D solid model necessary for the analysis of the implant geometry was obtained (Figure ). Since it is of great importance to minimize the amount of deviation between the resulting 3D model and point cloud data, deviation analysis was performed for all surfaces obtained with region definitions. As a result of the analysis, the deviation amount was determined as 0.05 mm. Finite element analysis (FEA) software was used to determine the stress distribution over the entire unit consisting of bone and implants. The 3D solid model obtained with CAD was transferred to FEA and an adaptable 3D solution network was created in the finite element model with Mesh Generation. The solution matrix was calculated as a tetrahedron mesh type and a parabolic element. The mesh size was calculated as 0.5 mm (Figure , Figure ). The mesh sizes used in the parts forming the whole, the elastic modulus and Poisson ratios of the materials used are given in Table and Table . In the continuation of the study, 250 N was applied vertically to the models. The stress distributions formed in two different models as a result of the application of force were examined. Data of maximum and minimum principal stresses (highest and lowest residual stress levels) in the bone were recorded. Additionally, stress distributions on the implants were evaluated with Von Mises data. In this way, the analyses made for intraosseous implants and the evaluations made with subperiosteal implants were handled separately. For comparability, the stress values of the implants were taken from the areas where intraosseous implants were applied.
The residual stress values formed in the bone as a result of vertical loading were measured as 19.86 MPa in the M2 model with 2.0 mm diameter. The stress in the M1 model which has screws with 1.5 mm diameter was measured as 22.15 MPa. Increasing the diameter of the connection holes drilled into the bone enabled the stresses on the bone to be absorbed in a wider area. On the other hand, thinner implants had a negative effect on residual stresses due to the displacement force of the pressure applied to the bone. In addition, this effect caused an increase in the stress on the implant (Figure ). The stress on the bone decreases as the hole diameters increase, regardless of the implant thickness. Similar loads at the connection points were distributed over wider surfaces with wider diameter holes (Figure ). Displacement value on the M2 model, which has the least residual stress on the bone, is given in Figure . When the axial and total displacement values were examined, the lowest displacement value on the implants was 0.42 mm in the M1 implant with 1.5 mm diameter and 2 mm thickness. In the M2 model configuration, a total displacement of 0.46 mm occurred in all directions (Figure ). When the von Mises stress levels in the implant were examined, the highest stresses were seen at the bolt connection interfaces in both models. It was determined that the highest stresses occurred in the M1 model with 2 mm thickness and 1.5 mm hole diameter. The stress accumulated in a high amount at thin-walled joint interfaces (Figure , Figure ).
The subperiosteal implant concept is a one-stage surgery system applied to patients with excessive bone loss due to various reasons without the need for secondary procedures such as bone horizontal and vertical augmentation. The biggest advantage of this system is that it can be produced individually to minimize compatibility problems. The first example of the subperiosteal implant concept in history can be seen in the studies of Dahl during the 1940s . This discovery was followed by the publication of the first case series by Goldberg in the United States . During these years, the size of the bone was measured by traditional methods for the construction of implants. After measurement subperiosteal implants were produced by casting in a laboratory environment. The surgical phase consisted of placing implants appropriate to the shape and size of the alveolar bone and covering these implants with the surrounding soft tissue. Post-operative stabilization was provided with the fibrous connection formed by the periosteum and the support of the neighbor regions. They were usually made of cobalt-chromium or titanium alloys, and the prosthesis was made using transmucosal abutments arising in the oral cavity. Due to the difficulties experienced in transferring the oral cavity to the model, laboratory problems in production and practice, its use could not spread to the general public so it was replaced by intraosseous implants . With the development of technology and 3D software, in the mid-1980s, Dr. Carl Deckard and Dr. Laser proposed their sintering method and it was further developed by Joe Beaman. In the 3D laser sintering method, the desired product is obtained by selectively sintering a blocky material with a laser. With this method, production can also be made using many different materials. In this way, errors caused by measurement and laboratory processes can be prevented . In addition, virtual design can be corrected in this technique in case of need, it is possible to drill screw holes on the implant for proper fixation so that the implant receives maximum resistance from the bone. Adequate clinical case studies have been conducted about subperiosteal implant surgeries but none of them focused on mechanical and morphological properties of these implants , . One of the shortcomings in this area is the effect of screw diameters on stabilization which has not been studied on yet. Similar to our approach, various studies using finite element analysis (FEA) have explored the mechanical behavior of different dental materials and implants, providing critical insights into their stability and performance - . Plates and mini screws, which were first recommended by Michelet in 1973 for osteosynthesis in the jaws, have become the gold standard for the healing of maxillofacial fractures and osteotomies over the years . The use of screws of different thicknesses has been suggested for different surgeries as it varies according to the type of the surgery or the place where it is performed. While micro screws are used in intraoral autogenous grafting, mini-screws are used in cases of fracture or orthognathic surgery. In cases of mandibular or maxillary resection, 2.7 mm thick screws are used for reconstruction . In our study, a comparison of 1.5 and 2 mm screws, which are frequently used in maxillary fractures and orthognathic surgery operations, was made. In clinical use, 1.5 mm screws are felt less by the patient than 2 mm screws as the 1.5 mm screws are fixed to the bone with a thinner-walled plate. Reconstruction screws were not included in the study as they would have caused bone traumas, especially to the very thin aperture piriform rim region. Studies conducted over the years have shown that the support provided by screws and bone is leading point in terms of stabilization and resistance to occlusal forces. In this study, the effect of occlusal forces was investigated by fixing two subperiosteal implants with the same design with 2 different screw diameters. Screws with 1.5 and 2 mm diameters were chosen to compare the effect of screw diameter on subperiosteal implants. The aperture piriformis and the zygomatic buttress regions are the areas that remain unaffected by ongoing bone resorption. Because of that feature, they were preferred as the most ideal areas where we could place the mini-screws. A force of 250 N was applied to the implants in our study. The residual and Von Mises forces on the implants under this force, displacement amounts and residual stresses on the bone were measured. The residual stresses in the M2 model, in which 2 mm thick screws were used, were found less than in the M1 model. The increase in the diameter of the holes drilled on the bone allowed the residual stress to spread over a wider area; therefore, the increase in the hole diameter caused a reduction in the residual stress value. This reduction is especially important in maintaining the integrity of the bone in thin bone areas such as the aperture priform area. When the movements in the implant after the applied chewing force were measured, the amount of movement in the M2 model was 0.04 mm more in the region in which the most movement occurs compared to the M1 model. Considering these results, it was concluded that 2 mm screws were better than 1.5 mm screws in terms of both the stress they create on the bone and the formation of the implant, but the 1.5 mm screw showed better results in terms of stabilization of the implant. Personalized subperiosteal implants have become an increasingly popular and promising topic in recent years. Publications concerning personalized subperiosteal implants are increasing in the literature, but in general, these publications focus on implant design and case follow-up , . There has been no study in the literature evaluating the effect of fixation screw diameters on the implant and residual bone. Orthognathic surgery, maxillomandibular fracture, or orthodontic anchor screws are the most common studies on screws. In a study, the diameter and length of the mini-screws used for orthodontic anchorage were examined histomorphometrically. It was found that the increase in the diameter of the mini screws had a positive effect on the stability, but the length of the mini screws did not have a significant effect on the stability . A three-dimensional finite element model of the mandible was developed to simulate and examine the biomechanical loads of osteosynthesis screws in bilateral sagittal osteotomy. A 2.0 mm diameter mini screw was capable of providing sufficient stability at the osteotomy site after ramus split osteotomy. Even screws with a diameter of 1.5 mm were capable of standing forces of up to 89.5 N, a force normally unavailable to patients after ramus split osteotomy during the early recovery period. The forces applied by the patients after bilateral ramus split osteotomy did not exceed these values. The 2 mm screw diameter was capable of withstanding maximum chewing forces, whereas the 1.5 mm screw diameter could only withstand low chewing forces. It was concluded that the use of 2.0 mm bicortical titanium screws placed following sagittal split osteotomy provided sufficient stability in the osteotomy line . Sindel et al. , in their study in 2014, examined the effect of bicortical screws of different thicknesses and numbers on stabilization after sagittal split osteotomy with the FEA method and concluded that no significant difference in stabilization between 1.5 and 2 mm screws existed . Molon et al. performed sagittal split osteotomy on in vitro models and compared 1.5 and 2 mm screws in terms of stabilization against occlusal forces, and they found no statistical difference between 1.5 and 2 mm screws . Nagasao et al. examined 20 maxilla models of different thicknesses using the finite element analysis method in their study and concluded that the highest success in stabilization of the Le fort osteotomy was obtained when the screw diameter and bone thickness were the same . This result may explain the reason why the 1.5 mm screw diameter is more stable in the maxilla anterior region. Thin implant systems provide more stable results in areas where the bone is so atrophic. For this reason, while custom-made implants are being produced, it can be planned with different screw diameters according to the bone thickness based on the tomography. Thick-diameter screws in the zygomatic buttress region and thin-diameter screws in the aperture piriformis region can be used in combination to achieve optimum results. The stabilization of the implant can be increased by providing additional supports with designs in which screws are placed in areas such as the palatal bone.
Mini screw-supported personalized subperiosteal implants are a very new topic in oral surgery literature. However, early studies show promising results in severe bone resorption. The biggest advantage of this method compared to other options is that the patient receives the prosthesis immediately and does not need additional grafting surgery. Moreover, operation time is shorter in these surgeries and the interventions are less traumatic for the patients. According to this finite element study, it was concluded that the 2 mm screw diameter is more useful than the 1.5 mm screw diameter. It was observed that as the diameter increases, the spread of the force over a wider area in the connection areas causes a reduction in the stress. However, the displacement value of the implant with a hole diameter of 2 mm was greater. As a result, the screws with 2 mm hole diameter caused less incoming and residual stress, but showed more displacement value than the 1.5 mm screws. It seems that we are still at the beginning of the use of this treatment option. This method will evolve in the light of more research, development and long-term clinical follow-up. We recommend finite element analysis as a method for clinical applications but it should also be supported by following clinical studies.
|
Population Pharmacokinetics, Pharmacogenomics, and Adverse Events of Osimertinib and its Two Active Metabolites, AZ5104 and AZ7550, in Japanese Patients with Advanced Non-small Cell Lung Cancer: a Prospective Observational Study | 829fb4d8-085c-410d-855c-2d8b59183a45 | 10030409 | Pharmacology[mh] | Osimertinib, a third-generation epidermal growth factor receptor (EGFR)-tyrosine kinase inhibitor (TKI), is a first-line treatment for patients with advanced EGFR-positive non-small cell lung cancer (NSCLC) . It has an excellent efficacy and safety profile; however, over 80% and 30% of patients administered osimertinib experience grade ≥ 2 and ≥ 3 adverse events (AEs), respectively . It was initially approved as a second or later-line treatment for patients with acquired EGFR T790M mutation, but the indication was extended to include it as a first-line treatment and adjuvant therapy after tumor resection . Therefore, long-term administration to asymptomatic early-stage patients is expected, increasing the importance of AE management for osimertinib treatment. Therapeutic drug monitoring (TDM) and the use of biomarkers can be novel strategies for AE management. The potential benefits of TDM in TKI treatments have been suggested , and osimertinib treatments may also benefit from TDM, as it has characteristics suitable for TDM: long-term administration is required ; a validated bioanalytical method is available ; high interindividual variability of exposure with coefficients of variation of 27.6–37.6% is observed ; and a correlation between the exposure to osimertinib parent compound and the occurrence of AEs exists . Osimertinib has two active metabolites, AZ5104 and AZ7550, which may be essential compounds for TDM. AZ5104 is produced by cytochrome P450 (CYP) 3A4 through demethylation of osimertinib indole N- methyl, whereas AZ7550 is produced by demethylation of the osimertinib terminal amine . Both active metabolites circulate at 10% exposure of the parent compound. However, AZ5104 has a 15-fold higher potency compared with the parent compound against wild-type EGFR. In contrast, AZ7550 has a similar potency but a longer half-life; therefore, higher accumulation is expected for it than for the parent compound . Thus, TDM of these two active metabolites may be beneficial for AE management, but the relevance of monitoring these compounds has not yet been elucidated. Some germline polymorphisms in the target EGFR , breast cancer resistance protein (BCRP/ ABCG2 ), drug transporter P-glycoprotein (MDR1/ ABCB1 ), and metabolism-related genes (cytochrome P450 oxidoreductase, POR ) are associated with the occurrence of AEs or pharmacokinetics (PK) of the first- and second-generation EGFR-TKIs (gefitinib, erlotinib, and afatinib) . These polymorphisms can act as novel biomarkers to predict osimertinib AEs, since the active metabolite of osimertinib has a high potency against wild-type EGFR , is a substrate of ABCG2 and ABCB1, and is metabolized by cytochrome P450, which requires electron transfer via POR . While there have been reports on the association between exposure to osimertinib parent compound and its efficacy/safety , no studies have explored the relevance of monitoring the two active metabolites of osimertinib for AE management. Additionally, germline polymorphisms in EGFR , ABCG2 , ABCB1 , or POR have a significant impact on the AEs of first- and second-generation EGFR-TKIs , but those of osimertinib have not been investigated. This study aimed to evaluate (1) the exposure–toxicity relationship and (2) the association of germline polymorphisms with osimertinib AEs to provide evidence for safe treatment and quality-of-life improvement for patients with NSCLC treated with long-term administration of osimertinib. Study design and patients This prospective, longitudinal observational study was designed and conducted at the Ageo Central General Hospital from February 2019 to July 2020 and Keio University Hospital from June 2020 to April 2022. The primary endpoint was the association of exposures to osimertinib, AZ5104, AZ7550, or germline polymorphisms with AE severity. Patients with EGFR -mutation-positive NSCLC aged ≥ 20 years who were orally administered osimertinib (standard dose: 80 mg tablet/day) were included in this study. The inclusion criteria did not restrict the type of EGFR mutation, disease or treatment history, or line of treatment. Patients who were mentally or physically incapable of providing informed consent were excluded from the study. The protocol of this study was reviewed and approved by the ethics committees of Ageo Central General Hospital (Approval No. 564), Keio University School of Medicine (Approval No. 20,200,098), and Keio University Faculty of Pharmacy (Approval No. 210,118–3 and 200,710–1), and written informed consent was obtained from all participants. The study was conducted with adherence to the Declaration of Helsinki. Data collection AEs were assessed during hospital stays for 3 months or at three outpatient visits, when blood was collected at 2 months or later after the initial osimertinib administration (onset of most EGFR-TKI AEs are reported to be within 2–4 months after initial administration ). The severity of AEs (Online Resource, Table D1) was scaled according to the Common Toxicity Criteria for Adverse Effects (CTCAE) version v5.0 by pharmacists and physicians. Serum samples for PK analysis and whole peripheral blood samples for genotyping were opportunistically collected (i.e., leftovers from routine laboratory blood analysis) once every 1–2 months after commencing osimertinib therapy. Patients were asked to determine the time of drug intake. The collected serum and whole peripheral blood samples were stored at − 80 °C until analysis. The serum concentrations of the osimertinib parent compound, AZ5104, and AZ7550 were analyzed as previously described . Briefly, osimertinib, AZ5104, and AZ7550 were extracted from 100 µL serum using a protein precipitation method and analyzed simultaneously using liquid chromatography–tandem mass spectrometry. Genotyping Polymorphisms in EGFR , ABCG2 , ABCB1 , and POR (Online Resource, Appendix A) were analyzed using TaqMan® probe-based assays (Applied Biosystems, Foster City, CA, USA), whereas the ABCG2 polymorphism (rs2231137) was studied using the CycleavePCR® assay (TaKaRa Bio Inc., Kusatsu, Japan). The detailed genotyping method is provided in the Online Resource (Appendix A). Exposure–toxicity and pharmacogenomics–toxicity relationship A population pharmacokinetic (PopPK) model was developed using Phoenix® NLME™ 8.3 software (Certara, Princeton, NJ, USA) to estimate the area under the serum concentration–time curve from 0 to 24 h (AUC 0–24 ) of the osimertinib parent compound, AZ5104, and AZ7550, applied as exposure measures. The detailed method for PopPK model development is provided in the Online Resource (Appendix B). The worst grade of AE that occurred in each patient scaled using CTCAE was used for the exposure–toxicity relationship analysis. The AUC 0–24 was simulated using the developed PopPK model based on individual predicted concentrations at the time closest to the occurrence of AEs. For patients who did not experience AEs, a median of three simulated AUC 0–24 from three concentration data points was applied for the analysis. Germline polymorphism and the worst grade of AE occurring in each patient were analyzed for pharmacogenomics–toxicity relationship analysis. Statistical analysis The cut-off date for data collection was April 4, 2022. Continuous variables such as AUC 0–24 and laboratory data are presented as median (interquartile range, IQR); estimates of PopPK parameters are presented as mean (standard error, SE). Exposure–toxicity and pharmacogenomics–exposure analyses (comparison of continuous variables) were performed using Mann–Whitney U test. Receiver operating characteristic (ROC) curve analysis was used to evaluate the discrimination potential of AUC 0–24 for grade ≥ 2 AEs. Significance of deviation of allele and genotype frequencies from Hardy–Weinberg equilibrium was tested, and pharmacogenomics–toxicity analysis (comparison of categorical variables) was performed using Fisher’s exact test. For the pharmacogenomics association analysis, multiple statistical analyses were performed using additive, recessive, and dominant genetic models. All p -values were two-sided, and statistical significance was set at p < 0.05. IBM® SPSS® statistics version 28.0 (SPSS, Inc., Chicago, IL, USA) was used for all the statistical analyses. This prospective, longitudinal observational study was designed and conducted at the Ageo Central General Hospital from February 2019 to July 2020 and Keio University Hospital from June 2020 to April 2022. The primary endpoint was the association of exposures to osimertinib, AZ5104, AZ7550, or germline polymorphisms with AE severity. Patients with EGFR -mutation-positive NSCLC aged ≥ 20 years who were orally administered osimertinib (standard dose: 80 mg tablet/day) were included in this study. The inclusion criteria did not restrict the type of EGFR mutation, disease or treatment history, or line of treatment. Patients who were mentally or physically incapable of providing informed consent were excluded from the study. The protocol of this study was reviewed and approved by the ethics committees of Ageo Central General Hospital (Approval No. 564), Keio University School of Medicine (Approval No. 20,200,098), and Keio University Faculty of Pharmacy (Approval No. 210,118–3 and 200,710–1), and written informed consent was obtained from all participants. The study was conducted with adherence to the Declaration of Helsinki. AEs were assessed during hospital stays for 3 months or at three outpatient visits, when blood was collected at 2 months or later after the initial osimertinib administration (onset of most EGFR-TKI AEs are reported to be within 2–4 months after initial administration ). The severity of AEs (Online Resource, Table D1) was scaled according to the Common Toxicity Criteria for Adverse Effects (CTCAE) version v5.0 by pharmacists and physicians. Serum samples for PK analysis and whole peripheral blood samples for genotyping were opportunistically collected (i.e., leftovers from routine laboratory blood analysis) once every 1–2 months after commencing osimertinib therapy. Patients were asked to determine the time of drug intake. The collected serum and whole peripheral blood samples were stored at − 80 °C until analysis. The serum concentrations of the osimertinib parent compound, AZ5104, and AZ7550 were analyzed as previously described . Briefly, osimertinib, AZ5104, and AZ7550 were extracted from 100 µL serum using a protein precipitation method and analyzed simultaneously using liquid chromatography–tandem mass spectrometry. Polymorphisms in EGFR , ABCG2 , ABCB1 , and POR (Online Resource, Appendix A) were analyzed using TaqMan® probe-based assays (Applied Biosystems, Foster City, CA, USA), whereas the ABCG2 polymorphism (rs2231137) was studied using the CycleavePCR® assay (TaKaRa Bio Inc., Kusatsu, Japan). The detailed genotyping method is provided in the Online Resource (Appendix A). A population pharmacokinetic (PopPK) model was developed using Phoenix® NLME™ 8.3 software (Certara, Princeton, NJ, USA) to estimate the area under the serum concentration–time curve from 0 to 24 h (AUC 0–24 ) of the osimertinib parent compound, AZ5104, and AZ7550, applied as exposure measures. The detailed method for PopPK model development is provided in the Online Resource (Appendix B). The worst grade of AE that occurred in each patient scaled using CTCAE was used for the exposure–toxicity relationship analysis. The AUC 0–24 was simulated using the developed PopPK model based on individual predicted concentrations at the time closest to the occurrence of AEs. For patients who did not experience AEs, a median of three simulated AUC 0–24 from three concentration data points was applied for the analysis. Germline polymorphism and the worst grade of AE occurring in each patient were analyzed for pharmacogenomics–toxicity relationship analysis. The cut-off date for data collection was April 4, 2022. Continuous variables such as AUC 0–24 and laboratory data are presented as median (interquartile range, IQR); estimates of PopPK parameters are presented as mean (standard error, SE). Exposure–toxicity and pharmacogenomics–exposure analyses (comparison of continuous variables) were performed using Mann–Whitney U test. Receiver operating characteristic (ROC) curve analysis was used to evaluate the discrimination potential of AUC 0–24 for grade ≥ 2 AEs. Significance of deviation of allele and genotype frequencies from Hardy–Weinberg equilibrium was tested, and pharmacogenomics–toxicity analysis (comparison of categorical variables) was performed using Fisher’s exact test. For the pharmacogenomics association analysis, multiple statistical analyses were performed using additive, recessive, and dominant genetic models. All p -values were two-sided, and statistical significance was set at p < 0.05. IBM® SPSS® statistics version 28.0 (SPSS, Inc., Chicago, IL, USA) was used for all the statistical analyses. Data collection Patient characteristics are summarized in Table . A total of 302 serum samples for PK analysis, median (IQR) time after administration of 6.3 (2.0–23.6) h, and 53 whole peripheral blood samples for genotyping were collected from 53 patients. The median number of serum samples per patient was five, collected within a median of 5 months. Osimertinib safety was evaluated in 51 patients; we were unable to collect enough safety data from 2 out of 53 patients because of transfer to another hospital and withdrawal of consent. The lymphocytes for one of the patients and creatine phosphokinase for 20 patients were not tested in routine clinical laboratory blood analysis. The most prevalent grade ≥ 2 AE was skin disorders (63%, Online Resource, Table D1). Approximately 25% of the patients experienced severe AEs (grade 3 or leading to dose discontinuation). The patient characteristic significantly associated with severe AEs was the EGFR-TKI treatment line, with higher frequency of severe AEs in patients receiving therapy as first-line treatment (Online Resource, Table D2, p = 0.004). Other characteristics, including sex, age, ECOG PS, and somatic EGFR mutation, were not associated with any grade ≥ 2 or ≥ 3 AEs. Development and evaluation of the PopPK model The final PopPK parameters are listed in Table . Albumin was identified as a significant covariate for the clearance of the parent compound and AZ5104, whereas body weight (BW) was identified as a significant covariate for the clearance of AZ7550 (Online Resource, Table C1–C3, p < 0.05). The PopPK parameters were estimated using models (1), (2), and (3) for osimertinib, AZ5104, and AZ7550, respectively. The estimated values were close to the mean values calculated from the bootstrap sampling of n = 974 (success rate, 97.4%), n = 1000 (success rate, 100%), and n = 1000 (success rate, 100%), respectively, and all fell within the 95% percentile confidence intervals (Table ). The detailed results for the development and evaluation of the PopPK model are provided in the Online Resource (Appendix C). Exposure–toxicity relationship The median (IQR) values of AUC 0–24 at steady state (estimated using the final PopPK model) of the osimertinib parent compound, AZ5104, and AZ7550 were 4278 (3328–5589) ng/mL*h, 414 (309–584) ng/mL*h, and 367 (275–516) ng/mL*h, respectively, regardless of dosage. There was a significant association between the AUC 0–24 of the active metabolite AZ7550 and grade ≥ 2 paronychia or grade ≥ 2 anorexia; AUC 0–24 of the osimertinib parent compound or active metabolite AZ5104 and grade ≥ 2 diarrhea; and AUC 0–24 of the parent or either of the two active metabolites and grade ≥ 2 increased creatinine. Overall, the AUC 0–24 of AZ5104 was significantly associated with any grade ≥ 2 AEs (Fig. ). The cut-off AUC 0–24 of the parent compound, AZ5104, and AZ7550 to predict grade ≥ 2 AEs were 4938 ng/mL*h, 540 ng/mL*h, and 462 ng/mL*h, respectively, based on the ROC curve. The areas under the curve (sensitivity and specificity) were 0.680 (51% and 90%), 0.705 (42% and 100%), and 0.651 (49% and 90%), respectively (Online Resource, Fig. E1). The frequency of the aforementioned AEs (grade ≥ 2) was higher in patients with ≥ 4938 ng/mL*h parent compound, ≥ 540 ng/mL*h AZ5104, or ≥ 462 ng/mL*h AZ7550 AUC 0–24 (Online Resource, Table E1–E3). Pharmacogenomics-toxicity relationship All analyzed genotypes were in Hardy–Weinberg equilibrium ( p > 0.050), except for EGFR rs4947492, EGFR rs2227983, ABCG2 rs2622604, and ABCB1 rs1045642 ( p = 0.047, p = 0.020, p = 0.049, and p = 0.028, respectively), but the allele frequency was similar to that reported by Togo Var . There was a significant relationship between polymorphisms in germline EGFR and severe AEs or anorexia. EGFR rs2293348 C > T (C/T genotype) and EGFR rs4947492 G > A (A/A genotype) were both associated with any severe AEs (Table , p = 0.019 in the additive model, and p = 0.050 in the recessive model). EGFR rs2293348 C > T (C/T genotype) was also significantly associated with grade ≥ 1 or ≥ 2 anorexia (Online Resource, Table F1, p < 0.001 or p = 0.023, respectively, in an additive model). The ABCG2 rs2231137 G > A polymorphism (G/A and A/A genotypes) was significantly associated with both AEs and PK. A higher frequency of any grade ≥ 2 AEs (Table , p = 0.008 in a dominant model) was observed in patients with ABCG2 rs2231137 G/A or A/A genotypes. Furthermore, patients with the G/A or A/A genotype had a significantly higher exposure (AUC 0–24 ) to the osimertinib parent compound (Fig. , p = 0.018). A significant association was observed between ABCB1 rs1128503 C > T polymorphism (C/T and T/T genotypes) and any grade ≥ 2 AEs (Table , p = 0.038 in a dominant model). Additionally, the ABCB1 rs2032582 G > T/A polymorphism (A/T and G/T genotypes) was associated with any grade ≥ 2 skin disorders and grade ≥ 2 lymphocyte count decrease (Online Resource, Table F2, p = 0.017; Table F3, p = 0.033; respectively, in an additive model). Polymorphisms in the POR were not significantly associated with any AEs or PK of the drug. Patient characteristics are summarized in Table . A total of 302 serum samples for PK analysis, median (IQR) time after administration of 6.3 (2.0–23.6) h, and 53 whole peripheral blood samples for genotyping were collected from 53 patients. The median number of serum samples per patient was five, collected within a median of 5 months. Osimertinib safety was evaluated in 51 patients; we were unable to collect enough safety data from 2 out of 53 patients because of transfer to another hospital and withdrawal of consent. The lymphocytes for one of the patients and creatine phosphokinase for 20 patients were not tested in routine clinical laboratory blood analysis. The most prevalent grade ≥ 2 AE was skin disorders (63%, Online Resource, Table D1). Approximately 25% of the patients experienced severe AEs (grade 3 or leading to dose discontinuation). The patient characteristic significantly associated with severe AEs was the EGFR-TKI treatment line, with higher frequency of severe AEs in patients receiving therapy as first-line treatment (Online Resource, Table D2, p = 0.004). Other characteristics, including sex, age, ECOG PS, and somatic EGFR mutation, were not associated with any grade ≥ 2 or ≥ 3 AEs. The final PopPK parameters are listed in Table . Albumin was identified as a significant covariate for the clearance of the parent compound and AZ5104, whereas body weight (BW) was identified as a significant covariate for the clearance of AZ7550 (Online Resource, Table C1–C3, p < 0.05). The PopPK parameters were estimated using models (1), (2), and (3) for osimertinib, AZ5104, and AZ7550, respectively. The estimated values were close to the mean values calculated from the bootstrap sampling of n = 974 (success rate, 97.4%), n = 1000 (success rate, 100%), and n = 1000 (success rate, 100%), respectively, and all fell within the 95% percentile confidence intervals (Table ). The detailed results for the development and evaluation of the PopPK model are provided in the Online Resource (Appendix C). The median (IQR) values of AUC 0–24 at steady state (estimated using the final PopPK model) of the osimertinib parent compound, AZ5104, and AZ7550 were 4278 (3328–5589) ng/mL*h, 414 (309–584) ng/mL*h, and 367 (275–516) ng/mL*h, respectively, regardless of dosage. There was a significant association between the AUC 0–24 of the active metabolite AZ7550 and grade ≥ 2 paronychia or grade ≥ 2 anorexia; AUC 0–24 of the osimertinib parent compound or active metabolite AZ5104 and grade ≥ 2 diarrhea; and AUC 0–24 of the parent or either of the two active metabolites and grade ≥ 2 increased creatinine. Overall, the AUC 0–24 of AZ5104 was significantly associated with any grade ≥ 2 AEs (Fig. ). The cut-off AUC 0–24 of the parent compound, AZ5104, and AZ7550 to predict grade ≥ 2 AEs were 4938 ng/mL*h, 540 ng/mL*h, and 462 ng/mL*h, respectively, based on the ROC curve. The areas under the curve (sensitivity and specificity) were 0.680 (51% and 90%), 0.705 (42% and 100%), and 0.651 (49% and 90%), respectively (Online Resource, Fig. E1). The frequency of the aforementioned AEs (grade ≥ 2) was higher in patients with ≥ 4938 ng/mL*h parent compound, ≥ 540 ng/mL*h AZ5104, or ≥ 462 ng/mL*h AZ7550 AUC 0–24 (Online Resource, Table E1–E3). All analyzed genotypes were in Hardy–Weinberg equilibrium ( p > 0.050), except for EGFR rs4947492, EGFR rs2227983, ABCG2 rs2622604, and ABCB1 rs1045642 ( p = 0.047, p = 0.020, p = 0.049, and p = 0.028, respectively), but the allele frequency was similar to that reported by Togo Var . There was a significant relationship between polymorphisms in germline EGFR and severe AEs or anorexia. EGFR rs2293348 C > T (C/T genotype) and EGFR rs4947492 G > A (A/A genotype) were both associated with any severe AEs (Table , p = 0.019 in the additive model, and p = 0.050 in the recessive model). EGFR rs2293348 C > T (C/T genotype) was also significantly associated with grade ≥ 1 or ≥ 2 anorexia (Online Resource, Table F1, p < 0.001 or p = 0.023, respectively, in an additive model). The ABCG2 rs2231137 G > A polymorphism (G/A and A/A genotypes) was significantly associated with both AEs and PK. A higher frequency of any grade ≥ 2 AEs (Table , p = 0.008 in a dominant model) was observed in patients with ABCG2 rs2231137 G/A or A/A genotypes. Furthermore, patients with the G/A or A/A genotype had a significantly higher exposure (AUC 0–24 ) to the osimertinib parent compound (Fig. , p = 0.018). A significant association was observed between ABCB1 rs1128503 C > T polymorphism (C/T and T/T genotypes) and any grade ≥ 2 AEs (Table , p = 0.038 in a dominant model). Additionally, the ABCB1 rs2032582 G > T/A polymorphism (A/T and G/T genotypes) was associated with any grade ≥ 2 skin disorders and grade ≥ 2 lymphocyte count decrease (Online Resource, Table F2, p = 0.017; Table F3, p = 0.033; respectively, in an additive model). Polymorphisms in the POR were not significantly associated with any AEs or PK of the drug. To the best of our knowledge, this is the first report to evaluate the association between exposures to the two active metabolites of osimertinib (AZ5104 and AZ7550) or germline polymorphisms ( EGFR , ABCG2 , ABCB1 , and POR ) and the severity of AEs. PopPK modeling was performed to estimate the AUC 0–24 , which was selected as an exposure measure given the suggested linear relationship between the AUC of osimertinib parent compound (at a steady state of any dosing interval) and AEs . The median estimated AUC 0–24 of osimertinib at steady state was consistent or slightly lower than that in previous reports because some patients included in our study received a reduced dose . Albumin level was identified as a significant covariate for the clearance of the parent compound and AZ5104 (positive correlation), consistent with a previous report . C-reactive protein (CRP) is a covariate of osimertinib clearance . Decrease in albumin level and an increase in CRP level occur during inflammation, which can decrease CYP3A4 activity . Therefore, decreased albumin level may be an indicator of increased exposures to the osimertinib parent compound and AZ5104—a consequence of decreased CYP3A4 activity during inflammation. However, elucidating these mechanisms is outside the scope of this study. To our knowledge, our study is the first to report a PopPK model and a significant covariate (BW) for AZ7550. The BW may have an impact on the clearance of AZ7550 because AZ7550 has a longer half-life than the parent compound and AZ5104 (72.7 h versus 59.7 and 52.6 h, respectively), and greater distribution and accumulation are expected . The exposure–toxicity relationships were different among the three compounds: osimertinib parent, AZ5104, and AZ7550. Exposure to AZ7550 was associated with grade ≥ 2 paronychia and anorexia, both of which are potentially related to direct epidermal or mucosal damage. The mechanisms underlying EGFR-TKI-induced skin disorders, such as paronychia, involve decreased keratinocyte proliferation and increased keratinocyte differentiation at the epidermis, resulting in impaired skin barrier function . However, anorexia can be caused by multiple factors, and assessing causal associations is difficult. Although the mechanism responsible for EGFR-TKI-induced anorexia has not been elucidated, one hypothesis for anorexia induction is gastric mucosal injury, since inhibition of EGFR in gastric parietal cells can interfere with gastric mucosal membrane protection and repair of mucosal injury . Moreover, cancer cachexia should be considered a confounder of the AZ7550 exposure–anorexia relationship. The main characteristics of cancer cachexia are anorexia and inflammation, and the elevated production of proinflammatory cytokines can cause a decrease in CYP3A4 activity, resulting in alterations in AZ7550 metabolism . However, because our PopPK analysis indicated that the albumin level, which may be related to inflammation, was not a significant covariate for AZ7550 clearance, we considered that inflammation was unlikely to affect the PK of AZ7550. It is more likely that the increase in exposure to AZ7550 increases direct epidermal or mucosal damage, leading to a higher severity of paronychia or anorexia. Our results also indicated that exposures to the osimertinib parent compound and AZ5104 were associated with grade ≥ 2 diarrhea and increase in creatinine level. Similarly, a previous report suggested a linear relationship between the exposure to the parent compound and the occurrence of diarrhea , whereas another report suggested that exposure to the parent compound and grade ≥ 1 diarrhea were not associated ; thus, exposure to the parent compound may be related to higher grade (grade ≥ 2) diarrhea. The potential mechanism responsible for both diarrhea and creatinine increase may be the activation of chloride secretion as a result of EGFR inhibition . The increased activation of chloride secretion by increased parent compound and AZ5104 exposures can enhance passive water movement through the gastrointestinal lumen, causing high-grade diarrhea and dehydration, which can lead to renal failure . Assessment of the causal association for the exposure–kidney failure relationship is difficult because kidney failure can influence drug excretion. On the one hand, only 1.7% of the dose is excreted in the urine as osimertinib, AZ5104, and AZ7550 (oral bioavailability of osimertinib: 69.8% ), and changes in renal clearance are unlikely to affect drug exposure . On the other hand, because the influence of renal impairment on CYP3A enzyme function has been reported , the potential impact on total clearance cannot be ignored, and further studies are warranted to confirm this influence. Overall, exposure to AZ5104 was significantly associated with any grade ≥ 2 AEs; the exposure to AZ5104 showed a higher area under the curve in ROC analysis than that to osimertinib parent compound and AZ7550. Likewise, an earlier study suggested an absence of a relationship between the trough concentration of the osimertinib parent compound and any toxicity ; thus, monitoring AZ5104 may be more beneficial than monitoring the parent compound for the management of any AEs. Additionally, the results of ROC analysis predicting any grade ≥ 2 AEs using AUC 0–24 of the parent compound, AZ5104, and AZ7550, showed low sensitivity but high specificity. Therefore, the AUC 0–24 levels may not be an absolute index for any AE prediction, but it may become a useful index to decide whether a patient experiencing AEs needs dose reduction. Since several reports suggested that there is no relationship between exposure to osimertinib and efficacy, dose reduction up to 50% of mean exposure may be considered for patients experiencing AEs and high AUC 0–24 levels; however, further validation is needed . The two intron variants in EGFR , rs2293348 C > T (C/T genotype) and rs4947492 G > A (A/A genotype), were associated with severe AEs. According to earlier studies, EGFR rs2293348 is related to erlotinib-induced rash, and rs4947492 is related to gefitinib-induced diarrhea . The role of these genetic variants has not been fully investigated; however, since the intron sequence is involved in the regulation of expression, these variants may cause alterations in EGFR expression . Thus, EGFR rs2293348 and rs4947492 may influence sensitivity to EGFR inhibition and severity of AEs. The germline polymorphisms in ABCG2 rs2231137 G > A (G/A and A/A genotypes) and ABCB1 rs1128503 C > T (C/T and T/T genotypes) were significantly associated with any grade ≥ 2 AEs, and ABCB1 rs2032582 G > T/A polymorphism (A/T and G/T genotypes) was significantly associated with any grade ≥ 2 skin disorders. Osimertinib is a substrate of ABCG2 and ABCB1; thus, polymorphisms in these genes may influence the distribution or PK of this drug . ABCG2 rs2231137 G > A (G/A and A/A genotypes) results in Val12Met substitution and is partly responsible for the functional impairment of ABCG2. The association of this variant with gefitinib-induced rash and the PK of gefitinib has been reported , which is consistent with our results for osimertinib. The two ABCB1 polymorphisms (rs1128503 and rs2032582) were associated with grade ≥ 2 AEs but not with the serum concentration of osimertinib. We hypothesized that ABCB1 polymorphisms may influence the tissue distribution and accumulation of osimertinib and its active metabolites, as an in vivo study suggested the involvement of ABCB1 and ABCG2 in tissue accumulation of other TKIs . Taken together, our results suggested that monitoring exposure to the parent compound, AZ5104, and AZ7550 and genotyping polymorphisms in EGFR, ABCG2 , and ABCB1 are potential new approaches for AE management. AEs can be effectively managed without dose interruption by adjusting the dose and increasing awareness about the need for AE management. The AEs we focused on in this study, such as paronychia, anorexia, and diarrhea, are not fatal but need to be managed to improve quality of life and encourage patients to continue osimertinib therapy for a long period. This study has several limitations to consider when interpreting the results. Because of the small sample size, we were unable to perform a multivariable analysis to compare the impact of multiple risk factors or check for potential confounding variables. A larger sample size is also required to evaluate the relationship between fatal or severe AEs (grade ≥ 3) and exposure to osimertinib or its active metabolites. Opportunistic collection of serum samples reduced the burden on patients but led to there being fewer concentration data points around the time of peak serum concentration (T max ) for PopPK analysis. Additional blood sampling is required to improve the accuracy of the volume of distribution estimate, but this was not essential for the purpose of this study. The exposure–efficacy relationship was not analyzed in this study because progression-free survival was not reached for many of the participants at the time of data cut-off. Although many reports have suggested that there is no relationship between osimertinib exposure and efficacy, further studies are needed to identify the optimal therapeutic window for osimertinib treatment. Our findings demonstrated for the first time that exposures to the two active metabolites of osimertinib (AZ5104 and AZ7550) were associated with AEs and may have a different impact than exposure to the osimertinib parent compound. Therefore, monitoring not only the parent compound but also the active metabolites is a potential approach for osimertinib AE management. Germline polymorphisms in EGFR (rs2293348 and rs4947492), ABCG2 (rs2231137), and ABCB1 (rs1128503 and rs2032582) were identified as potential biomarkers for predicting the severity of AEs and may help increase awareness of the importance of AE management for patients at higher risk. Below is the link to the electronic supplementary material. Supplementary Material 1 |
Chronic Post-Prandial Epigastric Pain Associated with Median Arcuate Ligament Syndrome and Atherosclerosis of the Celiac Trunkin An Elderly Woman: A Case Report | 99c3a9bb-12df-4339-ae33-00874045e983 | 11896898 | Surgical Procedures, Operative[mh] | The median arcuate ligament is an anatomical structure located in the upper abdomen. It is a tendinous portion of the diaphragm that arches over the aorta and forms the aortic hiatus, typically at the level of the T12 vertebra, just above the origin of the celiac trunk . In some individuals, the median arcuate ligament is positioned lower than normal and compresses the celiac trunk, leading to median arcuate ligament syndrome (MALS) . The syndrome is a relatively uncommon condition, with an estimated prevalence of approximately 2 cases per 100 000 individuals. MALS typically affects young women, with a female-to-male ratio of 2: 1 to 3: 1, and predominantly occurs in individuals within the age range of 20 to 40 years . The underlying reasons for the observed predominance in women remain uncertain; however, it may be related to a tendency for the celiac artery to originate at a more cephalad position in females than in males . The exact cause of this anatomical malformation is unknown. Patients with MALS experience a wide range of nonspecific symptoms, including post-prandial epigastric pain, nausea, vomiting, food fear, and weight loss . The diagnosis of MALS is a diagnosis of exclusion supported by imaging methods, such as computed tomography (CT) and Doppler ultra-sound . Atherosclerosis should be excluded, because the celiac artery is the most frequently affected site of mesenteric artery atherosclerosis . There is currently no consensus on the treatment of MALS. Treatment is mainly surgery to remove the median arcuate ligament (MAL), which relieves compression on the celiac trunk and plexus . Celiac artery revascularization is generally reserved for refractory cases if symptoms persist after surgery . Several cases of MALS have been reported, often presenting diagnostic challenges . However, to the best of our knowledge, there has been no documented case involving concurrent MALS and celiac artery atherosclerosis. Furthermore, an approach to managing these coexisting conditions may remain unclear. Herein, this report describes a 66-year-old woman with chronic post-prandial epigastric pain associated with atherosclerosis of the celiac trunk, managed with angioplasty and stenting, combined with MALS.
A 66-year-old female patient with a history of uncontrolled dyslipidemia presented to the gastrointestinal clinic with chronic abdominal pain for 4 years. Four years prior to this presentation, the patient reported experiencing mildly dull epigastric pain that typically occurred after meals, lasting for 1 to 2 h, accompanied by mild abdominal distension. The symptoms were initially minimal, occurring 1 to 3 times per month. The pain increased in severity over the years. Two years later, it worsened with large meals, and the symptoms gradually became more frequent and severe, occurring almost daily, leading the patient to restrict food intake and divide meals into smaller portions. The patient also sometimes experienced nausea; however, she denied weight loss, diarrhea, and fever. An esophagogastroduodenoscopy (EGD) and colonoscopy showed no significant abnormalities. The patient was empirically treated for gastritis and irritable bowel disease; however, symptoms still persisted. A year before the presentation, the patient underwent a resection of a gastric submucosal tumor, with benign pathology. The patient’s symptoms did not improve significantly after the tumor removal, prompting her to seek our hospital. At initial evaluation, the patient was in no distress, and the physical examination revealed only mild epigastric tenderness, with no bruit or other significant findings. Laboratory test results revealed no significant findings except for dyslipidemia (low-density lipoprotein cholesterol: 4.9 mmol/L). The abdominal ultrasound did not suggest any possible causes. A repeat EGD showed chronic gastritis, which did not correlate with the patient’s clinical scenario. Subsequently, an abdominal CT scan with contrast was performed to investigate other potential causes of the pain. The results revealed hook-shaped stenosis of the celiac trunk at the upper margin of the L1 vertebra and mild post-stenosis dilatation, highly suggestive of MALS . The patient exhibited a strong preference against surgical intervention, including minimally invasive laparoscopic surgery, due to concerns regarding the associated risks of surgery, despite discussions regarding the potential benefits of decompression intervention. Due to the patient’s advanced age and the presence of dyslipidemia, atherosclerosis was suspected to play a role in pathogenesis. A discussion between the patient, her family, and a team of physicians, including a gastroenterologist and a cardiovascular interventionist, was held to discuss the risks and benefits of each intervention, including surgery and angioplasty, leading to the agreement to perform endovascular angioplasty with stent implantation. Digital subtraction angiography confirmed the nearly total occlusion of the celiac trunk, with the position suggesting MALS . During the procedure with a 5×18 mm Herculink Elite (Abbott Vascular, Santa Clara, CA, USA) stent, intravascular ultrasound revealed concurrent moderately severe intra-luminal atherosclerotic stenosis of approximately 60% to 70% . Post-implantation angiography revealed a significant improvement in blood flow to the branches of the celiac trunk . After the intervention, the patient remained in stable condition but had mild urticaria. A follow-up abdominal CT showed a well-placed stent with good flow inside and a minor curve at the position of the ligament. The patient reported remarkable improvement in her epigastric pain and was discharged on day 5 of admission with a regimen that included low-dose aspirin, clopidogrel, rosuvastatin, and bisoprolol. At the 5-month follow-up, the patient was in stable condition and noted that she could consume larger portions of meals without experiencing discomfort or digestive issues. The patient will be reassessed every 3 months, and a CT scan will be performed after 6 months or if her symptoms recur.
To the best of our knowledge, this is the first case presenting chronic post-prandial epigastric pain associated with MALS and celiac artery atherosclerosis. Due to the low prevalence and nonspecific symptoms of MALS, it could be misdiagnosed. This case made physicians highly suspicious of this disease, especially in patients with post-prandial abdominal pain. Furthermore, angioplasty and stenting can be considered in selected patients with MALS and atherosclerosis of the celiac artery. Most patients with MALS are asymptomatic, and the high recurrence rate after decompression illustrates the complexity of its pathophysiology, suggesting that it involves more than just ischemia . As seen in the present case, finding another pathological process involved, such as atherosclerosis, is essential. The pathophysiological process of MALS includes ischemia to the organs supplied by the celiac trunk, with or without disturbances in the neural transmission of the celiac plexus . The compression of the celiac trunk can reduce the blood supply to the foregut and foregut-derived organs, including the distal esophagus, stomach, proximal duodenum, liver, pancreas, gallbladder, and spleen . Additionally, the pressure exerted on the nearby celiac plexus can disrupt the transmission of neural impulses from and to these organs, possibly playing a role in the pathophysiology of MALS. However, chronic compression could predispose to other secondary pathologic processes, especially atherosclerosis, which can worsen arterial constriction . In one study with 120 unselected Finnish autopsy patients, 29% of patients had stenosis of mesenteric arteries, with the celiac artery being the most common site . Furthermore, recent findings by Emre et al provided no evidence to support the hypothesis that a superiorly positioned celiac trunk contributes to MALS. The distance between the celiac trunk and the superior mesenteric artery was measured at a relatively shorter distance of 9.19 mm, compared with that in the existing literature . Symptoms of MALS are nonspecific and various range. The most common manifestation is chronic abdominal pain, usually located in the epigastrium and occurring post-prandially. Additionally, patients can experience unintentional weight loss, nausea, vomiting, and diarrhea . Physical examination is usually insignificant, but weight loss and epigastric tenderness can be present. An epigastric bruit may be heard on auscultation up to 35%, particularly during expiration when the median arcuate ligament moves downward . Many patients with MALS are asymptomatic because of the collateral circulation that supplied the foregut, while others can present life-threatening hemorrhage due to post-stenosis aneurysm rupture . Compared with other reported cases, our case shares similarities with 3 others regarding diagnostic challenges and initial misdiagnosis . However, our case did not present with weight loss or vomiting, which are part of the classic symptom triad (post-prandial epigastric pain, weight loss, nausea, and vomiting). This may be attributed to the chronic nature of the 4-year clinical course and differences in dietary patterns, compared with the 3-month symptom history observed in the other reports . Therefore, this syndrome should be considered in the differential list when patients present with chronic abdominal pain, especially when occurring post-prandially and when symptoms are not relieved with empiric treatment, which guides the prompt imaging methods. The diagnosis of MALS typically involves evidence of celiac trunk stenosis and excludes other causes of abdominal pain . EGD and CT scans could help rule out more common conditions. Evidence of celiac trunk stenosis can be gained with duplex ultrasound, CT angiography, or digital subtraction angiography . Duplex ultrasound is considered an initial investigation, because of cost-effectiveness, noninvasiveness, and lack of radiation exposure. Gruber et al highlighted that specific ultrasound findings, such as a celiac trunk deflection angle greater than 50° and an expiratory peak systolic velocity over 350 cm/s, can provide high sensitivity and specificity for MALS diagnosis . Classical CT findings of MALS include proximal celiac trunk narrowing with a characteristic hooked configuration (a visual fold angle < 135°) with post-stenotic dilatation, best viewed on sagittal images, which may differentiate it from atherosclerosis . However, these findings could not exclude the atherosclerosis of the celiac trunk. In our case, we did not perform duplex ultrasound, due to a thickening abdominal wall, resulting in a bad window for ultrasound. However, CT angiography revealed typical findings for MALS, similar to other reported cases . Therefore, imaging modalities, including Doppler ultrasound and CT scans, are essential for confirming the diagnosis . Decompression or intervention is recommended only when symptomatic MALS is confirmed, and other more common causes of symptoms are excluded . The primary goal of MALS management should be to intervene in the pathophysiological process, which requires individual investigation and tailored approaches . Currently, most patients with MALS are treated with laparoscopic MAL release. Symptomatic resolution is usually achieved within 6 weeks after the procedure . The initial success rate ranges from 93% to 94%, and revascularization can further increase the success rate . However, the symptomatic recurrence rate is also high, at 38% for laparoscopic decompression and 50% for open decompression . Furthermore, several complications of both the open and laparoscopic surgery were recorded, such as pneumothorax, celiac artery bleeding, aortic puncture, and splenic infarction . Due to the high recurrence rate and unclear associated factors of MALS, a recent study showed that hybrid laparoscopic and endovascular treatment during the same procedure had better outcomes, with no recurrent symptoms within a 6-month follow-up . Our patient’s case was quite challenging in terms of diagnosis and treatment. Due to aging and dyslipidemia, our patient had risk factors for atherosclerotic disease. Digital subtraction angiography and intravascular ultrasound confirmed the presence of moderately severe atherosclerosis of the celiac trunk, which may explain the progressive symptoms of this patient 1 year earlier. When stenosis is caused not only by external compression from the ligament but also by atherosclerosis, the optimal management approach remains controversial. Apparently, ligament incision alone may not fully resolve the stenosis in such cases. Angioplasty is significantly less invasive and could help relieve internal stenosis; however, it alone carries a higher risk of treatment failure and re-stenosis and requires long-term antiplatelet therapy . Therefore, a combination of laparoscopic release of the median arcuate ligament and angioplasty with stenting may have been the most appropriate approach for this patient. However, the patient opted to defer invasive procedures and surgery, which resulted in endovascular angioplasty and stent implantation as the alternative treatment. The patient was in stable condition at a 5-month follow-up, and a longer monitor is needed to observe the long-term prognosis and guide for further treatment.
In conclusion, median arcuate ligament syndrome is a rare condition with diverse presentations. Physicians should raise high suspicion of this disease in patients with chronic post-prandial epigastric pain. Imaging studies such as duplex ultra-sound and CT angiography are crucial in confirming the diagnosis. Angioplasty and stenting can be performed in selected patients with MALS and atherosclerosis of the celiac trunk.
|
Global epidemiology of | 88650842-8666-4ed0-938d-99c6cc565f44 | 11782492 | Biochemistry[mh] | Tuberculosis (TB) caused by Mycobacterium tuberculosis ( Mtb ) is a significant global health concern, with an estimated 10.6 million infections and 1.6 million casualties by 2021 . The COVID-19 pandemic has disrupted access to medical healthcare, including TB diagnosis and treatment programs, aggravating the TB burden, and compromising the progress in TB control achieved in recent decades , . Furthermore, the incidence of TB has been increasing in several countries, including Ecuador, where the rate reached 48 cases per 100,000 inhabitants in 2021 due, among other factors, to COVID-19 containment measures , . The increase in global migration has contributed to the spread of TB, mainly in high-income countries where migrants seek better economic, educational, or living opportunities. In the Ecuadorian context, migration is primarily directed toward the USA, Spain, Italy, Canada, and Chile – . Identifying the transmission routes of TB cases is essential for reducing potential transmission networks. Different public health organizations have implemented screening programs among household contacts to control the spread of the disease, including the TB-Directly Observed Treatment Short Course (TB‐DOTS) . Genomic approaches, such as whole-genome sequencing (WGS) of Mtb strains, have been a landmark in the traceability of transmitted TB cases and have provided invaluable information on drug resistance and sub-lineage patterns – . WGS and Bayesian phylogenetic approaches have reconstructed the historical patterns of TB spread in Central and South America, dating back to the introduction of Mtb strains in these regions. The apparent ancestral emergence and marked diversification of the L2 (ancestral and modern Beijing sub-lineages) and L4 lineages have also been emphasized, reflecting their close correlation with restricted geographic distribution, leading to the independent emergence of multiple sub-lineages and local adaptation to distinct human populations – . Studies conducted in Brazil, Paraguay, Mexico, and the United States have applied genomic approaches to identify frequent transmission clusters between prison inmates, drug users, migrants, and mixed groups – , highlighting the need to prioritize contact tracing to groups with a higher likelihood of retrospective clustering to improve TB control. The Mycobacterium tuberculosis complex (MTBC) comprises distinct phylogenetic lineages that have evolved over centuries – . The Euro-American lineage (L4) exhibits extensive sub-lineage diversity both within and across countries. In Central and South America, the Caribbean, Europe, and Middle Africa, the most prevalent L4 sub-lineages include Latin American (LAM), Haarlem, X-type, and T families – . Previous reports have revealed that the LAM and Haarlem families are the most prevalent in Ecuador, with few cases in the Beijing family – . Transmission clustering in these cases has often been linked to isolates from neighboring countries , . A pivotal WGS study on a limited dataset of Ecuadorian Mtb isolates identified the 4.3.2/3 (LAM) and 4.1.2 (Haarlem) sub-lineages, highlighting the significant genetic diversity present . This study underscores the importance of understanding the local transmission dynamics of TB within Ecuador, particularly considering the potential influence of events occurring outside national boundaries, associated with migration patterns, trade, and regional cooperation. Although studies have investigated transnational TB transmission between Ecuador and countries such as Colombia , evidence remains sparse. We analyzed the transmission networks of 88 Mtb isolates of L4 lineage collected from various cities across Ecuador in comparison with 415 publicly available genomes in 19 Latin American countries. To accomplish this, we used a combination of phylogenetics and composition analysis. Additionally, we explored the role of genetic variation in virulence genes in the recent transmission of TB in Ecuador. Our findings provide crucial insights into the genetic diversity of Mtb in Ecuador, highlighting how migration influences the TB burden and transmission dynamics locally. This study contributes to the development of more effective targeted TB control strategies tailored to the unique characteristics of Mtb populations.
Genomic and functional analysis of Ecuadorian Mtb isolates This study focused on genomic analysis of raw reads of 88 M . tuberculosis isolates collected in Ecuador . To extend our investigation of genomic variability and epidemiology, we combined these data with a selection of 415 M . tuberculosis isolates from different countries (see Methods). Initially, we used the raw sequences (average coverage of 61X) of the 88 Ecuadorian isolates to perform quality control and genome assembly for each isolate using Unicycler and Pilon while discarding contigs smaller than 300 bp. The quality of the genomes was assessed and an average N50 of 65,747. Annotation was performed using Prokka software. These analyses yielded approximately 4298 genes per isolate, including coding sequences and RNA gene transfer. Descriptive statistics for this analysis are presented in Table and Supplementary Table . Genomic analysis of the Ecuadorian M. tuberculosis isolates revealed that they belonged to the Euro-American Lineage (lineage 4). Among the sublineages, 4.3.3 (27.3%, 24/88) and 4.1.1 (23.9%, 21/88) were the most prevalent, followed by 4.4.1.1 and 4.1.2.1 (11.4%, 10/88 each). Clades 4.1.2, and 4.3.4.1/2 were present, but in smaller proportions. Functional annotation and metabolic insights We performed functional annotation of each of the 88 Ecuadorian genomes. A substantial proportion (77%) of the annotated genes was associated with specific functional roles (e.g., metabolic processes, cellular processes, energy, protein processing, and stress response, Defense, and Virulence). We enriched our dataset with functional annotations, assigning Enzyme Commission (EC) numbers to 1,061 proteins, Gene Ontology (GO) classifications to 918 proteins, linking 814 proteins to specific KEGG pathways, and identifying approximately 816 virulence factors distributed across three databases. This enrichment analysis provided deep insights into the metabolic functions and potential pathogenic mechanisms of the isolates (Supplementary Table ). Our subsystem analysis, categorized according to the KEGG database, showed that approximately 1,961 genes per isolate are involved in various biological processes and structural complexes (Supplementary Table ). The analysis revealed that the majority of the annotated protein-coding genes were related to metabolism (40.3%), with 37.9% of these genes being associated with cofactors, vitamins, and prosthetic groups (300 genes). This was followed by genes related to the stress response, Defense, Virulence (10.8%), and energy (10.5%). In our detailed metabolic pathway analysis, we observed a predominance of genes involved in amino acid (369 genes), carbohydrate (312 genes), lipid (262 genes), and xenobiotic biodegradation and metabolism (240 genes). These pathways are crucial for the pathogenicity and survival of Mtb (Supplementary Table – ). Notably, variability was observed among different sub-lineages. Sub-lineage 4.1.2.1 shows a higher proportion of proteins related to metabolism and a lower proportion of proteins related to protein processing. Conversely, sub-lineage 4.1.2 presents a higher proportion of proteins related to protein processing and fewer proteins related to metabolism (Supplementary Fig. A). Lineage 4.1.2.1 exhibits a higher number of gene variations related to metabolism, suggesting significant metabolic diversity, indicating a complex and adaptable metabolic network, whereas lineage 4.3.2/3 showed greater variations in genes associated with protein processing, suggesting a diverse set of genes involved in protein synthesis, folding, and degradation, reflecting the complexity of protein processing mechanisms (Table ). Pangenome analysis of the Ecuadorian Mtb isolates We also performed pan-genome analysis to understand the genetic diversity and conservation of 88 Mtb isolates from Ecuador. Pangenome reconstruction of the 88 Mtb isolates revealed 4,397 gene families, including 3,104 classified as core (present in at least 99% of isolates), 666 as accessory, and 270 as cloud gene families (Supplementary Fig. B and Supplementary Table ). Interestingly, 70.5% of the isolates belonged to core gene families, suggesting minimal within-genome variability and emphasizing the high level of genetic conservation among the isolates. According to pangenome calculations, a b value of 0.086 in the power-law regression model indicated a close pangenome for Mtb . Three isolates (S1454, S1477, and S1453) from sub-lineages 4.1.1 and 4.3.2/3 showed the highest number of cloud genes (261, 200, and 177, respectively). Most gene families within the core and accessory partitions belong to metabolic subsystems. Conversely, the majority of the unique gene families were classified within the Environmental Information Processing subsystem (Supplementary Fig. C,D). Genotypic drug-resistance analysis Furthermore, when analyzing the phylogenetic relationships and genetic diversity associated with resistance in Ecuadorian isolates (Fig. ), our findings revealed that sub-lineages 4.3.2/3 and 4.1.1, exhibited a higher prevalence of resistant variants (46.6% and 26.1%, respectively). We identified 42 single-nucleotide variants (SNVs) in resistance-related genes, with rpoB showing the most mutations. Additionally, SNVs were more frequently identified in genes related to intermediary metabolism and respiration (21.71%), followed by cell wall processes (21.34%) and conserved hypotheticals (20.44%). Exhaustive details of the SNV are provided in Supplementary Table . To gain a deeper insight into the genomic epidemiology of TB in Ecuador, we incorporated an additional 415 Mtb samples from 19 countries categorized as both continental and transoceanic migratory nations (See “ ” section). All these 503 Mtb genomes belonged to the L4 lineage, including 4.3.2/3 (35.4%), 4.1.2 (22.7%), 4.4.1 (12.7%), and 4.1.1 (10.7%) sub-lineages (Supplementary Table ). Moreover, 63.8% (321/503) of Mtb samples were genotypically susceptible to all anti-TB drugs. The remaining 36.2% (180/503) were resistant to at least one antibiotic. The clinical classification of these drug-resistant samples showed that 16.7% were identified as MDR-TB, 8.5% as HR-TB, 4.2% as pre-XDR-TB, and 1.6% as RR-TB. Table summarizes the drug-resistant clinical classification for the 503 Mtb samples per region and Table provides an overview of their drug-resistant canonical mutations. An extended list of canonical mutations is provided in Supplementary Table . Since multidrug-resistant TB (MDR-TB) and extensively drug-resistant TB (XDR-TB) can increase domestic transmission , we determined the frequency of resistance-associated canonical mutations in 503 Mtb isolates. The most frequent mutations were katG Ser315Thr (n = 92), rpoB Ser450Leu (n = 68), rpsL Lys43Arg (n = 21), embB Met306Ile (n = 17), pncA Gln10Pro (n = 14), and gyrA Ala90Val (n = 11), which confer resistance to INH, RIF, STR, EMB, PZA, and FQs, respectively. Other common canonical mutations included fabG 1 − 15C > T (n = 16), ahpC -74G > A (n = 15), and gid 329_330delTG (n = 16), which are associated with INH and STR resistance, respectively. TB population structure and transmission clusters To better understand the genetic diversity and potential transmission dynamics of TB in Ecuador, we performed a population structure analysis of 503 Mtb samples (Fig. ). Clusterization based on the rhierBAPS approach revealed four primary clusters at the initial hierarchical level, which were further divided into 15 sub-clusters at the second hierarchical level. Phylogenetic reconstruction of 503 isolates using 1,437 SNPs revealed four main clusters with different branch colors. Additionally, the sub-clusters (level 2), lineages, geographical distribution, and resistance are represented by colored concentric lines. Cluster 1 (pink) encompassed 128 isolates divided into three sub-clusters (1–3), containing samples belonging to 4, 4.4.1.1, 4.2.2, 4.6.2, and 4.7/8 sub-lineages. Cluster 2 (in orange) comprises 125 isolates, including three sub-clusters (4–6), with the majority belonging to the 4.1.2.1 sub-lineage, and some isolates belonging to the 4.1.2 sub-lineage (sub-cluster 6). Cluster 3 (in green) was composed of 178 isolates, distributed across three sub-clusters (7–9), primarily associated with 4.3.2, 4.3.3, 4.3.4.1, and 4.3.4.2, and, unexpectedly, a 4.1.2 isolate within sub-cluster 9. Finally, Cluster 4 (in blue) encompasses six sub-clusters (10–15), including 71 isolates predominantly from the 4.1.1, 4.1.1.1, and 4.1.1.3 sub-lineages. Notably, the isolates formed subclusters 13 and 15, with a small portion of subcluster 10 corresponding to the 4.1 sub-lineage. In addition, we analyzed the isolates for transmission clusters by pairwise comparisons using a minimum distance of 12 SNPs with MTBseq. Thus, strains in the same transmission genomic cluster (TGC) had fewer than 12 SNPs. Using this strategy, 92.8% (n = 467) of the isolates were classified into 51 TGCs, ranging from two to 63 members, whereas 35 isolates were not classified. Based on the number of members per TGC (see “ ” section), we classified seven members as small , 20 as medium , and 24 as large (Supplementary Table ). Most of the 51 TGCs resolved using this methodology were within Cluster 3 (37.3%) and Cluster 1 (29.4%), as determined using rhierBAPS (Fig. A). In particular, Mtb samples identified as susceptible were more prevalent in larger TGC than in resistant samples. Most pre-XDR isolates exhibited clonal relationships either among themselves or with isolates from other countries. Of the 35 isolates not grouped by MTBseq, 28.6% belonged to sub-lineage 4.3.2/3/4 and 25.7% to sub-lineage 4.8 (Fig. A and Table ). Upon closer examination of the five largest TGCs, specifically, TGC_1, TGC_2, TGC_9, TGC_18, and TGC_19, distinct geographical profiles were identified. Notably, TGC_1 (n = 63 isolates) predominantly comprised Mtb isolates from Paraguay (46.3%), Peru (19.0%), and Spain (15.9%). TGC_18 (n = 56) included isolates from Ecuador (41.0%) and Paraguay (26.8%), whereas TGC_19 (n = 31) consisted solely of isolates from Spain (38.7%). TGC_2 (n = 30) was predominantly isolated from Argentina (53.3%), whereas TGC_9 (n = 24) was isolated mainly from Canada (54.1%). Among the 88 Ecuadorian Mtb isolates, 84 were grouped into 15 of 51 identified TGCs. These included five medium-sized and ten large TGCs. Notably, ten of these TGCs (TGC_2, TGC_3, TGC_5, TGC_14, TGC_15, TGC_18, TGC_20, TGC_21, TGC_37, and TGC_47) included isolates from Ecuador alongside isolates from various countries, unveiling previously unrecognized potential continental and intercontinental connections. In particular, TGC_37 and TGC_14 exhibited a close genetic relationship among their members, showing distances of ≤ 9 SNPs and ≤ 6 SNPs, respectively. In addition, TGC_37 comprised five isolates from Ecuador and one from an Ecuadorian migrant in Spain, highlighting the impact of migration on TB spread. Conversely, TGC_14 grouped two isolates from Ecuador with five from Brazil, indicating active transmission between these nations (Supplementary Table ). The Ecuadorian isolates within each TGC exhibited a high degree of clonality. For example, approximately 90% of the Ecuadorian isolates in TGC_11 and 70% in TGC_15 and TGC_18 had a genetic distance of zero SNPs (Supplementary Tables , and Supplementary Figs. , ). Similarly, in some TGCs, many isolates remained distinct and did not cluster with other isolates (singletons) in various proportions. (Supplementary Table ). Regarding the sub-lineage composition, the most representative sub-lineages in the Ecuadorian isolates among the TGCs were 4.3.3 (46.9%, 23/49; TGC_18), 4.1.1 (32.7%, 16/49; TGC_15), and 4.4.1.1 (20.4%, 10/49; TGC_11). Concerning the potential transmission of isolates, we identified four pre-XDR TB cases in TGC_15 and three pre-XDR isolates in TGC_15 and TGC_18, corresponding to clonal samples (genetic distance < 1 SNP). Among these three TGCs, 41% were from previously treated patients with TB, 34% were from untreated patients, and 25% were from patients currently undergoing treatment. TB transmission networks analysis TB transmission networks, that is, the suggested transmission routes of TB considering the genetic distances between samples, their geographical location, and year of isolation, were inferred for the 503 Mtb isolates using TransFlow. Using this strategy, 46 transmission networks were identified; however, only 18 of these networks had more than three members, as shown in Fig. A. As can be seen, most of these transmission networks (with more than three members) correspond to different TGCs, meaning they present more than 12 SNPs between them. However, six transmission networks (7, 11–14, and 16) corresponded to TGC_18, indicating the low genetic distance between them (Fig. A). Considering the three clustering strategies employed (rhierBAPS, MTBseq, and TransFlow), we identified 17 convergent transmission groups to define potential TB transmission networks between Ecuadorian and non-Ecuadorian isolates. The geographic distribution of networks involving Ecuadorian isolates is shown in Fig. B, where TGCs that were confirmed and congruent with TransFlow are highlighted in red. Notably, some of these Ecuadorian samples (collected between 2019 and 2021) showed clonality with isolates from Colombia (n = 2) and Latin American migrants in Spain (n = 4), which were detected in 2014 and 2015, respectively (Fig. A, Supplementary Figs. , and Supplementary Table ). We identified several transmission networks showing connections among countries, including Paraguay, Argentina, and Brazil, with specific samples acting as potential index cases in the corresponding transmission networks (Supplementary Fig. ). The distribution of local and migrant Ecuadorian isolates revealed diverse representations spread across 12 TGCs (Fig. C). Some of these isolates showed clonality with samples from Colombia, Latin America, and Ecuadorian migrants in Spain, or were identified as potential index cases within transmission networks (Fig. C and Supplementary Fig. ). Notably, TGC18 and 19 exhibit diverse nationalities, with 10 and 8 different origins represented within these clusters, respectively. The short genetic distance between different TGCs indicated a close relationship, suggesting a possible joint origin. This relationship was supported by the genetic distances observed between the Ecuadorian isolates, implying a shared ancestry among the TGCs (Fig. D). These findings highlight the complex transmission dynamics of TB, with evidence of cross-border and international transmission networks involving Ecuadorian populations and those from other Latin American countries. The diversity of isolates and potential index cases underscore the need for targeted public health interventions to address the multifaceted nature of TB epidemiology. Genetic diversity within the virulence-associated genes To characterize mutations in the genes involved in host adaptability, we analyzed SNPs in genes commonly associated with MTBC virulence in Ecuadorian isolates . A total of 303 SNPs were identified in these virulence genes, and all 88 isolates had at least one SNP each in the mce1F, mmpL4, phoR, ctpV, pepD, mce3F, fadD13, and nuoG genes. The pks12, fadD5, mce3C, pks12, nuoG, and katG genes were mutated with at least one SNP in more than 50% of the isolates. The top six genes with the most significant SNPs were plcA, plcB, pks7, pks12, phoR, and PPE46 (Supplementary Tables , ). Among the isolates, 46.6% (41/88) had more than 40 SNPs, with a maximum of 79 SNPs identified within virulence genes. Isolates corresponding to sub-lineages 4.1.1 (TGC_15 and TGC_37), 4.4.1.1 (TGC_11), and 4.3.2/3/4 (TGC_18, _20, _21, _26, _29, and _30) presented higher numbers of virulence-associated polymorphisms. Additionally, two members of TGC_14 (S0017 and S0039) and one ungrouped sample (S2193) showed a high number of mutations in the virulence genes (70, 71, and 74 SNPs, respectively). Interestingly, the five isolates associated with sub-lineage 4.7/8 showed the lowest number of SNPs in the virulence genes (15–25 SNPs). Phylodynamics of TB in Ecuador The divergence time to the most recent common ancestor (TMRCA) estimated for the Ecuadorian isolates fell within a density interval between 697 and 1,475 years, with an estimated time of 1,054 years. We also estimated TMRCA for different TGCs, identifying TGC_15 as the oldest (218 years before present, YBP) and TGC_03 as the youngest (130 YBP). Notably, isolates corresponding to TGC_02, TGC_21, and TGC_47, each of which included only one Ecuadorian isolate, had TMRCAs greater than 150 YBP (191, 208, and 229 YBP, respectively). TMRCAs were also estimated for the ungrouped strains, ranging from 157 to 404 YBP (Supplementary Fig. ). Because most Ecuadorian Mtb isolates were closely related to the 4.3.2/3 and 4.1.1 sub-lineages, we focused on all isolates belonging to these sub-lineages. The TMRCA estimates for the 4.3.2/3 and 4.1.1 sub-lineages were 470 YBP (95% HPD, 358 to 720) and 450 YBP (95% HPD, 349 to 647), respectively. Within the 4.1.1 sub-lineage, TGC_15 members have a TMRCA of 291 YBP, while those from TGC_37 have a TMRCA of 254 YBP. Isolates from TGC_14 had a TMRCA of 238 YBP, whereas ungrouped isolates S0516 and S2193 had TMRCA estimates of 482 and 354 YBP, respectively. In the 4.3.2/3 sub-lineage, TMRCA estimates ranged from 242 to 295 YBP, covering six TGCs: TGC_26 (289 YBP), TGC_18 (242–265 YBP), TGC_30 (249 YBP), TGC_20 (271–293 YBP), TGC_21 (295 YBP), and TGC_29 (292 YBP), and one ungrouped isolate (S2192, 289 YBP). Notably, isolate S2192 was included in the clade containing TGC_26, and TGC_18 was divided into three clades. In the 4.1.2.1 sub-lineage, isolates have a TMRCA estimated to be between 237 and 292 YBP, including 263 YBP for isolates corresponding to TGC_05, 237 to 273 YBP for TGC_03, and 292 YBP for TGC_02. The S-type sub-lineage had a TMRCA of 257 YBP (TGC_11), while isolates corresponding to the 4.8 sub-lineage had a TMRCA of 278 YBP (TGC_48) and 269 YBP (TGC_47). Isolate S0137, which formed a clade with TGC_48, had a TMRCA of 340 YBP (Supplementary Fig. ).
This study focused on genomic analysis of raw reads of 88 M . tuberculosis isolates collected in Ecuador . To extend our investigation of genomic variability and epidemiology, we combined these data with a selection of 415 M . tuberculosis isolates from different countries (see Methods). Initially, we used the raw sequences (average coverage of 61X) of the 88 Ecuadorian isolates to perform quality control and genome assembly for each isolate using Unicycler and Pilon while discarding contigs smaller than 300 bp. The quality of the genomes was assessed and an average N50 of 65,747. Annotation was performed using Prokka software. These analyses yielded approximately 4298 genes per isolate, including coding sequences and RNA gene transfer. Descriptive statistics for this analysis are presented in Table and Supplementary Table . Genomic analysis of the Ecuadorian M. tuberculosis isolates revealed that they belonged to the Euro-American Lineage (lineage 4). Among the sublineages, 4.3.3 (27.3%, 24/88) and 4.1.1 (23.9%, 21/88) were the most prevalent, followed by 4.4.1.1 and 4.1.2.1 (11.4%, 10/88 each). Clades 4.1.2, and 4.3.4.1/2 were present, but in smaller proportions.
We performed functional annotation of each of the 88 Ecuadorian genomes. A substantial proportion (77%) of the annotated genes was associated with specific functional roles (e.g., metabolic processes, cellular processes, energy, protein processing, and stress response, Defense, and Virulence). We enriched our dataset with functional annotations, assigning Enzyme Commission (EC) numbers to 1,061 proteins, Gene Ontology (GO) classifications to 918 proteins, linking 814 proteins to specific KEGG pathways, and identifying approximately 816 virulence factors distributed across three databases. This enrichment analysis provided deep insights into the metabolic functions and potential pathogenic mechanisms of the isolates (Supplementary Table ). Our subsystem analysis, categorized according to the KEGG database, showed that approximately 1,961 genes per isolate are involved in various biological processes and structural complexes (Supplementary Table ). The analysis revealed that the majority of the annotated protein-coding genes were related to metabolism (40.3%), with 37.9% of these genes being associated with cofactors, vitamins, and prosthetic groups (300 genes). This was followed by genes related to the stress response, Defense, Virulence (10.8%), and energy (10.5%). In our detailed metabolic pathway analysis, we observed a predominance of genes involved in amino acid (369 genes), carbohydrate (312 genes), lipid (262 genes), and xenobiotic biodegradation and metabolism (240 genes). These pathways are crucial for the pathogenicity and survival of Mtb (Supplementary Table – ). Notably, variability was observed among different sub-lineages. Sub-lineage 4.1.2.1 shows a higher proportion of proteins related to metabolism and a lower proportion of proteins related to protein processing. Conversely, sub-lineage 4.1.2 presents a higher proportion of proteins related to protein processing and fewer proteins related to metabolism (Supplementary Fig. A). Lineage 4.1.2.1 exhibits a higher number of gene variations related to metabolism, suggesting significant metabolic diversity, indicating a complex and adaptable metabolic network, whereas lineage 4.3.2/3 showed greater variations in genes associated with protein processing, suggesting a diverse set of genes involved in protein synthesis, folding, and degradation, reflecting the complexity of protein processing mechanisms (Table ).
We also performed pan-genome analysis to understand the genetic diversity and conservation of 88 Mtb isolates from Ecuador. Pangenome reconstruction of the 88 Mtb isolates revealed 4,397 gene families, including 3,104 classified as core (present in at least 99% of isolates), 666 as accessory, and 270 as cloud gene families (Supplementary Fig. B and Supplementary Table ). Interestingly, 70.5% of the isolates belonged to core gene families, suggesting minimal within-genome variability and emphasizing the high level of genetic conservation among the isolates. According to pangenome calculations, a b value of 0.086 in the power-law regression model indicated a close pangenome for Mtb . Three isolates (S1454, S1477, and S1453) from sub-lineages 4.1.1 and 4.3.2/3 showed the highest number of cloud genes (261, 200, and 177, respectively). Most gene families within the core and accessory partitions belong to metabolic subsystems. Conversely, the majority of the unique gene families were classified within the Environmental Information Processing subsystem (Supplementary Fig. C,D).
Furthermore, when analyzing the phylogenetic relationships and genetic diversity associated with resistance in Ecuadorian isolates (Fig. ), our findings revealed that sub-lineages 4.3.2/3 and 4.1.1, exhibited a higher prevalence of resistant variants (46.6% and 26.1%, respectively). We identified 42 single-nucleotide variants (SNVs) in resistance-related genes, with rpoB showing the most mutations. Additionally, SNVs were more frequently identified in genes related to intermediary metabolism and respiration (21.71%), followed by cell wall processes (21.34%) and conserved hypotheticals (20.44%). Exhaustive details of the SNV are provided in Supplementary Table . To gain a deeper insight into the genomic epidemiology of TB in Ecuador, we incorporated an additional 415 Mtb samples from 19 countries categorized as both continental and transoceanic migratory nations (See “ ” section). All these 503 Mtb genomes belonged to the L4 lineage, including 4.3.2/3 (35.4%), 4.1.2 (22.7%), 4.4.1 (12.7%), and 4.1.1 (10.7%) sub-lineages (Supplementary Table ). Moreover, 63.8% (321/503) of Mtb samples were genotypically susceptible to all anti-TB drugs. The remaining 36.2% (180/503) were resistant to at least one antibiotic. The clinical classification of these drug-resistant samples showed that 16.7% were identified as MDR-TB, 8.5% as HR-TB, 4.2% as pre-XDR-TB, and 1.6% as RR-TB. Table summarizes the drug-resistant clinical classification for the 503 Mtb samples per region and Table provides an overview of their drug-resistant canonical mutations. An extended list of canonical mutations is provided in Supplementary Table . Since multidrug-resistant TB (MDR-TB) and extensively drug-resistant TB (XDR-TB) can increase domestic transmission , we determined the frequency of resistance-associated canonical mutations in 503 Mtb isolates. The most frequent mutations were katG Ser315Thr (n = 92), rpoB Ser450Leu (n = 68), rpsL Lys43Arg (n = 21), embB Met306Ile (n = 17), pncA Gln10Pro (n = 14), and gyrA Ala90Val (n = 11), which confer resistance to INH, RIF, STR, EMB, PZA, and FQs, respectively. Other common canonical mutations included fabG 1 − 15C > T (n = 16), ahpC -74G > A (n = 15), and gid 329_330delTG (n = 16), which are associated with INH and STR resistance, respectively.
To better understand the genetic diversity and potential transmission dynamics of TB in Ecuador, we performed a population structure analysis of 503 Mtb samples (Fig. ). Clusterization based on the rhierBAPS approach revealed four primary clusters at the initial hierarchical level, which were further divided into 15 sub-clusters at the second hierarchical level. Phylogenetic reconstruction of 503 isolates using 1,437 SNPs revealed four main clusters with different branch colors. Additionally, the sub-clusters (level 2), lineages, geographical distribution, and resistance are represented by colored concentric lines. Cluster 1 (pink) encompassed 128 isolates divided into three sub-clusters (1–3), containing samples belonging to 4, 4.4.1.1, 4.2.2, 4.6.2, and 4.7/8 sub-lineages. Cluster 2 (in orange) comprises 125 isolates, including three sub-clusters (4–6), with the majority belonging to the 4.1.2.1 sub-lineage, and some isolates belonging to the 4.1.2 sub-lineage (sub-cluster 6). Cluster 3 (in green) was composed of 178 isolates, distributed across three sub-clusters (7–9), primarily associated with 4.3.2, 4.3.3, 4.3.4.1, and 4.3.4.2, and, unexpectedly, a 4.1.2 isolate within sub-cluster 9. Finally, Cluster 4 (in blue) encompasses six sub-clusters (10–15), including 71 isolates predominantly from the 4.1.1, 4.1.1.1, and 4.1.1.3 sub-lineages. Notably, the isolates formed subclusters 13 and 15, with a small portion of subcluster 10 corresponding to the 4.1 sub-lineage. In addition, we analyzed the isolates for transmission clusters by pairwise comparisons using a minimum distance of 12 SNPs with MTBseq. Thus, strains in the same transmission genomic cluster (TGC) had fewer than 12 SNPs. Using this strategy, 92.8% (n = 467) of the isolates were classified into 51 TGCs, ranging from two to 63 members, whereas 35 isolates were not classified. Based on the number of members per TGC (see “ ” section), we classified seven members as small , 20 as medium , and 24 as large (Supplementary Table ). Most of the 51 TGCs resolved using this methodology were within Cluster 3 (37.3%) and Cluster 1 (29.4%), as determined using rhierBAPS (Fig. A). In particular, Mtb samples identified as susceptible were more prevalent in larger TGC than in resistant samples. Most pre-XDR isolates exhibited clonal relationships either among themselves or with isolates from other countries. Of the 35 isolates not grouped by MTBseq, 28.6% belonged to sub-lineage 4.3.2/3/4 and 25.7% to sub-lineage 4.8 (Fig. A and Table ). Upon closer examination of the five largest TGCs, specifically, TGC_1, TGC_2, TGC_9, TGC_18, and TGC_19, distinct geographical profiles were identified. Notably, TGC_1 (n = 63 isolates) predominantly comprised Mtb isolates from Paraguay (46.3%), Peru (19.0%), and Spain (15.9%). TGC_18 (n = 56) included isolates from Ecuador (41.0%) and Paraguay (26.8%), whereas TGC_19 (n = 31) consisted solely of isolates from Spain (38.7%). TGC_2 (n = 30) was predominantly isolated from Argentina (53.3%), whereas TGC_9 (n = 24) was isolated mainly from Canada (54.1%). Among the 88 Ecuadorian Mtb isolates, 84 were grouped into 15 of 51 identified TGCs. These included five medium-sized and ten large TGCs. Notably, ten of these TGCs (TGC_2, TGC_3, TGC_5, TGC_14, TGC_15, TGC_18, TGC_20, TGC_21, TGC_37, and TGC_47) included isolates from Ecuador alongside isolates from various countries, unveiling previously unrecognized potential continental and intercontinental connections. In particular, TGC_37 and TGC_14 exhibited a close genetic relationship among their members, showing distances of ≤ 9 SNPs and ≤ 6 SNPs, respectively. In addition, TGC_37 comprised five isolates from Ecuador and one from an Ecuadorian migrant in Spain, highlighting the impact of migration on TB spread. Conversely, TGC_14 grouped two isolates from Ecuador with five from Brazil, indicating active transmission between these nations (Supplementary Table ). The Ecuadorian isolates within each TGC exhibited a high degree of clonality. For example, approximately 90% of the Ecuadorian isolates in TGC_11 and 70% in TGC_15 and TGC_18 had a genetic distance of zero SNPs (Supplementary Tables , and Supplementary Figs. , ). Similarly, in some TGCs, many isolates remained distinct and did not cluster with other isolates (singletons) in various proportions. (Supplementary Table ). Regarding the sub-lineage composition, the most representative sub-lineages in the Ecuadorian isolates among the TGCs were 4.3.3 (46.9%, 23/49; TGC_18), 4.1.1 (32.7%, 16/49; TGC_15), and 4.4.1.1 (20.4%, 10/49; TGC_11). Concerning the potential transmission of isolates, we identified four pre-XDR TB cases in TGC_15 and three pre-XDR isolates in TGC_15 and TGC_18, corresponding to clonal samples (genetic distance < 1 SNP). Among these three TGCs, 41% were from previously treated patients with TB, 34% were from untreated patients, and 25% were from patients currently undergoing treatment.
TB transmission networks, that is, the suggested transmission routes of TB considering the genetic distances between samples, their geographical location, and year of isolation, were inferred for the 503 Mtb isolates using TransFlow. Using this strategy, 46 transmission networks were identified; however, only 18 of these networks had more than three members, as shown in Fig. A. As can be seen, most of these transmission networks (with more than three members) correspond to different TGCs, meaning they present more than 12 SNPs between them. However, six transmission networks (7, 11–14, and 16) corresponded to TGC_18, indicating the low genetic distance between them (Fig. A). Considering the three clustering strategies employed (rhierBAPS, MTBseq, and TransFlow), we identified 17 convergent transmission groups to define potential TB transmission networks between Ecuadorian and non-Ecuadorian isolates. The geographic distribution of networks involving Ecuadorian isolates is shown in Fig. B, where TGCs that were confirmed and congruent with TransFlow are highlighted in red. Notably, some of these Ecuadorian samples (collected between 2019 and 2021) showed clonality with isolates from Colombia (n = 2) and Latin American migrants in Spain (n = 4), which were detected in 2014 and 2015, respectively (Fig. A, Supplementary Figs. , and Supplementary Table ). We identified several transmission networks showing connections among countries, including Paraguay, Argentina, and Brazil, with specific samples acting as potential index cases in the corresponding transmission networks (Supplementary Fig. ). The distribution of local and migrant Ecuadorian isolates revealed diverse representations spread across 12 TGCs (Fig. C). Some of these isolates showed clonality with samples from Colombia, Latin America, and Ecuadorian migrants in Spain, or were identified as potential index cases within transmission networks (Fig. C and Supplementary Fig. ). Notably, TGC18 and 19 exhibit diverse nationalities, with 10 and 8 different origins represented within these clusters, respectively. The short genetic distance between different TGCs indicated a close relationship, suggesting a possible joint origin. This relationship was supported by the genetic distances observed between the Ecuadorian isolates, implying a shared ancestry among the TGCs (Fig. D). These findings highlight the complex transmission dynamics of TB, with evidence of cross-border and international transmission networks involving Ecuadorian populations and those from other Latin American countries. The diversity of isolates and potential index cases underscore the need for targeted public health interventions to address the multifaceted nature of TB epidemiology.
To characterize mutations in the genes involved in host adaptability, we analyzed SNPs in genes commonly associated with MTBC virulence in Ecuadorian isolates . A total of 303 SNPs were identified in these virulence genes, and all 88 isolates had at least one SNP each in the mce1F, mmpL4, phoR, ctpV, pepD, mce3F, fadD13, and nuoG genes. The pks12, fadD5, mce3C, pks12, nuoG, and katG genes were mutated with at least one SNP in more than 50% of the isolates. The top six genes with the most significant SNPs were plcA, plcB, pks7, pks12, phoR, and PPE46 (Supplementary Tables , ). Among the isolates, 46.6% (41/88) had more than 40 SNPs, with a maximum of 79 SNPs identified within virulence genes. Isolates corresponding to sub-lineages 4.1.1 (TGC_15 and TGC_37), 4.4.1.1 (TGC_11), and 4.3.2/3/4 (TGC_18, _20, _21, _26, _29, and _30) presented higher numbers of virulence-associated polymorphisms. Additionally, two members of TGC_14 (S0017 and S0039) and one ungrouped sample (S2193) showed a high number of mutations in the virulence genes (70, 71, and 74 SNPs, respectively). Interestingly, the five isolates associated with sub-lineage 4.7/8 showed the lowest number of SNPs in the virulence genes (15–25 SNPs).
The divergence time to the most recent common ancestor (TMRCA) estimated for the Ecuadorian isolates fell within a density interval between 697 and 1,475 years, with an estimated time of 1,054 years. We also estimated TMRCA for different TGCs, identifying TGC_15 as the oldest (218 years before present, YBP) and TGC_03 as the youngest (130 YBP). Notably, isolates corresponding to TGC_02, TGC_21, and TGC_47, each of which included only one Ecuadorian isolate, had TMRCAs greater than 150 YBP (191, 208, and 229 YBP, respectively). TMRCAs were also estimated for the ungrouped strains, ranging from 157 to 404 YBP (Supplementary Fig. ). Because most Ecuadorian Mtb isolates were closely related to the 4.3.2/3 and 4.1.1 sub-lineages, we focused on all isolates belonging to these sub-lineages. The TMRCA estimates for the 4.3.2/3 and 4.1.1 sub-lineages were 470 YBP (95% HPD, 358 to 720) and 450 YBP (95% HPD, 349 to 647), respectively. Within the 4.1.1 sub-lineage, TGC_15 members have a TMRCA of 291 YBP, while those from TGC_37 have a TMRCA of 254 YBP. Isolates from TGC_14 had a TMRCA of 238 YBP, whereas ungrouped isolates S0516 and S2193 had TMRCA estimates of 482 and 354 YBP, respectively. In the 4.3.2/3 sub-lineage, TMRCA estimates ranged from 242 to 295 YBP, covering six TGCs: TGC_26 (289 YBP), TGC_18 (242–265 YBP), TGC_30 (249 YBP), TGC_20 (271–293 YBP), TGC_21 (295 YBP), and TGC_29 (292 YBP), and one ungrouped isolate (S2192, 289 YBP). Notably, isolate S2192 was included in the clade containing TGC_26, and TGC_18 was divided into three clades. In the 4.1.2.1 sub-lineage, isolates have a TMRCA estimated to be between 237 and 292 YBP, including 263 YBP for isolates corresponding to TGC_05, 237 to 273 YBP for TGC_03, and 292 YBP for TGC_02. The S-type sub-lineage had a TMRCA of 257 YBP (TGC_11), while isolates corresponding to the 4.8 sub-lineage had a TMRCA of 278 YBP (TGC_48) and 269 YBP (TGC_47). Isolate S0137, which formed a clade with TGC_48, had a TMRCA of 340 YBP (Supplementary Fig. ).
Human-adapted MTB strains show a high degree of genomic conservation but vary in geographic distribution, virulence, transmissibility, and drug resistance patterns . To better understand the transmission pathways within our study population, we analyzed the sequences of 503 Mtb isolates of the L4 lineage, mostly from neighboring regions in Ecuador and other areas worldwide. It should be noted that the number of isolates varies significantly between countries. For instance, El Salvador and Chile have only one sample each, while countries such as Brazil and Paraguay have more than one hundred. This heterogeneity may affect the analysis of geographic distribution and limit the interpretation of the results. Our analysis of high-quality genomes revealed that 63.8% of the isolates were sensitive to all drugs used for TB treatment, whereas the remaining 36.2% were resistant to at least one drug. The genomes of all the isolates studied belonged to the Euro-American lineage, with the most common sub-lineages being 4.3.2/3 (35.4%), 4.1.2.1 (22.7%), 4.4.1 (12.7%), and 4.1.1. (10.7%). These results were congruent with those of previous studies in Ecuador that used the MIRU-VNTR strategy for genotyping circulating MTBC strains and also identified other Mtb lineages, including L2.1, and L4.2 , , . The distribution of these sub-lineages suggests that both historical European roots , , and migratory processes , , contributed to their presence in South and Central America and the Caribbean. Furthermore, we identified 19 genes that harbor mutations associated with resistance, predominantly linked to resistance to first-line drugs, thus enhancing the local genetic data for TB research , , . Functional characterization of genes is crucial for understanding how microorganisms such as M. tuberculosis adapt and survive in their hosts. Annotation of protein-encoding genes revealed that genes mainly associated with Cofactors, Vitamins, and Prosthetic groups, fatty acids, Lipids, and Isoprenoids, and Amino acids, and derivatives were most representative. Similar findings have been reported in other M. tuberculosis populations, where genes related to energy production and conversion, amino acid transport and metabolism, and lipid transport and metabolism are highly represented – . The wide conservation of these genes indicates their importance in interactions between bacteria and their human hosts. Specifically, during mycobacterial persistence, when the host–pathogen struggles for nutrient and immune recognition, these genes play a crucial role in ensuring the survival and adaptability of bacteria. On the other hand, we found that certain genes associated with the virulence of Mycobacterium tuberculosis like ESAT-6-like protein EsxS (63.6%), Acid and phagosome-regulated protein Apr AB (69.3%), and Chorismate mutase I were absent in the Ecuadorian isolates which would suggest possible affectation in adaptation and persistence of Mtb , due ESAT-6-like protein EsxS is related with the modulation of host immune responses , ; the Apr ABC locus modulates pH-driven adaptation to the macrophage phagosome , , and Chorismate mutase I is involved in inhibiting intrinsic apoptotic cell death of macrophages, playing a key role in the pathogenesis of TB , . Utilizing WGS in spanning network analysis enables researchers to connect TB transmission events , , providing insights into the spatial and temporal dynamics of TB transmission and identifying individuals and locations that play a critical role , . A recent population-based sequencing approach realizes a critical analysis of the utility of a pairwise distance threshold of < 12 SNPs and suggests that in scenarios with higher transmission rates, it is necessary to comprehend long-term transmission dynamics, adhering to a strict transmission 12 SNPs threshold is not advisable ; however, contrary to several studies supporting the utility of pairwise distance threshold of < 12 SNPs to identify recent transmission events – . We examined the possible transmission networks between Mtb isolates from Ecuador and from other 19 countries. Evidence suggests the potential spread of Ecuadorian Mtb isolates to individuals in different countries, based on the most significant clustering proportion confined to specific geographical locations. Combining genetic and epidemiological data could facilitate TB transmission management, particularly in migrant communities where socio-epidemiological changes due to migration may increase transmission complexity . Migration fluxes to and from high-burden countries significantly influence TB incidence, potentially leading to disease reactivation , . TB case clusters may involve autochthonous, mixed multinational, or cases among foreign-born individuals concentrated in a specific country , . Furthermore, Mtb can exhibit clonal transmission between hosts and establish clonal infections within a single host, with limited genetic diversification during infection or reactivation – . Our study observed clonality among Ecuadorian isolates with Colombian or Latin American migrants in Spain, indicating a potential transmission route involving direct contact between migrants, particularly in shared workplaces, such as plantations, factories, or restaurants. Surprisingly, three isolates corresponding to female patients were found in both scenarios, indicating that there might have been a relationship among the individuals, which might have been the cause of disease transmission. These findings suggest possible transnational transmission events involving Ecuador and its border countries and frequent migration destinations, highlighting the need to strengthen disease surveillance to reduce the possibility of more dangerous strains of tuberculosis entering the country , . Genome comparisons offer valuable insights into the molecular mechanisms that bacteria employ to survive and multiply within intracellular or extracellular host environments and to induce lesions and diseases . However, our understanding of the virulence factors expressed by Mtb is limited, and genetic variations resulting from selection pressure may affect their expression. Despite these challenges, performing this type of analysis may contribute to our understanding of how these factors function under local conditions . Our study identified 303 SNPs located within 103 genomic regions associated with virulence, with particular emphasis on genes such as plcA, plcB, pks7, pks12, phoR , and PPE46 which displayed the highest number of SNPs. These genes play pivotal roles in lipid metabolism, which is crucial for Mtb virulence of Mtb and is an integral component of the complex mycobacterial cell envelope , . Variations in these genes could affect their functionality, affecting the ability of the bacterium to evade host immune defense, induce necrosis in macrophages, and modulate virulence, particularly the PhoP-PhoR two-component system, potentially contributing to the pathogenic capabilities of the bacterium , . Phylodynamic analysis is useful for understanding TB strain variation and dynamics in various countries . The TMCRA and many TGC datasets suggest that TB isolates sampled in Ecuador can trace their ancestry back hundreds of years, indicating a complex evolutionary situation and implying the possibility that different TB isolates have caused infections at various time intervals from varied sources. The Ecuadorian Mtb data reported frequencies of 4.3.2/3 and 4.1.1, with TMRCA estimated to range from 450 to 470 YBP. These data indicate the possible ancient roots of the isolates. Nonetheless, it should be mentioned that these calculations only depict TB ancestral isolates within each sub-lineage, and the origin of TB in Ecuador could have been even earlier, before the arrival of Europeans, indicating a long history of human-pathogen co-evolution in the region, consistent with that reported in the ancient Andean population . While some isolates appear to have signs of long-living existence and evolution, and thus historical transmission, others seem to have appeared more recently, suggesting current transmission. Genetic diversity among Ecuadorian Mtb isolates is likely to arise from transcontinental interactions, human migration, and various factors that promote the circulation of local and global TB isolates. Thus, additional studies integrating genomic data from various geographic sites as well as comprehensive epidemiological information could help trace the origin and dissemination routes of TB isolates in Ecuador more accurately. Although our study offers valuable insights, it had several limitations. The primary limitation is the modest sample size of 88 Mtb isolates from cultured samples, a small fraction compared to the total number of TB cases reported in Ecuador in 2021 (5595). Furthermore, the lack of comprehensive epidemiological information, including contact tracing details, restricted our analysis of transmission dynamics. To enhance our understanding, future research should aim to combine genomic data with additional epidemiological details to uncover potential Mtb transmission pathways for Mtb . Moreover, our sequencing focused on cultured sputum isolates, a standard approach in Mtb genomic epidemiology that may only partially capture the complete spectrum of mycobacteria in the lungs, thus limiting our ability to capture the intricacies of within-host variations in individual infections Longitudinal studies have the potential to enhance sequencing analyses and to uncover variations in resistance genes and virulence factors, thereby assisting in refining treatment strategies. These insights hold promise for alleviating the global TB burden. To the best of our knowledge, this is the first study to establish local Mtb transmission networks in Ecuador using whole-genome analysis. Our findings reinforce and contribute to the knowledge of transmission networks previously characterized in Ecuador based on the MIRU-VNTR approach , , , . Our study provides valuable insights into the genomic and epidemiological characteristics of Mtb isolates from Ecuador. Additionally, our analysis identified drug-resistant isolates and transmission events between individuals and across borders, underscoring the need for more extensive whole-genome sequencing and network analyses to guide public health interventions.
Genome database of the Mtb samples This study analyzed raw reads from the genomes of 88 clinical Mtb isolates of the L4 lineage, sequenced in a previous study , corresponding to BioProject PRJNA827129. These isolates were collected conveniently between 2019 and 2021 from private laboratories and the National Reference of Mycobacteria at the National Institute of Public Health Research "Leopoldo Izquieta Pérez" (INSPI-LIP) across the different provinces in Ecuador. A significant proportion of the samples came from Guayaquil, accounting for 81.8% of the isolates. This city represents epidemiologically more than half of the TB cases in the country . The remaining samples were collected from Babahoyo (5.6%), El Empalme (3.4%), and Quito (2.3%), highlighting the geographical spread and prevalence of TB in these regions. A small number of isolates came from Chone, Duran, Guaranda, Machala, and Nueva Loja, accounting for 1.1% of the total isolates. Additionally, the study included 415 publicly available sequences previously characterized as Mtb isolates of the L4 lineage from 19 countries identified as significant in the Ecuadorian context of continental and transoceanic migration. Among the continental countries, the study included isolates from Argentina (n = 18), Brazil (n = 84), Canada (n = 41), Colombia (n = 8), Guatemala (n = 16), Mexico (n = 35), Panama (n = 4), Paraguay (n = 67), Peru (n = 44), and the USA (n = 6). Isolates from Hungary (n = 5), the Netherlands (n = 12), Portugal (n = 8), Spain (n = 59), and the United Kingdom (n = 5) were included. Notably, among the 59 Spanish isolates, some were identified previously from Latin American migrants who had settled in Spain years earlier, including individuals from Bolivia (n = 20), Colombia (n = 8), Ecuador (n = 8), Chile (n = 1), and Honduras (n = 1), as detailed in reference . The accession numbers and distributions of these countries are listed in Supplementary Table , . All protocols used in this article were approved by The University Espiritu Santo Review Board under code 2022-001A. Genome assembly and annotation To ensure high-quality genomic data, raw reads from the 503 Mtb genomes were initially processed with rigorous quality control using Fastp v0.23.4 , and Kraken v2 for species confirmation and contamination screening, ensuring that only Mtb -specific reads were processed. Pangenome construction Pangenomic analysis was carried out using the Panaroo pipeline from the GFF archives annotated by Prokka v1.14.16 using the H37Rv (NC_000962.3) M. tuberculosis reference genome and default parameters for clustering to define gene families and identify core genes present in 99% of the isolates. The pangenome was divided into core, accessory, and unique genes, which were categorized based on their presence in isolates. Variant calling analysis The cleaned reads were processed using the MTBseq pipeline with standard input parameters to map the reads to the Mtb H37Rv reference genome (NC_000962.3). Briefly, this tool involves BWA-mem and SAMtools for mapping, GATK v3 for base call recalibration, and realigning reads around insertions and deletions (InDels), followed by SAMtools mpileup for Single Nucleotide Polymorphisms (SNPs) and InDel calling (using parameters B and d 1000). Only high-quality genomes were processed after accomplishing the following criteria: mean coverage greater than 20x, read depth less than 5x, and reference genome coverage of > 95%. MTBseq analysis facilitated sub-lineage classification and the construction of a genetic distance matrix to identify transmission groups. Virulence analysis Ecuadorian Mtb isolates were analyzed to identify mutations in the genes involved in host adaptability and virulence. Genomes from the isolates were assembled from high-quality reads using the Unicycler assembly pipeline with a minimum contig size of 300 bp and polished using Pilon . SNPs within high-confidence and repetitive genomic regions, such as the PE/PPE gene family, were confirmed via visualization using Integrated Genome Viewer , . Virulence-related proteins were analyzed using the Virulence Factor Database (VFDB) in the Pathosystems Resource Integration Center (PATRIC) online ( https://www.patricbrc.org/ ). Drug resistance prediction and lineage classification The TB-Profiler v4 pipeline uses the BCF tool for variant calling and predicts drug resistance-associated mutations. These include resistance to first-line drugs such as rifampicin (RIF), isoniazid (INH), pyrazinamide (PZA), and ethambutol (EMB), and second-line drugs such as fluoroquinolones (FQs), streptomycin (STR), ethionamide (ETH), and aminoglycosides (including second-line drugs, amikacin, kanamycin, and capreomycin). According to World Health Organization guidelines, isolates exhibiting resistance solely to INH are classified as isoniazid-resistant TB (HR-TB), those resistant to RIF as rifampicin-resistant TB (RR-TB), and those with resistance to both INH and RIF as multidrug-resistant TB (MDR-TB). Isolates that demonstrated resistance to any FQ in addition to MDR or RR status were categorized as pre-extensively drug-resistant TB (pre-XDR-TB). Phylogenetic analysis and population structure of Mtb isolates Phylogenetic reconstruction was conducted using concatenated SNPs aligned genomically using the MTBseq tool. Repetitive regions (PPE/PE-PGRS genes), consecutive indels, and genes involved in antibiotic resistance were excluded from further phylogenetic analysis. The optimal substitution model for this SNP alignment was determined using ModelTest-NG v0.1.7 . Following this, the phylogenetic tree was reconstructed using the maximum-likelihood (ML) method via RAxML-NG , utilizing the general time-reversible model with gamma-distributed rate heterogeneity (GTR + GAMMA) and supported by 1000 bootstrap replicates. To address the potential ascertainment bias resulting from the exclusive use of polymorphic sites and consequent rescaling of tree branches, ascertainment bias correction was applied by specifying the number of invariant sites , as outlined at https://github.com/conmeehan/pathophy . The phylogenetic tree was visualized using the Interactive Tree Of Life (iTOL) v6.6 . The genome of Mycobacterium microti genome (accession number: SRR3647357) was used as an outgroup to root the tree. Additionally, a population structure analysis was conducted using the rhierBAPS package , which was set to a maximum depth of two and n.pops of 20. This analysis employed a hierarchical nested clustering approach based on genetic data, specifically prioritizing SNP loci that displayed a minor allele in at least two sequences, to effectively identify subpopulations or clusters. Transmission cluster analysis and modeling of the genetic clustering network Transmission genomic clusters (TGCs) were identified using concatenated high-quality SNPs processed using MTBseq. We applied a pairwise distance threshold of 12 SNPs between isolates to achieve optimal population clustering, a standard established in previous genomic TB studies – , . TGCs were categorized based on their size into small (fewer than three isolates), medium (three to five isolates), and large (more than five isolates). Visualization of the resulting SNP alignment for each cluster was used to infer a genetic network. We used a parsimony-based algorithm for network reconstruction using PopART software because of the monomorphic and non-recombining behavior of Mtb, as well as the potential of the sample dataset, including the original genotype. We compared the distribution network of isolates using a median-spanning network (MSN) and median-joining network (MJN). For clusters of at least three samples, we utilized TransFlow to reconstruct the transmission network, enhancing our understanding of local transmission dynamics. TransFlow integrates genomic data with epidemiological factors, such as sampling dates and geographic coordinates, to map spatial connectivity among isolates. Additionally, to mitigate the bias arising from the use of lineage-specific reference genomes, TransFlow incorporates the PANPASCO pipeline . This pipeline employs a computational pan-genome consisting of 146 complete MTBC genomes from major lineages 1–4, facilitating accurate pairwise SNP distance calculations. This methodology ensures a thorough and representative analysis of genetic variation across MTBC populations. Phylodynamics of TB transmission clusters To study the temporal dynamics of the different sub-lineages identified in the Ecuadorian Mtb L4 isolates, with a particular focus on uncovering historical introduction events, a time-calibrated phylogeny was inferred using BEAST v1.10.4 , utilizing collection and tip dates from the isolates. The XML input file necessary for analysis was generated by concatenating SNPs derived from MTBseq and processed using BEAUTi. This file was adjusted to specify the number of invariant sites following guidance from the BEAST User Forum. To assess the temporal signal of the sequence alignments, we used TempEst v1.5.3 . The dating analysis employed the general time-reversible plus gamma distribution (GTR + GAMMA) substitution model coupled with a strict molecular clock and coalescent constant-size demographic model. Markov chain Monte Carlo (MCMC) simulations were conducted for 250 million iterations, with a 10% burn-in phase and samples taken every 10,000 generations. This approach facilitates independent evaluation of chain convergence. The analysis results were summarized and convergence was confirmed using Tracer v1.6 , ensuring that all essential parameters achieved an effective sample size (ESS) of over 200. The Maximum Clade Credibility (MCC) tree was then computed using TreeAnnotator v2.5.0, providing a statistically supported phylogenetic tree with time-calibrated estimates.
This study analyzed raw reads from the genomes of 88 clinical Mtb isolates of the L4 lineage, sequenced in a previous study , corresponding to BioProject PRJNA827129. These isolates were collected conveniently between 2019 and 2021 from private laboratories and the National Reference of Mycobacteria at the National Institute of Public Health Research "Leopoldo Izquieta Pérez" (INSPI-LIP) across the different provinces in Ecuador. A significant proportion of the samples came from Guayaquil, accounting for 81.8% of the isolates. This city represents epidemiologically more than half of the TB cases in the country . The remaining samples were collected from Babahoyo (5.6%), El Empalme (3.4%), and Quito (2.3%), highlighting the geographical spread and prevalence of TB in these regions. A small number of isolates came from Chone, Duran, Guaranda, Machala, and Nueva Loja, accounting for 1.1% of the total isolates. Additionally, the study included 415 publicly available sequences previously characterized as Mtb isolates of the L4 lineage from 19 countries identified as significant in the Ecuadorian context of continental and transoceanic migration. Among the continental countries, the study included isolates from Argentina (n = 18), Brazil (n = 84), Canada (n = 41), Colombia (n = 8), Guatemala (n = 16), Mexico (n = 35), Panama (n = 4), Paraguay (n = 67), Peru (n = 44), and the USA (n = 6). Isolates from Hungary (n = 5), the Netherlands (n = 12), Portugal (n = 8), Spain (n = 59), and the United Kingdom (n = 5) were included. Notably, among the 59 Spanish isolates, some were identified previously from Latin American migrants who had settled in Spain years earlier, including individuals from Bolivia (n = 20), Colombia (n = 8), Ecuador (n = 8), Chile (n = 1), and Honduras (n = 1), as detailed in reference . The accession numbers and distributions of these countries are listed in Supplementary Table , . All protocols used in this article were approved by The University Espiritu Santo Review Board under code 2022-001A.
To ensure high-quality genomic data, raw reads from the 503 Mtb genomes were initially processed with rigorous quality control using Fastp v0.23.4 , and Kraken v2 for species confirmation and contamination screening, ensuring that only Mtb -specific reads were processed.
Pangenomic analysis was carried out using the Panaroo pipeline from the GFF archives annotated by Prokka v1.14.16 using the H37Rv (NC_000962.3) M. tuberculosis reference genome and default parameters for clustering to define gene families and identify core genes present in 99% of the isolates. The pangenome was divided into core, accessory, and unique genes, which were categorized based on their presence in isolates.
The cleaned reads were processed using the MTBseq pipeline with standard input parameters to map the reads to the Mtb H37Rv reference genome (NC_000962.3). Briefly, this tool involves BWA-mem and SAMtools for mapping, GATK v3 for base call recalibration, and realigning reads around insertions and deletions (InDels), followed by SAMtools mpileup for Single Nucleotide Polymorphisms (SNPs) and InDel calling (using parameters B and d 1000). Only high-quality genomes were processed after accomplishing the following criteria: mean coverage greater than 20x, read depth less than 5x, and reference genome coverage of > 95%. MTBseq analysis facilitated sub-lineage classification and the construction of a genetic distance matrix to identify transmission groups.
Ecuadorian Mtb isolates were analyzed to identify mutations in the genes involved in host adaptability and virulence. Genomes from the isolates were assembled from high-quality reads using the Unicycler assembly pipeline with a minimum contig size of 300 bp and polished using Pilon . SNPs within high-confidence and repetitive genomic regions, such as the PE/PPE gene family, were confirmed via visualization using Integrated Genome Viewer , . Virulence-related proteins were analyzed using the Virulence Factor Database (VFDB) in the Pathosystems Resource Integration Center (PATRIC) online ( https://www.patricbrc.org/ ).
The TB-Profiler v4 pipeline uses the BCF tool for variant calling and predicts drug resistance-associated mutations. These include resistance to first-line drugs such as rifampicin (RIF), isoniazid (INH), pyrazinamide (PZA), and ethambutol (EMB), and second-line drugs such as fluoroquinolones (FQs), streptomycin (STR), ethionamide (ETH), and aminoglycosides (including second-line drugs, amikacin, kanamycin, and capreomycin). According to World Health Organization guidelines, isolates exhibiting resistance solely to INH are classified as isoniazid-resistant TB (HR-TB), those resistant to RIF as rifampicin-resistant TB (RR-TB), and those with resistance to both INH and RIF as multidrug-resistant TB (MDR-TB). Isolates that demonstrated resistance to any FQ in addition to MDR or RR status were categorized as pre-extensively drug-resistant TB (pre-XDR-TB).
Phylogenetic reconstruction was conducted using concatenated SNPs aligned genomically using the MTBseq tool. Repetitive regions (PPE/PE-PGRS genes), consecutive indels, and genes involved in antibiotic resistance were excluded from further phylogenetic analysis. The optimal substitution model for this SNP alignment was determined using ModelTest-NG v0.1.7 . Following this, the phylogenetic tree was reconstructed using the maximum-likelihood (ML) method via RAxML-NG , utilizing the general time-reversible model with gamma-distributed rate heterogeneity (GTR + GAMMA) and supported by 1000 bootstrap replicates. To address the potential ascertainment bias resulting from the exclusive use of polymorphic sites and consequent rescaling of tree branches, ascertainment bias correction was applied by specifying the number of invariant sites , as outlined at https://github.com/conmeehan/pathophy . The phylogenetic tree was visualized using the Interactive Tree Of Life (iTOL) v6.6 . The genome of Mycobacterium microti genome (accession number: SRR3647357) was used as an outgroup to root the tree. Additionally, a population structure analysis was conducted using the rhierBAPS package , which was set to a maximum depth of two and n.pops of 20. This analysis employed a hierarchical nested clustering approach based on genetic data, specifically prioritizing SNP loci that displayed a minor allele in at least two sequences, to effectively identify subpopulations or clusters.
Transmission genomic clusters (TGCs) were identified using concatenated high-quality SNPs processed using MTBseq. We applied a pairwise distance threshold of 12 SNPs between isolates to achieve optimal population clustering, a standard established in previous genomic TB studies – , . TGCs were categorized based on their size into small (fewer than three isolates), medium (three to five isolates), and large (more than five isolates). Visualization of the resulting SNP alignment for each cluster was used to infer a genetic network. We used a parsimony-based algorithm for network reconstruction using PopART software because of the monomorphic and non-recombining behavior of Mtb, as well as the potential of the sample dataset, including the original genotype. We compared the distribution network of isolates using a median-spanning network (MSN) and median-joining network (MJN). For clusters of at least three samples, we utilized TransFlow to reconstruct the transmission network, enhancing our understanding of local transmission dynamics. TransFlow integrates genomic data with epidemiological factors, such as sampling dates and geographic coordinates, to map spatial connectivity among isolates. Additionally, to mitigate the bias arising from the use of lineage-specific reference genomes, TransFlow incorporates the PANPASCO pipeline . This pipeline employs a computational pan-genome consisting of 146 complete MTBC genomes from major lineages 1–4, facilitating accurate pairwise SNP distance calculations. This methodology ensures a thorough and representative analysis of genetic variation across MTBC populations.
To study the temporal dynamics of the different sub-lineages identified in the Ecuadorian Mtb L4 isolates, with a particular focus on uncovering historical introduction events, a time-calibrated phylogeny was inferred using BEAST v1.10.4 , utilizing collection and tip dates from the isolates. The XML input file necessary for analysis was generated by concatenating SNPs derived from MTBseq and processed using BEAUTi. This file was adjusted to specify the number of invariant sites following guidance from the BEAST User Forum. To assess the temporal signal of the sequence alignments, we used TempEst v1.5.3 . The dating analysis employed the general time-reversible plus gamma distribution (GTR + GAMMA) substitution model coupled with a strict molecular clock and coalescent constant-size demographic model. Markov chain Monte Carlo (MCMC) simulations were conducted for 250 million iterations, with a 10% burn-in phase and samples taken every 10,000 generations. This approach facilitates independent evaluation of chain convergence. The analysis results were summarized and convergence was confirmed using Tracer v1.6 , ensuring that all essential parameters achieved an effective sample size (ESS) of over 200. The Maximum Clade Credibility (MCC) tree was then computed using TreeAnnotator v2.5.0, providing a statistically supported phylogenetic tree with time-calibrated estimates.
Supplementary Information 1. Supplementary Information 2.
|
Mastering your fellowship: Part 3, 2023 | 98eae18c-ade3-4421-b256-310566f218d5 | 10244956 | Family Medicine[mh] | This section in the South African Family Practice journal is aimed at helping registrars in preparing for the Fellowship of the College of Family Physicians of South Africa (FCFP [SA]) Final Part A examination and will provide examples of the question formats encountered in the written examination: multiple choice question (MCQ) in the form of single best answer (SBA – Type A) and extended matching question (EMQ – Type R); short answer question (SAQ), questions based on the Critical Reading of a Journal article (CRJ: evidence-based medicine) and an example of an objectively structured clinical examination (OSCE) question. Each of these question types is presented based on the College of Family Physicians blueprint and the key learning outcomes of the FCFP (SA) programme. The MCQs are based on the 10 clinical domains of family medicine, the SAQs are aligned with the five national unit standards and the critical reading section will include evidence-based medicine and primary care research methods. This month’s edition is based on unit standard one (Effectively manage themselves, their team and their practice, in any sector, with visionary leadership and self-awareness, to ensure the provision of high-quality, evidence-based care), unit standard two (Evaluate and manage patients with both undifferentiated and more specific problems cost-effectively according to the bio-psycho-social approach), unit standard three (Improve the health and quality of life of the community) and unit standard five (Conduct all aspects of healthcare in an ethical and professional manner). The clinical domain covered in this edition is eye health. We suggest that you attempt to answer the questions (by yourself or with peers or supervisors), before finding the model answers online: http://www.safpj.co.za/ . Please visit the Colleges of Medicine website for guidelines on the Fellowship examination: https://www.cmsa.co.za/view_exam.aspx?QualificationID=9 . We are keen to hear about how this series is assisting registrars and their supervisors in preparing for the FCFP (SA) examination. Please email ([email protected]) us with your feedback and suggestions.
A 65-year-old male presented with a red eye associated with a clear watery discharge, itchiness and pain for the last 2 days. Only one eye is affected. He reports mild blurring of vision. An image of the eye is shown in . What would you prescribe as the most appropriate next step? Acyclovir ophthalmic ointment Chloromycetin ophthalmic ointment Oxymetazoline eye drops Reassurance and advice on cold compressors Steroid eye drops Answer: ( a ) Discussion The red eye is a frequent presentation to primary care, so it is essential to distinguish potentially serious eye conditions from less severe causes of acute red eye. Common causes include conjunctivitis, keratitis, uveitis, episcleritis, scleritis, acute angle closure glaucoma, a foreign body and subconjunctival haemorrhage. In assessing the red eye, one needs to elucidate if the following symptoms and signs are present: Is there pain and irritation? Is there a discharge? If yes, describe the type of discharge. Also, assess whether the discharge is present unilateral or bilateral. Is photophobia present? Is there associated visual loss? What is the severity of the red eye? Are there pupillary changes? Describe the distribution of redness. Is it diffuse, localised, or is it a ciliary flush? Conjunctivitis is an inflammation of the conjunctiva that is usually bilateral, has variable pain and irritation and often has diffuse redness. The discharge is common and can be watery, mucoid, purulent or mucopurulent. Allergic conjunctivitis may be related to seasonal changes, that is, spring catarrh. The discharge is often mucoid or watery. There may be palpebral follicles or papillae present. The inflammation is prominently perilimbal. Treatment is with antihistamine, cromoglycate or steroid drops if very severe. Viral conjunctivitis is often associated with a profuse watery discharge and is very common. Keratitis and tarsal follicles may be present. It may be associated with pre-auricular lymphadenopathy. Adenovirus infection may present with epidemic keratoconjunctivitis and pharyngo-conjunctival fever. Treatment for this is supportive management. A more severe form of viral conjunctivitis is because of the herpes simplex virus. Infection is often because of primary infection. Infection is usually unilateral and is associated with pain and blurring of vision. If such symptoms are present, fluorescein staining is advised. Illuminate the cornea with an ophthalmoscope with a cobalt blue filter, with the brightest setting. Punctate stains or linear branching (dendritic) ulcers may be seen in herpes simplex keratoconjunctivitis. A referral is advised if the corneal lesion is not clean or clear, or has whitish areas within the epithelial scar bed; if the corneal staining area is not smaller within 24 h; if there is a history of recurrence; and, if you are worried about disciform keratitis. Bacterial conjunctivitis is associated with a mucopurulent discharge and may be because of Staphylococcus epidermidis, Staphylococcus aureus, Haemophilus influenzae, Streptococcus spp. and Pseudomonas aeruginosa (contact lens wearers) and Neisseria gonorrhoea . The treatment involved the frequent instillation of antibiotic drops, for example, chloramphenicol. For N. gonorrhoea , intramuscular ceftriaxone is also prescribed. Chlamydial infection causes adult inclusion conjunctivitis and is because of serotypes D to K. This infection presents with conjunctival follicles and lid swelling. Treatment is with topical and systemic tetracycline. The unilateral nature of the case presented with the blurring of vision should alert the clinician of the possibility of herpes conjunctivitis and keratitis. Then fluorescein staining and prompt initiation of the definite treatment are mandatory. Further reading Pons J. Chapter 80: How to examine the eye. In: Mash B, Brits H, Naidoo M, Ras T, editors. South African family practice manual. 4th ed. Braamfontein: Van Schaik; 2023. Pons J. Chapter 81: How to treat the eye. In: Mash B, Brits H, Naidoo M, Ras T, editors. South African family practice manual. 4th ed. Braamfontein: Van Schaik; 2023. South African Department of Health. Hospital-level standard treatment guidelines and rssential medicines list. Pretoria: South African National Department of Health; 2019. Albrecht S. Conjunctivitis [homepage on the Internet]. New York, NY: Medscape; 2011 [cited 2023 Jan 23]. Available from: https://www.medscape.com/viewarticle/741938_1
The red eye is a frequent presentation to primary care, so it is essential to distinguish potentially serious eye conditions from less severe causes of acute red eye. Common causes include conjunctivitis, keratitis, uveitis, episcleritis, scleritis, acute angle closure glaucoma, a foreign body and subconjunctival haemorrhage. In assessing the red eye, one needs to elucidate if the following symptoms and signs are present: Is there pain and irritation? Is there a discharge? If yes, describe the type of discharge. Also, assess whether the discharge is present unilateral or bilateral. Is photophobia present? Is there associated visual loss? What is the severity of the red eye? Are there pupillary changes? Describe the distribution of redness. Is it diffuse, localised, or is it a ciliary flush? Conjunctivitis is an inflammation of the conjunctiva that is usually bilateral, has variable pain and irritation and often has diffuse redness. The discharge is common and can be watery, mucoid, purulent or mucopurulent. Allergic conjunctivitis may be related to seasonal changes, that is, spring catarrh. The discharge is often mucoid or watery. There may be palpebral follicles or papillae present. The inflammation is prominently perilimbal. Treatment is with antihistamine, cromoglycate or steroid drops if very severe. Viral conjunctivitis is often associated with a profuse watery discharge and is very common. Keratitis and tarsal follicles may be present. It may be associated with pre-auricular lymphadenopathy. Adenovirus infection may present with epidemic keratoconjunctivitis and pharyngo-conjunctival fever. Treatment for this is supportive management. A more severe form of viral conjunctivitis is because of the herpes simplex virus. Infection is often because of primary infection. Infection is usually unilateral and is associated with pain and blurring of vision. If such symptoms are present, fluorescein staining is advised. Illuminate the cornea with an ophthalmoscope with a cobalt blue filter, with the brightest setting. Punctate stains or linear branching (dendritic) ulcers may be seen in herpes simplex keratoconjunctivitis. A referral is advised if the corneal lesion is not clean or clear, or has whitish areas within the epithelial scar bed; if the corneal staining area is not smaller within 24 h; if there is a history of recurrence; and, if you are worried about disciform keratitis. Bacterial conjunctivitis is associated with a mucopurulent discharge and may be because of Staphylococcus epidermidis, Staphylococcus aureus, Haemophilus influenzae, Streptococcus spp. and Pseudomonas aeruginosa (contact lens wearers) and Neisseria gonorrhoea . The treatment involved the frequent instillation of antibiotic drops, for example, chloramphenicol. For N. gonorrhoea , intramuscular ceftriaxone is also prescribed. Chlamydial infection causes adult inclusion conjunctivitis and is because of serotypes D to K. This infection presents with conjunctival follicles and lid swelling. Treatment is with topical and systemic tetracycline. The unilateral nature of the case presented with the blurring of vision should alert the clinician of the possibility of herpes conjunctivitis and keratitis. Then fluorescein staining and prompt initiation of the definite treatment are mandatory. Further reading Pons J. Chapter 80: How to examine the eye. In: Mash B, Brits H, Naidoo M, Ras T, editors. South African family practice manual. 4th ed. Braamfontein: Van Schaik; 2023. Pons J. Chapter 81: How to treat the eye. In: Mash B, Brits H, Naidoo M, Ras T, editors. South African family practice manual. 4th ed. Braamfontein: Van Schaik; 2023. South African Department of Health. Hospital-level standard treatment guidelines and rssential medicines list. Pretoria: South African National Department of Health; 2019. Albrecht S. Conjunctivitis [homepage on the Internet]. New York, NY: Medscape; 2011 [cited 2023 Jan 23]. Available from: https://www.medscape.com/viewarticle/741938_1
This question was previously used in a FCFP(SA) written article. You are the family physician working in a community health centre (CHC) in South Africa. Disability because of cataracts seems to be increasing in the community. The nursing staff suggest that the CHC assists the community in addressing this disability. List five questions you would ask to decide if this is a priority in the community. (5 marks) How would you go about establishing answers to three of the questions that you listed in question 1 above? (3 marks) If you decide that visual loss is a priority in this community, what primary and secondary prevention strategies could you implement in the community? (5 marks) If you decide that visual loss is a priority in this community, what facility-based primary and secondary prevention interventions could you implement? (5 marks) Who could you approach as potential partners in addressing this disability burden for the community? (2 marks) Total: 20 marks Model answers 1. List five questions you would ask to decide if this is a priority in the community. (5 marks) ( The approach to this question requires the answer in the following points for prioritisation as per the Family Practice Manual ) How common is the problem of cataracts? How serious is the problem? To what extent is the community concerned about it? Is it feasible to intervene? Will an intervention be effective ? 2. How would you go about establishing answers to three of the questions that you listed in question 1 above? (3 marks) One point for one mark under any of the three questions chosen as listed below: How common is the problem of cataracts?: ■ Suggest a measure of prevalence such as the percentage of people seen at the CHC diagnosed with cataracts in the last year. Establish how many of those diagnosed with cataracts also have a significant visual disability as per visual acuity scores (and possibly the reasons given for cataracts). ■ Use ward-based primary health care (PHC) outreach teams while doing their house visits to identify those with vision loss that have not yet sought help at the CHC. ■ Start looking at incidence: the number of new cases developing cataracts per year. How serious is the problem? (Weighed against other health issues faced by the community): ■ Impact of the disability on the community and families if no intervention is taken now. ■ Backlog: The number of cataract cases identified and not operated on at present in the community (Backlog + received surgery = burden of cataract). To what extent is the community concerned about it?: ■ Does the community feel the same concern the nursing staff have observed? Explore what the community feels, thinks and does regarding this problem and if they see it as a priority area for intervention. Discuss with community elders, leaders, and community health workers (CHWs). Assess and identify to address the barriers (distance to hospital, fear of surgery, long waiting list, insufficient manpower, materials) to cataract surgery with the community. Is it feasible to intervene?: ■ Is it possible to come up with interventions as a multidisciplinary team with the community, health facilities, ophthalmologists, optometrists and outside partners (alternative healers or traditional healers) to address the issue of cataracts? Do the community leaders buy into the proposed intervention plans? Will an intervention be effective?: ■ Can the CHC be able to come up with realistic strategies with the resources available locally, with community support and help from networking with other resources available? Discuss strategies for monitoring if the intervention is working in the future and establish a cost-effective plan that is agreeable to all the stakeholders. ■ Identify modifiable risk factors associated with cataracts among the patients seen in the CHC (diabetes, smoking, chronic steroid use, chronic ultraviolet exposure). Train staff to screen such patients for cataracts. 3. If you decide that visual loss is a priority in this community, what primary and secondary prevention strategies could you implement in the community? (5 marks) One mark for any of the five points given below Health education in the community may consist of primary prevention (preventing cataracts from forming) and secondary prevention (preventing progression to cataract blindness in patients with cataracts): Motivate people to have regular eye checks after the age of 40 years and seek help with changes in vision at the local CHC. Monitor the changes in vision in family members by other family members (e.g. change in baseline while reading, watching television, etc.) and seek help early. Health education strategies in the community should be orientated to modify risk factors for conditions associated with visual loss. This can be in partnership with any other services providing eye care in the community such as the private sector, non-governmental organisations (NGOs) and CHWs. Diabetes: Tight control of blood sugar at home and compliance to regular planned visits to CHC. Ultraviolet exposure: Use sunglasses and hats when there is prolonged exposure. Awareness campaigns: Eye care education delivered in homes, schools and other community settings. Integrate eye-health preservation in health promotion agendas carried out by the CHWs in the community. 4. If you decide that visual loss is a priority in this community, what facility-based primary and secondary prevention interventions could you implement? (5 marks) One mark for any of the five points given below. Facility-based activities: Improving accessibility to services to assess visual loss with the initiation of yearly screening days in partnership with optometrists, NGOs or ophthalmologists from the tertiary hospitals. Early detection of cataracts and a local standard operating procedure to do eye checks as part of routine examinations for all above the age of 50 years when presenting to the CHC with any complaints. Early diagnosis and control of people with diabetes. All diabetic patients should have a fundoscopy examination by trained personnel at least once a year. Continuing Professional Development (CPD) activity in the CHC to train staff on picking up and managing conditions that lead to visual loss. Regular measurement of data in the CHC, establishing a system of recording and monitoring the patients presenting with visual loss in the CHC. Consider patient-centred integrated eye care service in the CHC with emphasis on patient self-reporting: education pamphlets in the local languages that can be distributed to the patients on when/why to report to CHC with regard to visual loss. Engage with the clinic managers to make essential medicines and medical supplies available within health systems for eye care. 5. Who could you approach as potential partners in addressing this disability burden for the community? (2 marks) One mark for any of the two points given below Regional hospital ophthalmological services, for example, ‘cataract blitzes’. Non-governmental organisations that would be able to provide corrective glasses for those operated, cataract trains and camps that mobilise services into the community. International humanitarian organisations, for example, Christoffel Blinden Mission. Media including social media can help spread awareness of the availability of services to the public. Further reading Buso D, Reid S. Chapter 152: How to make a community diagnosis and prioritise health issues. In: Mash B, Blitz J, editors. South African family practice manual. 3rd ed. Pretoria: Van Schaik, 2015; p. 499–500. Marcus TS. Community orientated primary care, level 2. Topic 1, Unit 1. Pearson, 2013; p. 10. National guideline on the prevention of blindness in South Africa. Department of Health Directorate: Chronic Diseases, Disabilities and Geriatrics; 2002. World report on vision [homepage on the Internet]. Geneva: World Health Organization; 2019 [cited 2023 Jan 22]. Available from: https://www.who.int/publications/i/item/9789241516570 Mash R, Gaede B, Hugo JJ. The contribution of family physicians and primary care doctors to community-orientated primary care. S Afr Fam Pract. 2021;63(1):1–5. https://doi.org/10.4102/safp.v63i1.5281
Read the accompanying article carefully and then answer the following questions. As far as possible use your own words. Do not copy out chunks from the article. Be guided by the allocation of marks concerning the length of your responses. Tshivhase SE, Khoza LB, Tshitangano TG. Application of the information-motivation-behavioural skills model to strengthen eye care follow-up amongst glaucoma patients. Afr Vision Eye Health. 2021;80(1):8. https://doi.org/10.4102/aveh.v80i1.642 Total: 30 marks Did the study address a focused question? Discuss. (2 marks) Critically appraise the authors’ choice of study design to answer the research question. (4 marks) Critically appraise the authors’ choice of the conceptual framework used to underpin the study’s design and data interpretation. (3 marks) Critically appraise the sampling strategy. (5 marks) Critically appraise how well the authors describe the data collection process. (4 marks) Critically appraise the process followed to develop and validate the instrument used in this study. (3 marks) Critically appraise the analysis and presentation of study data. (3 marks) Use a structured approach (e.g. READER) to discuss the value of these findings to your practice. (6 marks) Model answers Did the study address a focused question? Discuss. (2 marks) In the abstract, the authors mentioned that they applied the information-motivation and behavioural skills model (IMBSM) in strengthening eye care follow-up among glaucoma patients in the Limpopo province of South Africa. In the introduction section, the authors expand on this aim by stating that they analysed a specific theoretical model for improving glaucoma patients’ knowledge, attitude and practices as a springboard for strategy development. Therefore, the study investigated how the application of the IMBSM can strengthen eye care follow-up in this patient group. One may conclude that the research question for this observational study is focused, as it describes the population, the risk factor(s), and the outcome of interest. The population studied is focused, as these are the glaucoma patients attending eye care services in Limpopo. They introduced a specific theoretical model or lens to understand adherence issues to follow-up (the risk factors, which may result in lapses in care continuity) and they had a certain outcome in mind, namely to strengthen eye care follow-up. Critically appraise the authors’ choice of study design to answer the research question. (4 marks) The authors use a descriptive cross-sectional observational study to answer the research question. When considering the strength and appropriateness of this study design, one may conclude that this design is appropriate as it can assess the prevalence of an outcome of interest such as non-adherence to glaucoma medication, as well as the prevalence of any factors associated with it. Although a qualitative exploratory study could also be used, this would not provide insights into the prevalence of factors but provide rich data on aspects of non-adherence that may not have been anticipated. Given the extensive literature on adherence to chronic medication and the factors associated with it, the authors were justified to presume that most of the categories of factors contributing to non-adherence were considered. Critically appraise the authors’ choice of the conceptual framework used to underpin the study’s design and data interpretation. (3 marks) A conceptual framework is a helpful way to illustrate the relationships between variables and concepts in a study and is often informed by the literature and any related hypotheses. In this study, the authors use the IMBSM model that is described and established in the literature to explain and connect the various categories of factors that contribute to adherence to chronic medication. The IMBSM model used in this study connects the prevention of non-adherence to three areas/categories, namely information and knowledge factors, motivation factors and behavioural skills factors. There are numerous models of understanding adherence, including the Health Belief Model or the World Health Organization’s Multidimensional Adherence Model. The latter model categorises patient-related factors, condition-related factors, social and economic factors, and therapy-related factors as dimensions of adherence. However, this study’s conceptual framework narrows the focus to knowledge, attitude and behavioural factors. While this may lead to the exclusion of other social, economic and health system factors, it is still consistent with the study’s overall aim of exploring areas of realistic intervention through information provision and behavioural modification. Nevertheless, the reader needs to be made aware of the complex mix of contextual factors that influence the individual’s ability to respond to ‘cues to action’; this should have been mentioned in the manuscript’s limitations section. Critically appraise the sampling strategy. (5 marks) The authors employed a non-probability convenience sampling strategy and described the sample size of 450 as adequate for achieving reliable results. While the practicalities of being unable to randomly sample all patients with glaucoma on treatment in the district are understandably difficult, a systematic random sample could have still been attempted among the groups of patients approached, who attended the hospital for care and met the criteria. The hospital is the only ophthalmology referral centre in the entire region. This means that almost all public patients diagnosed with glaucoma would have been in the hospital’s care. Hospital records might have been used to ascertain the total number of patients with glaucoma to assist with calculating a representative sample size. It may have been useful if the authors included a clearer description in the study setting of how the local health system is organised and how this facilitates patient access to the services at Elim Hospital. A description of the distances between facilities and whether the services consist of only follow-up at the hospital or whether there is an outreach service, as these affect follow-up. Importantly, the authors do not describe in more replicable detail, how they calculated the sample size. Typically, in purely descriptive cross-sectional studies, a reference would be made to similar studies where a particular prevalence of the outcome in question (e.g. percentage of patients with poor adherence) would have been used to calculate their study’s sample size. It is also notable that the authors choose to exclude patients who were not attending consistently for 3 years. This means that patients who were very poorly adherent were not included in the study, thereby contributing to selection bias. Critically appraise how well the authors describe the data collection process. (4 marks) The data collection process involved trained interviewers who provided the interviews. However, the authors indicate that many of the participants required assistance as they could not read or write and that the ‘researchers’ assisted them. It is not clear why the trained interviewers did not fulfil this role. Perhaps the authors implied that these interviewers were the researchers. Furthermore, it is also not clear how the interviewers were trained and supervised and how consent was taken. The professional background and language(s) of the interviewers were not described. Were these trained interviewers familiar with the community context and were they able to administer the instrument in the participants’ preferred or home language? Lastly, the authors do not describe where in the hospital the questionnaires were administered and whether confidentiality and anonymity were maintained or ensured. Critically appraise the process followed to develop and validate the instrument used in this study. (3 marks) The study’s process for developing and validating the instrument is described in fair detail. They describe the establishment of face validity through their piloting of the questionnaire on 45 patients or participants. They also describe that they ensured alignment of the instrument questions with the objectives and the literature, which include alignment with the conceptual framework (content validity). However, they do not describe the details of how they established content validity. These could include the description of how many panel experts were used to review the instrument. The instrument was developed in the English language and subsequently translated into the ‘local language’. Neither the process of translation was described nor the local language was defined. Typically, the process of translation requires a series of forward and back translations, to ensure that the data collection instrument is aligned appropriately with the language and cultural context. Critically appraise the analysis and presentation of study data. (3 marks) The study data are appropriately presented in tables using categories that are aligned with the conceptual framework. Participants’ responses are presented in proportions (percentages), which enables comparison. Unfortunately, not many inferences can be made from the findings as no inferential statistical analysis was conducted, nor were sample sizes calculated for such analyses. For example, no conclusions can be made about whether the patients with lower levels of knowledge of glaucoma demonstrated statistically significant lower levels of adherence. Some of the study data raise questions about the construct validity of the instrument. For example, it is not fully clear how the questions on patients’ beliefs about the cause of glaucoma revolved around only three possible answers (i.e. witchcraft, hereditary, and unknown). Use a structured approach (e.g. READER) to discuss the value of these findings to your practice. (6 marks) The READER format may be used to answer this question: Relevance to family medicine and primary care? Education – does it challenge existing knowledge or thinking? Applicability – are the results applicable to my practice? Discrimination – is the study scientifically valid enough? Evaluation – given the aforementioned, how would I score or evaluate the usefulness of this study to my practice? Reaction – what will I do with the study findings? The answer may be a subjective response but should be one that demonstrates a reflection on the possible changes within the student’s practice within the South African public healthcare system. It is acceptable for the student to suggest how their practice might change, within other scenarios after graduation (e.g. private general practice). The reflection on whether all important outcomes were considered is, therefore, dependent on the reader’s perspective (is there other information you would have liked to see?). A model answer could be written from the perspective of the family physician employed in the South African district health system: R: This study is relevant to the African primary care context because many patients with glaucoma are also followed up at primary care clinics and district hospitals while under the care of the regional hospital or tertiary ophthalmology service. E: While the challenge of adherence to chronic medication is not new, this study does reveal some local features that are arguably helpful to clinicians in this region. However, the challenges with representativity and the lack of more inferential statistical analysis limit what can be meaningfully drawn from the study. A: It is not possible to generalise the study’s findings to the wider South African setting, as the study was conducted in a specialist ophthalmology service using a non-probability sampling method. D: Regarding discrimination, the study has fundamental design flaws, bringing its validity into question. The lack of a description of how the sample size was calculated is problematic, which in turn affects what conclusions can be drawn even from the basic description of the statistical findings. Furthermore, those patients who are lost to follow-up are not represented at all and their reasons for non-adherence may be entirely different and unknown, which presents a selection bias to the findings that are not addressed. There is also an opportunity lost concerning applying inferential statistical methods. E: The topic and the findings are essential for informing any strategies for tailoring interventions that support adherence in this area and setting. Knowing the specific details of the major contributors to non-adherence to glaucoma medication in this group of patients is crucial. Given the design flaws in the study, this is not satisfactorily clear. R: Although the study has design and analytical flaws, the challenge of adherence to chronic medications, in general, is significant. These preliminary findings may be helpful for a follow-up explanatory study with a subset of these patients for richer details on adherence. More importantly, attempting to contact and follow-up with those who have not followed up with the service at all for such a study may reveal richer findings that could support intervention strategies for addressing and preventing non-adherence. Total: 30 marks Further reading Pather M. Evidence-based family medicine. In: Mash B, editor. Handbook of family medicine. 4th ed. Cape Town: Oxford University Press, 2017; p. 430–453. MacAuley D. READER: An acronym to aid critical reading by general practitioners. Br J Gen Pract. 1994;44(379):83–85. The Critical Appraisals Skills Programme (CASP). CASP checklists [homepage on the Internet]. 2023 [cited 2023 Feb 04]. Available from: https://casp-uk.net/casp-tools-checklists/ Goodyear-Smith F, Mash B, editors. How to do primary care research. Boca Raton, FL: CRC Press, Taylor and Francis Group; 2019.
Objective of station This station tests the candidate’s ability to manage a patient with new onset unilateral visual loss. Type of station Integrated consultation. Role player Simulated patient: Adult male or female. Instructions to the candidate You are the family physician working at a district hospital. Please consult with this patient, who presents to the emergency unit. Your task: Please consult with this patient and develop a comprehensive management plan. You do not need to do an examination on this patient. All examination findings will be provided on request. Instructions for the examiner This is an integrated consultation station in which the candidate has 15 min. Familiarise yourself with the assessor guidelines, which detail the required responses expected from the candidate. No marks are allocated. In the mark sheet (Table 1), tick off one of the three responses for each of the competencies listed. Make sure you are clear on what the criteria are for judging a candidate’s competence in each area. Provide the following information to the candidate when requested: examination findings: ■ Please switch off your cell phone. ■ Please do not prompt the student. ■ Please ensure that the station remains tidy and is reset between candidates. Guidance for examiners regarding The aim is to establish that the candidate has an effective and safe approach to managing acute persistent visual loss in an adult. Working definition of competent performance: The candidate effectively completes the task within the allotted time, in a manner that maintains patient safety, even though the execution may not be efficient and well structured: ■ Not competent : Patient safety is compromised (including ethical-legally) or the task is not completed. ■ Competent : The task is completed safely and effectively. ■ Good : In addition to displaying competence, the task is completed efficiently and in an empathic, patient-centred manner (acknowledges patient’s ideas, beliefs, expectations, concerns/fears). Establishes and maintains a good clinician–patient relationship. ■ The competent candidate is respectful and engages with the patient in a dignified manner. ■ The good candidate is empathic, compassionate and collaborative, facilitating patient participation in key areas of the consultation. Gathering information: ■ The competent candidate gathers sufficient information to establish a working diagnosis ( acute persistent visual loss secondary to acute angle-closure glaucoma ). ■ The good candidate additionally has a structured and holistic approach ( excludes other causes such as trauma, infection, foreign body, vascular insufficiency or neural; pays attention to pain and the accompanying emotional distress ). Clinical reasoning ■ The competent candidate identifies the diagnosis ( acute closed-angle glaucoma needing emergency care ) and acknowledges the accompanying extreme distress . ■ The good candidate makes a specific diagnosis ( acute closed-angle glaucoma ) and has a structured approach to addressing the patient’s illness experience ( analgesia; normalises distress; recognises immediacy of analgesic need). Explaining and planning ■ The competent candidate uses clear language to explain the problem to the patient and uses strategies to ensure patient understanding ( questions OR feedback OR reverse summarising ). ■ The good candidate additionally ensures that the patient is actively involved in decision-making, paying particular attention to knowledge-sharing and empowerment, given the emergency of the situation and the further assessments needed to confirm the diagnosis. Management ■ The competent candidate proposes appropriate intervention (immediate discussion and referral to an ophthalmologist for complete assessment and rapid intervention). ■ The good candidate provides counselling to the patient on what she or he can expect at the ophthalmology department: Slit lamp or gonioscope examination; tomography. Mentions a discussion with an ophthalmologist for possible administration of timolol and pilocarpine eyedrops and intravenous acetazolamide while awaiting transfer . Role player instructions Appearance and behaviour Adult male or female, 40–60 years old, in pain and anxious. Opening statement: ‘Doctor, I have had this terrible pain in my right eye since this morning, and now I can’t see properly … please help me … Am I going blind?’ History Open responses: Freely tell the doctor: ■ You have no idea why you have this problem. It started on its own yesterday and made you nauseous and is very painful. This is the first time ever that you have had this problem. You took paracetamol and codeine about 3 h ago, with no effect. Closed responses: Only tell the doctor if asked: ■ Vision: Very hazy vision and lights have halos around them. ■ The pain is the worst pain you’ve ever felt, with a throbbing deep behind your right eye into your head. ■ There is no family history of glaucoma. ■ You have no other medical problems. ■ You live a healthy lifestyle – eat well and exercise regularly. ■ You have been smoking since the age of 20 years – a pack lasts about 2 days. Your medical history You have no medical issues that you know of. Ideas, concerns and expectations: You’re very worried that you will go blind and the pain is overwhelming. Examination findings Vital signs Heart rate: 123/min. Easily palpable pulses. Blood pressure: 140/85 mmHg. Temperature: 36.5 °C. Random blood glucose level: 5.2 mmol/L. Point of care haemoglobin: 13.3 g/dL. Examination findings Red, tearing right eye. Pupil not reacting to light. Cornea seems hazy and swollen. The globe feels harder than the right eye, and painful to palpation. Acuity: sees light, and hands at 30 cm only. Left eye, including visual acuity, is normal. All systemic examination reveals no abnormalities. Further reading Leveque T. Approach to the person with acute persistent visual loss [homepage on the Internet]. UpToDate; 2021 [cited 2023 Feb 08]. Available from: https://www.uptodate.com/contents/approach-to-the-adult-with-acute-persistent-visual-loss?search=retinal%20detachment&source=search_result&selectedTitle=1~150&usage_type=default&display_rank=1 Weizer JS. Angle-closure glaucoma [homepage on the Internet]. UpToDate; 2021 [cited 2023 Feb 08]. Available from: https://www.uptodate.com/contents/angle-closure-glaucoma?search=retinal%20detachment&topicRef=6902&source=see_link
This station tests the candidate’s ability to manage a patient with new onset unilateral visual loss.
Integrated consultation.
Simulated patient: Adult male or female.
You are the family physician working at a district hospital. Please consult with this patient, who presents to the emergency unit. Your task: Please consult with this patient and develop a comprehensive management plan. You do not need to do an examination on this patient. All examination findings will be provided on request.
This is an integrated consultation station in which the candidate has 15 min. Familiarise yourself with the assessor guidelines, which detail the required responses expected from the candidate. No marks are allocated. In the mark sheet (Table 1), tick off one of the three responses for each of the competencies listed. Make sure you are clear on what the criteria are for judging a candidate’s competence in each area. Provide the following information to the candidate when requested: examination findings: ■ Please switch off your cell phone. ■ Please do not prompt the student. ■ Please ensure that the station remains tidy and is reset between candidates.
The aim is to establish that the candidate has an effective and safe approach to managing acute persistent visual loss in an adult. Working definition of competent performance: The candidate effectively completes the task within the allotted time, in a manner that maintains patient safety, even though the execution may not be efficient and well structured: ■ Not competent : Patient safety is compromised (including ethical-legally) or the task is not completed. ■ Competent : The task is completed safely and effectively. ■ Good : In addition to displaying competence, the task is completed efficiently and in an empathic, patient-centred manner (acknowledges patient’s ideas, beliefs, expectations, concerns/fears). Establishes and maintains a good clinician–patient relationship. ■ The competent candidate is respectful and engages with the patient in a dignified manner. ■ The good candidate is empathic, compassionate and collaborative, facilitating patient participation in key areas of the consultation. Gathering information: ■ The competent candidate gathers sufficient information to establish a working diagnosis ( acute persistent visual loss secondary to acute angle-closure glaucoma ). ■ The good candidate additionally has a structured and holistic approach ( excludes other causes such as trauma, infection, foreign body, vascular insufficiency or neural; pays attention to pain and the accompanying emotional distress ). Clinical reasoning ■ The competent candidate identifies the diagnosis ( acute closed-angle glaucoma needing emergency care ) and acknowledges the accompanying extreme distress . ■ The good candidate makes a specific diagnosis ( acute closed-angle glaucoma ) and has a structured approach to addressing the patient’s illness experience ( analgesia; normalises distress; recognises immediacy of analgesic need). Explaining and planning ■ The competent candidate uses clear language to explain the problem to the patient and uses strategies to ensure patient understanding ( questions OR feedback OR reverse summarising ). ■ The good candidate additionally ensures that the patient is actively involved in decision-making, paying particular attention to knowledge-sharing and empowerment, given the emergency of the situation and the further assessments needed to confirm the diagnosis. Management ■ The competent candidate proposes appropriate intervention (immediate discussion and referral to an ophthalmologist for complete assessment and rapid intervention). ■ The good candidate provides counselling to the patient on what she or he can expect at the ophthalmology department: Slit lamp or gonioscope examination; tomography. Mentions a discussion with an ophthalmologist for possible administration of timolol and pilocarpine eyedrops and intravenous acetazolamide while awaiting transfer .
Appearance and behaviour Adult male or female, 40–60 years old, in pain and anxious. Opening statement: ‘Doctor, I have had this terrible pain in my right eye since this morning, and now I can’t see properly … please help me … Am I going blind?’ History Open responses: Freely tell the doctor: ■ You have no idea why you have this problem. It started on its own yesterday and made you nauseous and is very painful. This is the first time ever that you have had this problem. You took paracetamol and codeine about 3 h ago, with no effect. Closed responses: Only tell the doctor if asked: ■ Vision: Very hazy vision and lights have halos around them. ■ The pain is the worst pain you’ve ever felt, with a throbbing deep behind your right eye into your head. ■ There is no family history of glaucoma. ■ You have no other medical problems. ■ You live a healthy lifestyle – eat well and exercise regularly. ■ You have been smoking since the age of 20 years – a pack lasts about 2 days. Your medical history You have no medical issues that you know of. Ideas, concerns and expectations: You’re very worried that you will go blind and the pain is overwhelming.
Adult male or female, 40–60 years old, in pain and anxious. Opening statement: ‘Doctor, I have had this terrible pain in my right eye since this morning, and now I can’t see properly … please help me … Am I going blind?’
Open responses: Freely tell the doctor: ■ You have no idea why you have this problem. It started on its own yesterday and made you nauseous and is very painful. This is the first time ever that you have had this problem. You took paracetamol and codeine about 3 h ago, with no effect. Closed responses: Only tell the doctor if asked: ■ Vision: Very hazy vision and lights have halos around them. ■ The pain is the worst pain you’ve ever felt, with a throbbing deep behind your right eye into your head. ■ There is no family history of glaucoma. ■ You have no other medical problems. ■ You live a healthy lifestyle – eat well and exercise regularly. ■ You have been smoking since the age of 20 years – a pack lasts about 2 days. Your medical history You have no medical issues that you know of. Ideas, concerns and expectations: You’re very worried that you will go blind and the pain is overwhelming.
Vital signs Heart rate: 123/min. Easily palpable pulses. Blood pressure: 140/85 mmHg. Temperature: 36.5 °C. Random blood glucose level: 5.2 mmol/L. Point of care haemoglobin: 13.3 g/dL. Examination findings Red, tearing right eye. Pupil not reacting to light. Cornea seems hazy and swollen. The globe feels harder than the right eye, and painful to palpation. Acuity: sees light, and hands at 30 cm only. Left eye, including visual acuity, is normal. All systemic examination reveals no abnormalities. Further reading Leveque T. Approach to the person with acute persistent visual loss [homepage on the Internet]. UpToDate; 2021 [cited 2023 Feb 08]. Available from: https://www.uptodate.com/contents/approach-to-the-adult-with-acute-persistent-visual-loss?search=retinal%20detachment&source=search_result&selectedTitle=1~150&usage_type=default&display_rank=1 Weizer JS. Angle-closure glaucoma [homepage on the Internet]. UpToDate; 2021 [cited 2023 Feb 08]. Available from: https://www.uptodate.com/contents/angle-closure-glaucoma?search=retinal%20detachment&topicRef=6902&source=see_link
|
Synthesis of 3-Amino-4-substituted Monocyclic ß-Lactams—Important Structural Motifs in Medicinal Chemistry | bd66009a-5ce0-4200-959c-b41da6336ea5 | 8745335 | Pharmacology[mh] | Ever since Alexander Fleming’s serendipitous discovery of the first broad-spectrum ß-lactam antibiotic (i.e., penicillin G) in the late 1920s, ß-lactams arguably remain the single most clinically useful class of antibiotics discovered to date, in some countries making up over 60% of all antibiotic sales . Their excellent safety and efficacy profiles and the highly reactive nature of the CO-N bond in the ß-lactam ring have propelled this structural motif in many drug discovery initiatives, besides their primary use as antibacterial agents . Not only presence of the bicyclic ring system of penicillins is essential for their antibacterial activity, but it can also be replaced by a monocyclic ß-lactam. Appropriately decorated 3-aminoazetidin-2-ones serve as mimics of the D-Ala-D-Ala subunit of the stem peptide in the nascent peptidoglycan. ß-Lactams act as mechanism-based inhibitors of the transpeptidase activity of penicillin-binding proteins (PBPs), thus inhibiting the cross-linking step in peptidoglycan chains, ultimately leading to bacterial cell death . Aztreonam, the first clinically approved synthetic monobactam (i.e., N -sulfonated monocyclic ß-lactam) in the 1980s, is still in use worldwide because of its suitable activity against Gram-negative bacteria and ß-lactamase stability . Another approved monocyclic ß-lactam drug in clinical use is ezetimibe, which acts as a cholesterol absorption inhibitor and is used to treat hypercholesterolemia . Azetidin-2-one, a fundamental building block of all ß-lactam antibiotics, is a four-membered cyclic lactam (i.e., ß-lactam ring) with an oxo group at the C-2 position and various substituents at the N-1, C-3, and C-4 positions . The highly strained nature of the core ß-lactam skeleton is largely responsible for its chemical reactivity, which can be enhanced further by the nature of the substituents present on the ß-lactam ring, and in particular on the endocyclic nitrogen (N-1) . Structural modifications of this scaffold with different substituents on the monocyclic ß-lactam core can lead to a wide range of biological activities . Derivatives of monocyclic ß-lactams have been explored as agents for the treatment of atherosclerotic coronary heart disease, allergic and inflammatory conditions, autoimmune diseases, neurodegenerative diseases, diabetes, arterial thrombosis, microbial infections, and cancer . Appropriately decorated azetidine-2-ones act as inhibitors of different nucleophilic enzymes, most commonly serine, cysteine, and threonine proteases . Monocyclic ß-lactams are explored as cholesterol absorption inhibitors, activators and inhibitors of lecithin-cholesterol acyltransferase, vasopressin V1a antagonists, tryptase and chymase inhibitors, thrombin inhibitors, factor XIa or kallikrein inhibitors, cathepsin K inhibitors, 20S proteasome inhibitors, human leukocyte elastase inhibitors, beta lactam combretastatin mimetics, carbonic anhydrase inhibitors, N -acyl ethanolamine acid amidase inhibitors, inhibitors of dengue and West Nile virus NS2B-NS3 protease, human cytomegalovirus protease inhibitors, RORγt (retinoid-related orphan receptor gamma t) modulators and glutamate uptake modulators . In addition to their own therapeutic potential, they are useful as synthetic synthons in the preparation of various other compounds in modern organic chemistry . Since the discovery of the first monocyclic ß-lactams, which were originally isolated from bacteria and were, as such, not suitable for further chemical modifications, many efforts have been directed toward the development of new synthetic methods for their production, which would enable the desirable structural diversification . These can be broadly categorized into cycloadditions, cyclizations, and other transformations . Monocyclic β-lactams are most often synthesized by the Staudinger [2+2] cycloaddition between ketenes, generated in situ by treating acyl chlorides with a mild base and imines ( , reaction A) . This three-step reaction cascade consists of (i) nucleophilic addition of the imine nitrogen to the electrophilic carbon of the ketene, (ii) the formation of a zwitterionic intermediate, and (iii) ring closure . More recently, enolate-imine cycloaddition and the Kingunsa reaction with rearrangement have been studied as means to access decorated monocyclic β-lactams. Various other cyclizations involving N1-C2, C3-C4, and N1-C4 ring closures have also been reported . In addition, there are also other less commonly used procedures, such as CO insertion [3+1] on aziridine ring, Ugi four-component reactions, and diiodomethane additions to amide dianions . The most widely used protocols are based on either Mitsunobu-mediated cyclization of α-hydroxy-ß-amino acid hydroxamates ( , reaction B), or bromine-induced cyclization of γ,δ-unsaturated hydroxamates ( , reaction C). While both synthetic strategies rely on the acidity of the N-H in hydroxamates to facilitate the desirable cyclization and allow for a wide range of different substitutions at C-3 and C-4 of the newly formed ring, the number and complexity of synthetic steps required are often preventing synthetic success . Further structural modifications of monocyclic ß-lactams are necessary due to emerging bacterial resistance and increased production of ß-lactamases , hydrolytic enzymes that can inactivate ß-lactams by hydrolysis . We, therefore, believe that there is scope for improving the pharmacological profiles of azetidine-2-ones, especially in antibacterial and anticancer applications, with the introduction of appropriate substituents on the C-4 position. Herein we report our synthetic efforts to prepare a set of novel monocyclic ß-lactams using Staudinger cycloaddition reaction and studies on the optimization of the challenging N-1/N-3 deprotection reactions. The main objective of this work was, therefore, to establish a more convenient and reliable method for the synthesis of 3-amino-4-substituted azetidin-2-ones, which are important intermediates in the development of pharmacologically relevant monocyclic ß-lactams.
Our initial efforts to prepare the desired 3-amino-4-substituted monocyclic ß-lactams through N1–C4 ring closure reactions (e.g., via Mitsunobu cyclization or bromine-induced cyclization ) were unproductive. In a subsequent approach, the convergent methodology for the stereoselective synthesis of functionalized β-lactams with a broad substrate scope developed by Staudinger et al. was explored. The preparation of the ketene and imine building blocks required for the Staudinger [2+2] cycloaddition to provide target ß-lactams, as well as our studies on the necessary N-1 and N-3 deprotections, are discussed in more detail below. 2.1. Methods for Cyclization of 2-Azetidinone (Monocyclic Beta Lactam Core) The requisite imines ( 1–14 ) were formed by the condensation of appropriate primary amines and aldehydes in dichloromethane or methanol at room temperature, using anhydrous sodium sulfate as a drying agent . To demonstrate the possibility of incorporating a variety of different substituents at the C-4 position of the monocyclic ß-lactam, we selected several aldehydes from our in-house library of chemicals. The selected amines were previously described in the synthesis of monocyclic ß-lactams. In the case of the more reactive aliphatic aldehydes ( 15–17 ), the condensation reactions were carried out on ice, and the imines were used directly without evaporation of the solvent. We initially focused our efforts on the preparation of monocyclic ß-lactams using ketenes obtained from t -butylcarbamate- or benzylcarbamate-protected glycine and imines derived from aromatic aldehydes to test reactivity in Staudinger model reactions. However, the expected [2+2] cycloaddition products (i.e., 2-azetidinones) were not observed with any of the evaluated carbamates. This may be due to the competing formation of 1,3-oxazin-4-ones, which are highly stable and cannot react further to form 2-azetidinones . Therefore, we have elected to use a phthalimido group to protect the glycine-amino group instead. Ketenes, prepared in situ from an acyl chloride with N -phthalimido protecting group (e.g., 18 ), were prone to undergo the desired cycloadditions . The reactions proceeded smoothly when the nitrogen of the amino acid residue was protected by substitution of both hydrogen atoms, as in the case of phthalimido-protected glycine. An acyl chloride was added dropwise to a mixture of imine and a base in toluene at 80 °C, and the product formed was easily isolated by precipitation or column chromatography. Because of the instability of ketenes, the order of the addition of the reactants was also an important factor. The imines ( 1 – 17 ) were mainly obtained by the reactions of aromatic aldehydes (which were substituted by electron-withdrawing groups) with dimethoxybenzylamine or benzylamine. The products of the Staudinger reaction ( 19 – 27 ) in the case of an aromatic or heterocyclic substituent at the C-4 position of the ring were mainly isolated as cis -isomers; trans -stereoisomers were either not detected or were only present in traces that we could not isolate. Staudinger cycloaddition is a stepwise reaction initiated by the nucleophilic attack of an imine on a ketene, leading to a zwitterionic intermediate, followed by ring closure of this intermediate. Direct ring closure leads to the cis -stereoisomer, while indirect ring closure with further isomerization leads to the trans -stereoisomer. As previously reported in the literature, we found that electron-withdrawing groups on the imine facilitate the progress of the reaction, and electron-donating groups slow down the cyclization. Improved yields and exclusive formation of cis -stereoisomer were obtained with imines bearing aromatic substituents on the imine moiety, compared to imines formed from aliphatic aldehydes, which provided much lower yields and lower diastereoselectivity. The cis- configuration of newly synthesized monocyclic β-lactams was deduced using 1 H NMR coupling constants ( J values) of the β-lactam ring hydrogens H-3 and H-4; for cis -β-lactams J 3,4 ~ 5 Hz, and for trans- β-lactams J 3,4 ~ 2 Hz . Since the removal of the phthalimide protecting group requires relatively harsh conditions, we opted to prepare ß-lactam analogs bearing carbamate protecting groups at the N-3 position instead, which we hoped would be more easily removed. Alternatively, functionalized 2-azetidinones can also be prepared via microwave-assisted coupling of imines with diazoketones, which can be derived from t -butylcarbamate- or benzylcarbamate-protected α-amino acids . Such monocyclic ß-lactams are structurally different from analogous derivatives prepared via the previously described acyl chloride method by having an additional methylene unit present at C-3 of the ß-lactam ring . The monocyclic ß-lactams ( 32 – 34 ), which were prepared using this methodology, were isolated as trans- isomers, as opposed to the otherwise cis- isomers, which are formed via, e.g., Staudindger synthesis. A significant disadvantage of this method is the preparation of diazoketone ( 31 ), as most methods require the use of highly toxic diazomethane or expensive trimethylsilyldiazometane . Finally, another convenient method was used to synthesize the ß-lactam ring from t -butylcarbamate or benzylcarbamate-protected α-amino acids by Staudinger reaction. Cycloaddition was carried out with the ketenes derived from the mixed anhydride at −70 °C in dry tetrahydrofuran ( 35 – 37 , ). Again, the cis isomer was a major product but with lower yields, which could not be improved by changing the addition order of the reactants. 2.2. Deprotection of C3-NH 2 Protecting Group Since the phthalimide (Phth) moiety is the most commonly used amino protecting group in the ß-lactam ring cyclization reaction (because the cyclization of such Phth-protected ketenes proceeds in high yields), we wanted to optimize the conditions for its deprotection. However, the deprotection methods are quite harsh as they usually involve the use of a very strong base, such as hydrazine hydrate. A variety of Phth deprotection reagents were surveyed, including ethylenediamine, ethanolamine, methylhydrazine, and hydrazine hydrate. The highest product isolated yield was obtained when hydrazine hydrate was used (in contrast, the yield was considerably lower with ethylenediamine and ethanolamine, which are also milder reagents). In the case of methyl hydrazine, the reaction was very slow, even with a high excess of reagent used. The problem, which has not been described in the literature, is that the reaction of the phthalimide-protected monocyclic ß-lactams ( 39 – 40 ) with the hydrazine hydrate very likely stops after 1 h because a salt forms with the hydrazine ( 38 ). Removal of the excess hydrazine and further addition of a few drops of concentrated hydrochloric acid breaks down the salt formed . Once the HCl is removed, deprotection of the phthalimide group can continue, and the deprotected ß-lactams with the free amine group at C-3 can be isolated in high yields. Again, in the case of aliphatic substituents ( 41 ), which have an electron donor electronic effect than aromatic ones, deprotection with hydrazine hydrate proceeded rapidly and without any adjustments. 2.3. Deprotection of N1 Protecting Group With the optimized conditions for the [2+2] cycloaddition in hand , we moved our attention to the identification of the most optimal protecting group for the lactam amide nitrogen (i.e., N-1), that would (i) favor the cyclization, and (ii) be easily removable at the end. The preparation of target monocyclic ß-lactams was highly dependent on the success of N-1 deprotection. The deprotection conditions had to be harsh enough to remove the protecting group without concurrent opening of the highly sensitive ß-lactam ring. In the initial studies, we prepared a small set of N -benzyl ß-lactams ( 19 , 23 , 32 – 33 ) because we expected to be able to remove this protecting group easily with catalytic hydrogenation. Unfortunately, none of the traditional catalysts and hydrogen sources employed (e.g., Pd/C, Pd(OH) 2 with cyclohexene) yielded any product. We have also attempted the aforementioned catalytic hydrogenation under elevated pressure (30 bar); LC-MS and NMR analyses of the reaction mixtures revealed that under these conditions, the N-1 benzyl group was cleaved, but the product yield was too low to enable the isolation and purification of the desired compounds . Since deprotection of the benzyl group proved highly problematic, we prepared some ß-lactams with dimethoxybenzyl protecting group at the N-1 position ( 20 – 22 , 29 – 30 , 35 – 37 ). There are several published procedures for removing the para -methoxybenzyl or di-methoxybenzyl group from the amide nitrogen. The procedures that we investigated are summarized in . First, we attempted to treat N-1-dimethoxybenzyl ß-lactam with strong acids, such as p -toluenesulfonic acid and trifluoroacetic acid (at 60 °C), but this yielded only starting material, and a side product that we assumed (based on NMR) was an opened ß-lactam ring . Next, we have attempted to deprotect N-1 via oxidative cleavage of the dimethoxybenzyl protecting group. Several procedures using persulfate salts (e.g., potassium and/or ammonium persulfate, under various conditions including heating and acid addition ), which are known to be strong oxidizing agents, provided poor yields and numerous side products, making the isolation of the desired product by column chromatography extremely challenging. A process commonly used to deprotect lactam nitrogen in the literature was Birch reduction . Since the standard process requires the use of toxic liquid ammonia and is often very time consuming, we turned our attention to the more recently published ammonia-free Birch reduction . While the reaction under ammonia-free Birch reduction conditions provided no desired product in the case of dimethoxybenzyl, and benzyl ß-lactam derivates with aromatic C-4 substituents, an opened ß-lactam ring with eliminated phthalimide group has been isolated as an exclusive product. In the case of the trifluorophenyl group ( 42 ), the fluorine atoms were exchanged for hydrogen . The situation was quite different for compounds with aliphatic substituents, where the above deprotection could be performed in excellent yields and with almost no side products detected . Finally, the best and most reliable deprotection approach was achieved by using a milder oxidant, cerium ammonium nitrate ( 46 – 50 ) . Oxidative cleavage of N -dimethoxybenzyl protection with cerium ammonium nitrate in aqueous acetonitrile was achieved at the temperature of −10 °C, with the minimum formation of side product (<10%; e.g., compound 45 , ). We found that the absence of atmospheric oxygen and the water/acetonitrile ratio were important factors in the amount of side product formed. Various relative amounts of acetonitrile/water were tried (from the ratio MeCN/H 2 O = 2:1 to 1:3), with the proportion of product varying from 14% to 52%. The best yield was obtained with a 1:1 ratio of water: acetonitrile, with minimal formation of oxidized, non-deprotected side products observed. However, oxidative dimethoxybenzyl cleavage with cerium ammonium nitrate was unsuccessful for monocyclic ß-lactams that had aliphatic substituents at the C-4 position. Finally, as an example of the synthetic potential of the methods described in this manuscript, we have prepared a fully deprotected 3-amino-4-substituted azetidin-2-one 54 . The first step after cyclization was the cleavage of the phthalimide protecting group, as this requires the harshest conditions for deprotection, and the ß-lactam ring still protected at the lactam nitrogen is the most stable. Since oxidative cleavage of the dimethoxybenzyl protecting group with a free amino group at the C-3 position was not possible, we protected it again with a t-butyl carbamate protecting group that is stable to oxidation. For this purpose, we used di- tert -butyl dicarbonate with triethylamine in dichloromethane. After successful conversion of the phthalimide to the t -butyl carbamate protecting group ( 43 , 51 – 53 , shown in ), we used cerium ammonium nitrate to remove the dimethoxybenzyl protecting group from the ring nitrogen or ammonia-free Birch reduction in case of aliphatic substituent on C-4 position. Deprotection of the t -butyl carbamate protecting group with hydrochloric acid (4N HCl/dioxane) failed and resulted in the isolation of an opened monocyclic ß-lactam ring. However, the use of trifluoroacetic acid with anisole as a scavenger agent removed the Boc-protecting group in high yield ( , compound 54 ).
The requisite imines ( 1–14 ) were formed by the condensation of appropriate primary amines and aldehydes in dichloromethane or methanol at room temperature, using anhydrous sodium sulfate as a drying agent . To demonstrate the possibility of incorporating a variety of different substituents at the C-4 position of the monocyclic ß-lactam, we selected several aldehydes from our in-house library of chemicals. The selected amines were previously described in the synthesis of monocyclic ß-lactams. In the case of the more reactive aliphatic aldehydes ( 15–17 ), the condensation reactions were carried out on ice, and the imines were used directly without evaporation of the solvent. We initially focused our efforts on the preparation of monocyclic ß-lactams using ketenes obtained from t -butylcarbamate- or benzylcarbamate-protected glycine and imines derived from aromatic aldehydes to test reactivity in Staudinger model reactions. However, the expected [2+2] cycloaddition products (i.e., 2-azetidinones) were not observed with any of the evaluated carbamates. This may be due to the competing formation of 1,3-oxazin-4-ones, which are highly stable and cannot react further to form 2-azetidinones . Therefore, we have elected to use a phthalimido group to protect the glycine-amino group instead. Ketenes, prepared in situ from an acyl chloride with N -phthalimido protecting group (e.g., 18 ), were prone to undergo the desired cycloadditions . The reactions proceeded smoothly when the nitrogen of the amino acid residue was protected by substitution of both hydrogen atoms, as in the case of phthalimido-protected glycine. An acyl chloride was added dropwise to a mixture of imine and a base in toluene at 80 °C, and the product formed was easily isolated by precipitation or column chromatography. Because of the instability of ketenes, the order of the addition of the reactants was also an important factor. The imines ( 1 – 17 ) were mainly obtained by the reactions of aromatic aldehydes (which were substituted by electron-withdrawing groups) with dimethoxybenzylamine or benzylamine. The products of the Staudinger reaction ( 19 – 27 ) in the case of an aromatic or heterocyclic substituent at the C-4 position of the ring were mainly isolated as cis -isomers; trans -stereoisomers were either not detected or were only present in traces that we could not isolate. Staudinger cycloaddition is a stepwise reaction initiated by the nucleophilic attack of an imine on a ketene, leading to a zwitterionic intermediate, followed by ring closure of this intermediate. Direct ring closure leads to the cis -stereoisomer, while indirect ring closure with further isomerization leads to the trans -stereoisomer. As previously reported in the literature, we found that electron-withdrawing groups on the imine facilitate the progress of the reaction, and electron-donating groups slow down the cyclization. Improved yields and exclusive formation of cis -stereoisomer were obtained with imines bearing aromatic substituents on the imine moiety, compared to imines formed from aliphatic aldehydes, which provided much lower yields and lower diastereoselectivity. The cis- configuration of newly synthesized monocyclic β-lactams was deduced using 1 H NMR coupling constants ( J values) of the β-lactam ring hydrogens H-3 and H-4; for cis -β-lactams J 3,4 ~ 5 Hz, and for trans- β-lactams J 3,4 ~ 2 Hz . Since the removal of the phthalimide protecting group requires relatively harsh conditions, we opted to prepare ß-lactam analogs bearing carbamate protecting groups at the N-3 position instead, which we hoped would be more easily removed. Alternatively, functionalized 2-azetidinones can also be prepared via microwave-assisted coupling of imines with diazoketones, which can be derived from t -butylcarbamate- or benzylcarbamate-protected α-amino acids . Such monocyclic ß-lactams are structurally different from analogous derivatives prepared via the previously described acyl chloride method by having an additional methylene unit present at C-3 of the ß-lactam ring . The monocyclic ß-lactams ( 32 – 34 ), which were prepared using this methodology, were isolated as trans- isomers, as opposed to the otherwise cis- isomers, which are formed via, e.g., Staudindger synthesis. A significant disadvantage of this method is the preparation of diazoketone ( 31 ), as most methods require the use of highly toxic diazomethane or expensive trimethylsilyldiazometane . Finally, another convenient method was used to synthesize the ß-lactam ring from t -butylcarbamate or benzylcarbamate-protected α-amino acids by Staudinger reaction. Cycloaddition was carried out with the ketenes derived from the mixed anhydride at −70 °C in dry tetrahydrofuran ( 35 – 37 , ). Again, the cis isomer was a major product but with lower yields, which could not be improved by changing the addition order of the reactants.
2 Protecting Group Since the phthalimide (Phth) moiety is the most commonly used amino protecting group in the ß-lactam ring cyclization reaction (because the cyclization of such Phth-protected ketenes proceeds in high yields), we wanted to optimize the conditions for its deprotection. However, the deprotection methods are quite harsh as they usually involve the use of a very strong base, such as hydrazine hydrate. A variety of Phth deprotection reagents were surveyed, including ethylenediamine, ethanolamine, methylhydrazine, and hydrazine hydrate. The highest product isolated yield was obtained when hydrazine hydrate was used (in contrast, the yield was considerably lower with ethylenediamine and ethanolamine, which are also milder reagents). In the case of methyl hydrazine, the reaction was very slow, even with a high excess of reagent used. The problem, which has not been described in the literature, is that the reaction of the phthalimide-protected monocyclic ß-lactams ( 39 – 40 ) with the hydrazine hydrate very likely stops after 1 h because a salt forms with the hydrazine ( 38 ). Removal of the excess hydrazine and further addition of a few drops of concentrated hydrochloric acid breaks down the salt formed . Once the HCl is removed, deprotection of the phthalimide group can continue, and the deprotected ß-lactams with the free amine group at C-3 can be isolated in high yields. Again, in the case of aliphatic substituents ( 41 ), which have an electron donor electronic effect than aromatic ones, deprotection with hydrazine hydrate proceeded rapidly and without any adjustments.
With the optimized conditions for the [2+2] cycloaddition in hand , we moved our attention to the identification of the most optimal protecting group for the lactam amide nitrogen (i.e., N-1), that would (i) favor the cyclization, and (ii) be easily removable at the end. The preparation of target monocyclic ß-lactams was highly dependent on the success of N-1 deprotection. The deprotection conditions had to be harsh enough to remove the protecting group without concurrent opening of the highly sensitive ß-lactam ring. In the initial studies, we prepared a small set of N -benzyl ß-lactams ( 19 , 23 , 32 – 33 ) because we expected to be able to remove this protecting group easily with catalytic hydrogenation. Unfortunately, none of the traditional catalysts and hydrogen sources employed (e.g., Pd/C, Pd(OH) 2 with cyclohexene) yielded any product. We have also attempted the aforementioned catalytic hydrogenation under elevated pressure (30 bar); LC-MS and NMR analyses of the reaction mixtures revealed that under these conditions, the N-1 benzyl group was cleaved, but the product yield was too low to enable the isolation and purification of the desired compounds . Since deprotection of the benzyl group proved highly problematic, we prepared some ß-lactams with dimethoxybenzyl protecting group at the N-1 position ( 20 – 22 , 29 – 30 , 35 – 37 ). There are several published procedures for removing the para -methoxybenzyl or di-methoxybenzyl group from the amide nitrogen. The procedures that we investigated are summarized in . First, we attempted to treat N-1-dimethoxybenzyl ß-lactam with strong acids, such as p -toluenesulfonic acid and trifluoroacetic acid (at 60 °C), but this yielded only starting material, and a side product that we assumed (based on NMR) was an opened ß-lactam ring . Next, we have attempted to deprotect N-1 via oxidative cleavage of the dimethoxybenzyl protecting group. Several procedures using persulfate salts (e.g., potassium and/or ammonium persulfate, under various conditions including heating and acid addition ), which are known to be strong oxidizing agents, provided poor yields and numerous side products, making the isolation of the desired product by column chromatography extremely challenging. A process commonly used to deprotect lactam nitrogen in the literature was Birch reduction . Since the standard process requires the use of toxic liquid ammonia and is often very time consuming, we turned our attention to the more recently published ammonia-free Birch reduction . While the reaction under ammonia-free Birch reduction conditions provided no desired product in the case of dimethoxybenzyl, and benzyl ß-lactam derivates with aromatic C-4 substituents, an opened ß-lactam ring with eliminated phthalimide group has been isolated as an exclusive product. In the case of the trifluorophenyl group ( 42 ), the fluorine atoms were exchanged for hydrogen . The situation was quite different for compounds with aliphatic substituents, where the above deprotection could be performed in excellent yields and with almost no side products detected . Finally, the best and most reliable deprotection approach was achieved by using a milder oxidant, cerium ammonium nitrate ( 46 – 50 ) . Oxidative cleavage of N -dimethoxybenzyl protection with cerium ammonium nitrate in aqueous acetonitrile was achieved at the temperature of −10 °C, with the minimum formation of side product (<10%; e.g., compound 45 , ). We found that the absence of atmospheric oxygen and the water/acetonitrile ratio were important factors in the amount of side product formed. Various relative amounts of acetonitrile/water were tried (from the ratio MeCN/H 2 O = 2:1 to 1:3), with the proportion of product varying from 14% to 52%. The best yield was obtained with a 1:1 ratio of water: acetonitrile, with minimal formation of oxidized, non-deprotected side products observed. However, oxidative dimethoxybenzyl cleavage with cerium ammonium nitrate was unsuccessful for monocyclic ß-lactams that had aliphatic substituents at the C-4 position. Finally, as an example of the synthetic potential of the methods described in this manuscript, we have prepared a fully deprotected 3-amino-4-substituted azetidin-2-one 54 . The first step after cyclization was the cleavage of the phthalimide protecting group, as this requires the harshest conditions for deprotection, and the ß-lactam ring still protected at the lactam nitrogen is the most stable. Since oxidative cleavage of the dimethoxybenzyl protecting group with a free amino group at the C-3 position was not possible, we protected it again with a t-butyl carbamate protecting group that is stable to oxidation. For this purpose, we used di- tert -butyl dicarbonate with triethylamine in dichloromethane. After successful conversion of the phthalimide to the t -butyl carbamate protecting group ( 43 , 51 – 53 , shown in ), we used cerium ammonium nitrate to remove the dimethoxybenzyl protecting group from the ring nitrogen or ammonia-free Birch reduction in case of aliphatic substituent on C-4 position. Deprotection of the t -butyl carbamate protecting group with hydrochloric acid (4N HCl/dioxane) failed and resulted in the isolation of an opened monocyclic ß-lactam ring. However, the use of trifluoroacetic acid with anisole as a scavenger agent removed the Boc-protecting group in high yield ( , compound 54 ).
Using Staudinger [2+2] cycloaddition, we successfully synthesized a series of diprotected monocyclic ß-lactams with different substituents at the C-4 position. These initial ß-lactams had phthalimido-protected 3-amino group and dimethoxybenzyl protected ring nitrogen (N-1). Through an extensive study of previously published methods and their subsequent optimization, we have achieved the selective deprotection of both protecting groups in high yield. Oxidative cleavage with cerium ammonium nitrate selectively removed the N-1 protecting group when the aromatic substituents were at the C-4 position, while ammonia-free Birch reduction provided the highest yields for compounds with aliphatic C-4 substituents. For the removal of the phthalimido group, hydrazine hydrate provided the best yield, but in the case of aromatic substituents at the C-4 position, synthetic modification by HCl addition was required. The presented methods and the synthesized protected and partially deprotected 3-amino-4-substituted monocyclic ß-lactams are an important step toward new ß-lactams with potential pharmacological activities.
4.1. Chemistry and Chemical Characterization of Compounds Unless otherwise stated, all reactions were carried out under argon atmosphere in flame-dried glassware. Chemicals and solvents were obtained from commercial sources (Sigma-Aldrich, Acros Organics, TCI Europe, fluorochem, and Apollo Sci) and were used as supplied. Dry solvents were prepared by distillation from CaH 2 (CH 2 Cl 2 ) or from a mixture of sodium and benzophenone (tetrahydrofuran). Other solvents (dimethylformamide, toluene, methanol, and CH 3 CN) were used directly from anhydrous Aldrich Sure/Seal bottles. Evaporation of the solvent was carried out under reduced pressure. Reactions were monitored by thin-layer chromatography (TLC) on silica gel aluminum plates (Merck DC Fertigplatten Kieselgel 60 GF254), visualized under UV light (254 nm), and stained with appropriate TLC stains for detection (ninhydrin, dinitrophenylhydrazine, and phospho-molybdic acid). The products were purified by flash column chromatography performed on Merck silica gel 60 (mesh size, 70–230) using the indicated solvents. Yields are reported for the purified products. 1 H NMR and 13 C NMR spectra were recorded at 295 K using a Bruker Avance III NMR spectrometer equipped with a Broadband decoupling inverse 1 H probe, at 400 MHz and 100 MHz, respectively. Chemical shifts (δ) are given in parts per million (ppm) and refer to tetramethylsilane (TMS) as an internal standard. The coupling constants ( J ) are given in Hertz (Hz), and the splitting patterns are reported as: s, singlet; br s, broad singlet; d, doublet; dd, double doublet; t, triplet, and m, multiplet. Mass spectra were recorded using an ADVION Expres-sion CMSL mass spectrometer (Advion Inc., Ithaca, NY, USA). High-resolution, accurate mass measurements were performed using the ExactiveTM Plus Orbitrap mass spectrometer (Thermo Fisher Scientific Inc., Waltham, MA, USA). 4.2. General Procedure for the Synthesis of Schiff Bases ( 1 – 17 ) To a solution of an appropriate aldehyde (1 EQ) in dry dichloromethane or dry methanol was added an amine (1 EQ). The resultant solution was stirred for 15 min before Na 2 SO 4 (4 EQ) was added. The reaction mixture was then stirred at room temperature until TLC showed complete consumption of the starting material (30 min to 16 h). Next, the drying agent was removed by filtration, and the volatiles were removed under reduced pressure to afford the desired products, which were used in the next step without further purification. N -Benzyl-1-(4-(trifluoromethyl)phenyl)methanimine ( 1 ), quantitative yield, brown oil. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.62 (s, 1H), 8.01 (d, J = 8.7 Hz, 2H), 7.83 (d, J = 8.7 Hz, 2H), 7.33 (m, 5H), 4.84 (s, 2H); Rf = 0.66 (EtOAc/Hexane = 1:1, v / v ) as reported . N -(2,4-Dimethoxybenzyl)-1-(4-(trifluoromethyl)phenyl)methanimine ( 2 ), quantitative yield, brown oil. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.50 (s, 1H), 7.97 (d, J = 8.0 Hz, 2H), 7.81 (d, J = 8.2 Hz, 2H), 7.16 (d, J = 8.3 Hz, 1H), 6.58 (d, J = 2.4 Hz, 1H), 6.51 (dd, J = 8.3, 2.4 Hz, 1H), 4.71 (s, 2H), 3.79 (s, J = 4.7 Hz, 3H), 3.76 (s, J = 3.6 Hz, 3H); Rf = 0.60 (EtOAc/Hexane = 1:1, v / v ) as reported . 4-(((2,4-Dimethoxybenzyl)imino)methyl)benzonitrile ( 3 ), quantitative yield, colorless amorphous solid. 1 H NMR (400 MHz, CDCl 3 ) δ 8.33 (s, 1H), 7.89–7.79 (m, 2H), 7.74–7.61 (m, 2H), 7.21–7.13 (m, 1H), 6.53–6.42 (m, 2H), 4.80 (s, 2H), 3.81 (app s, 3H); Rf = 0.54 (EtOAc/Hexane = 1:1, v / v ) as reported . 1-(3-Bromo-4-fluorophenyl)- N -(2,4-dimethoxybenzyl)methanimine ( 4 ), quantitative yield, yellow oil. 1 H NMR (400 MHz, CDCl 3 ) δ 8.21 (s, 1H), 8.00 (dd, J = 6.8, 2.1 Hz, 1H), 7.68–7.58 (m, 1H), 7.21–7.14 (m, 1H), 7.15–7.05 (m, 1H), 6.50–6.46 (m, 2H), 4.74 (s, 2H), 3.80 (app s, 6H). 13 C NMR (100 MHz, CDCl 3 ) δ 161.61, 160.32, 158.86, 158.37, 133.10, 130.33, 128.99, 128.91, 119.29, 116.64, 116.41, 104.16, 98.56, 58.78, 55.40, 55.40. HRMS (ESI+) m/z calc. for C 16 H 15 BrFNO 2 351.0270, found [M + H] + 352.0338. Rf = 0.86 (EtOAc/Hexane = 1:1 v / v ). N -Benzyl-1-(4-nitrophenyl)methanimine ( 5 ), quantitative yield, yellow amorphous solid. 1 H NMR (400 MHz, CDCl 3 ) δ 8.46 (s, 1H), 8.26 (d, J = 8.7 Hz, 2H), 7.94 (d, J = 8.7 Hz, 2H), 7.43–7.15 (m, 5H), 4.88 (s, 2H); Rf = 0.66 (EtOAc/Hexane = 1:1, v / v ) as reported . N -(2,4-Dimethoxybenzyl)-1-(4-nitrophenyl)methanimine ( 6 ), quantitative yield, yellow amorphous solid. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.54 (s, 1H), 8.31–8.26 (m, 2H), 8.04–7.98 (m, 2H), 7.16 (d, J = 8.3 Hz, 1H), 6.58 (d, J = 2.4 Hz, 1H), 6.51 (dd, J = 8.3, 2.4 Hz, 1H), 4.73 (s, 2H), 3.79 (s, 3H), 3.76 (s, 3H); Rf = 0.46 (EtOAc/Hexane = 1:1, v / v ) as reported . N -(2,4-Dimethoxybenzyl)-1-(4-(methylsulfonyl)phenyl)methanimine ( 7 ), quantitative yield, pale yellow amorphous solid. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.51 (s, 1H), 8.00 (m, 4H), 7.16 (d, J = 8.3 Hz, 1H), 6.58 (d, J = 2.4 Hz, 1H), 6.51 (dd, J = 8.3, 2.4 Hz, 1H), 4.72 (s, 2H), 3.79 (s, 3H), 3.76 (s, 3H), 3.25 (s, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 160.40, 159.59, 158.40, 141.76, 141.24, 130.37, 128.93, 127.63, 119.01, 104.21, 98.57, 59.13, 55.41, 44.46. HRMS (ESI+) m/z calc. for C 17 H 19 NO 4 S 333.1035, found [M + H] + 334.1104. Rf = 0.25 (EtOAc:Hex = 1:1, v / v ). 4-(((2,4-Dimethoxybenzyl)imino)methyl)- N,N -dimethylaniline ( 8 ), quantitative yield, colorless amorphous solid. 1 H NMR (400 MHz, CDCl 3 ) δ 8.19 (d, J = 1.4 Hz, 1H), 7.63 (d, J = 8.9 Hz, 2H), 7.18 (d, J = 8.9 Hz, 2H), 6.68 (d, J = 8.9 Hz, 1H), 6.48–6.42 (m, 2H), 4.68 (s, 2H), 3.79 (s, 3H), 3.78 (s, 3H), 2.98 (s, 6H). 13 C NMR (100 MHz, CDCl 3 ) δ 162.19, 159.96, 158.22, 152.06, 129.96, 129.70, 124.44, 120.58, 111.61, 111.01, 104.01, 98.42, 58.71, 55.37, 50.39, 40.21, 40.06. HRMS (ESI+) m/z calc. for C 18 H 22 N 2 O 2 298.1681, found [M + H] + 299.1751; Rf = 0.63 (EtOAc/Hexane = 1:1, v / v ). N -Benzyl-1-phenylmethanimine ( 9 ), quantitative yield, brown oil. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.49 (s, 1H), 7.84–7.75 (m, 2H), 7.50–7.22 (m, 8H), 4.77 (s, J = 1.2 Hz, 2H); Rf = 0.65 (EtOAc/Hex = 1:1 v / v ) as reported . N -(2,4-Dimethoxybenzyl)-1-(furan-2-yl)methanimine ( 10 ), quantitative yield, dark brown oil. 1 H NMR (400 MHz, CDCl 3 ) δ 8.08 (s, 1H), 7.49 (s, 1H), 7.21–7.16 (m, 1H), 6.73 (d, J = 3.4 Hz, 1H), 6.49–6.44 (m, 3H), 4.74 (s, 2H), 3.80 (s, 3H), 3.79 (s, 3H); Rf = 0.36 (EtOAc/Hexane = 1:1, v / v ) as reported . N -(2,4-Dimethoxybenzyl)-1-(1 H -imidazol-5-yl)methanimine ( 11 ), quantitative yield, colorless amorphous solid. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.22 (s, 1H), 7.72–7.69 (m, 1H), 7.41 (s, 1H), 7.13 (d, J = 8.3 Hz, 1H), 6.56 (d, J = 2.4 Hz, 1H), 6.49 (dd, J = 8.3, 2.4 Hz, 1H), 4.57 (s, 2H), 3.77 (s, 3H), 3.75 (s, 3H). 13 C NMR (100 MHz, DMSO- d 6 ) δ 160.21, 158.35, 130.56, 120.00, 104.92, 98.69, 58.39, 55.82, 55.63. HRMS (ESI+) m/z calc. for C 13 H 15 N 3 O 2 245.1164, found [M + H] + 246.1234. Rf = 0.1 (EtOAc). 1-(Benzo[ b ]thiophen-2-yl)- N -benzylmethanimine ( 12 ), quantitative yield, yellow amorphous solid. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.76 (s, 1H), 7.98–7.8 (m 2H), 7.87 (s, 1H), 7.44–7.40 (m, 2H), 7.38–7.21 (m, 5H), 4.80 (s, 2H); Rf = 0.67 (EtOAc/Hexane = 1:1, v / v ) as reported . 1-(Benzo[ d ][1,3]dioxol-5-yl)- N -benzylmethanimine ( 13 ), quantitative yield, colorless amorphous solid. 1 H NMR (400 MHz, CDCl 3 ) δ 8.26 (t, J = 1.4 Hz, 1H), 7.41 (d, J = 1.4 Hz, 1H), 7.35–7.22 (m, 5H), 7.14 (dd, J = 8.0, 1.6 Hz, 1H), 6.82 (d, J = 7.9 Hz, 1H), 5.98 (s, 2H), 4.77 (s, 2H); Rf = 0.60 (EtOAc/Hexane = 1:1, v / v ) as reported . 1-(Benzo[d][1,3]dioxol-5-yl)- N -(2,4-dimethoxybenzyl)methanimine ( 14 ), quantitative yield, colorless amorphous solid. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.27 (s, 1H), 7.30 (d, J = 1.5 Hz, 1H), 7.20 (dd, J = 7.9, 1.5 Hz, 1H), 7.12 (d, J = 8.3 Hz, 1H), 6.97 (d, J = 8.0 Hz, 1H), 6.56 (d, J = 2.4 Hz, 1H), 6.49 (dd, J = 8.3, 2.4 Hz, 1H), 6.07 (s, 2H), 4.60 (s, 2H), 3.78 (s, 3H), 3.75 (s, 3H); Rf = 0.65 (EtOAc/Hexane = 1:1, v / v ) as reported . N -Benzyl-3-methylbutan-1-imine ( 15 ), quantitative yield, light orange oil. 1 H NMR (400 MHz, CDCl 3 ) δ 7.38–7.25 (m, 5H), 6.32 (s, 1H), 3.90 (s, 2H), 2.03–1.81 (m, 3H), 0.99–0.76 (m, 6H). 13 C NMR (100 MHz, CDCl 3 ) δ 179.39, 138.07, 128.76, 127.95, 127.81, 45.70, 44.53, 25.90, 22.58. HRMS (ESI+) m/z calc. for C 12 H 17 N 175.1361, found [M + H] + 176.1435. Rf = 0.65 (EtOAc/Hex = 1:1, v / v ). N -(2,4-Dimethoxybenzyl)-3-methylbutan-1-imine ( 16 ), quantitative yield, yellow oil. 1 H NMR (400 MHz, CDCl 3 ) δ 7.15 (d, J = 8.1 Hz, 1H), 6.73 (s, 1H), 6.46–6.38 (m, 2H), 3.86 (s, 2H), 3.81 (s, 3H), 3.79 (s, 3H), 2.00–1.88 (m, 3H), 0.88 (s, 3H), 0.87 (s, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 179.20, 160.92, 158.55, 130.46, 118.49, 104.03, 98.55, 55.38, 55.29, 45.95, 40.20, 25.93, 22.64. Rf = 0.63 (EtOAc/Hex = 1:1, v / v ). N -(2,4-Dimethoxybenzyl)heptan-1-imine ( 17 ), quantitative yield, orange oil. 1 H NMR (400 MHz, CDCl 3 ) δ 9.48 (s, 1H), 7.18–7.15 (m, 1H), 6.46–6.39 (m, 2H), 3.92 (s, 2H), 3.84–3.73 (m, 6H), 1.93–1.79 (m, 4H), 1.51–1.20 (m, 6H), 0.95–0.82 (m, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 176.84, 161.52, 158.65, 131.30, 114.65, 104.24, 98.38, 55.32, 55.27, 39.00, 22.73, 14.03. HRMS (ESI+) m/z calc. for C 16 H 25 NO 2 263.1885, found [M + H] + 264.1954. Rf = 0.85 (EtOAc/Hex = 1:1, v / v ). 4.3. General Procedures for the Synthesis of Ketene Precursors 4.3.1. General Procedure for the Synthesis of Acid Chloride ( 18 ) N -phthaloylglycine (2.00 g, 9.75 mmol, 1 EQ) was dissolved in dry dichloromethane (10 mL), and the solution was cooled to 0 °C using an ice bath before oxalyl chloride (0.95 mL, 10.73 mmol, 1.1 EQ) was added dropwise over 30 min. Upon complete addition, the reaction mixture was stirred at 0 °C for an additional 2 h, and the solvent was removed under reduced pressure without heating. The acyl chlorides thus obtained were used in the subsequent step without further purification. 2-(1,3-Dioxoisoindolin-2-yl)acetyl chloride ( 18 ), quantitative yield, yellow amorphous solid. 1 H NMR (400 MHz, CDCl 3 ) 7.95–7.89 (m, 1H), 7.82–7.76 (m, 1H), 4.83 (s, 1H); as reported . 4.3.2. General Procedure for the Synthesis of Diazoketone ( 31 ) N -Benzyloxycarbonylglycine (2.09 g, 10.0 mmol, 1 EQ) was dissolved in dry tetrahydrofuran (20 mL), and the resultant solution was cooled to −20 °C using a sodium chloride ice bath before triethylamine (1.39 mL, 10.0 mmol, 1 EQ) was added in one portion. Ethyl chloroformate (1.91 mL, 10.0 mmol, 1 EQ) was then added dropwise, and the reaction mixture was stirred for another 1 h. The white precipitate formed was removed by filtration. To the filtrate were slowly added dry acetonitrile (80 mL) (4:1 solution in THF) and (trimethylsilyl)diazomethane (2.0 M solution in hexane, 10 mL, 20.0 mmol, 2 EQ). The resultant reaction mixture was then stirred at 4 °C for 24–48 h. The reaction was quenched by the addition of diethyl ether and 10% (m/m) aqueous citric acid. The organic phase was then washed with saturated aqueous NaHCO 3 and brine. The organic layer was dried over Na 2 SO 4, and the solvents were evaporated. The diazoketone was purified by silica gel column chromatography using EtOAc:Hex = 1:1, v / v as eluent. Benzyl (3-diazo-2-oxopropyl)carbamate ( 31 ), quantitative yield, transparent amorphous solid. 1 H NMR (400 MHz, CDCl 3 ) δ (ppm) = 7.42–7.31 (m, 5H), 5.46 (s, 1H), 5.39 (s, 1H), 5.13 (s, 2H), 3.99 (s, 1H); as reported . 4.3.3. General Procedure for the Synthesis of Mixed Anhydride In a flame-dried flask, N -( tert -butoxycarbonyl)glycine (3.00 g, 17.13 mmol, 1 EQ) was dissolved in dry tetrahydrofuran (20 mL) and placed under an argon atmosphere. The solution was cooled to −60 °C, and triethylamine (2.62 mL, 18.84 mmol, 1.1 EQ) was added in one portion. Then ethyl chloroformate (2.13 mL, 22.27 mmol, 1.3EQ) was added dropwise over a period of 30 min. After the complete addition of the reagent, the reaction mixture was stirred at −40 °C for another 2 h. The resultant reaction mixture was then directly used in the next step without any further purification. The same reaction conditions were used for the synthesis of 2-(((benzyloxy)carbonyl)amino)acetic anhydride from ((benzyloxy)carbonyl)glycine. 4.4. General Procedure for the Synthesis of Monocyclic Beta Lactam Core I ( 19 – 30 ) Schiff base (1 EQ) was dissolved in dry toluene (0.1–0.2 mmol/mL) in a flame-dried flask and placed under an argon atmosphere. Triethylamine (2.5 EQ) was then added in one portion, and the resultant solution was heated to 80 °C, before 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride (1.3 EQ), dissolved in in dry toluene, was added dropwise over a period of 30 min. Upon complete addition, the reaction was stirred at 80 °C for a further 1.5–3.5 h. The reaction mixture was then cooled to room temperature, and the volatiles were removed in vacuo. The solid residue thus obtained was redissolved in ethyl acetate. The organic phase was washed with 10% aq. citric acid solution, saturated NaHCO 3, and brine. The organic phase was dried (Na 2 SO 4 ), filtered, then concentrated in vacuo. Some cyclized ß-lactams were purified by silica gel column chromatography using EtOAc: Hex as eluent. 2-(1-Benzyl-2-oxo-4-(4-(trifluoromethyl)phenyl)azetidin-3-yl)isoindoline-1,3-dione ( 19 ), yield: 51%, colorless amorphous solid. The reaction was carried out according to General Procedure I using N -benzyl-1-(4-(trifluoromethyl)phenyl)methanimine ( 1 ), (1.32 g, 5 mmol, 1.0 EQ), triethylamine (1.74 mL, 12.5 mmol, 2.5 EQ) and 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride (1.45 g, 6.5 mmol, 1.3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 2:1 ( v / v ) as eluent. 1 H NMR (400 MHz, DMSO- d 6 ) δ 7.80–7.70 (m, 4H), 7.51 (d, J = 8.2 Hz, 2H), 7.39–7.28 (m, 7H), 5.77 (d, J = 5.4 Hz, 1H), 5.18 (d, J = 5.4 Hz, 1H), 4.82 (d, J = 15.4 Hz, 1H), 4.47 (d, J = 15.4 Hz, 1H). 13 C NMR (100 MHz, CDCl 3 ) δ 166.73, 163.56, 137.29, 134.53, 134.42, 131.04, 129.09, 128.64, 128.26, 127.79, 125.50, 125.47, 123.59, 60.16, 59.84, 45.83. HRMS (ESI+) m/z calc. for C 25 H 17 F 3 N 2 O 3 450.1191, found [M + H] + 451.1260. Rf = 0.42 (EtOAc/n-Hex; 2:1, v / v ). 2-(1-(2,4-Dimethoxybenzyl)-2-oxo-4-(4-(trifluoromethyl)phenyl)azetidin-3-yl)isoindoline-1,3-dione ( 20 ), yield: 55%, light brown amorphous solid. The reaction was carried out according to General Procedure I using N -(2,4-dimethoxybenzyl)-1-(4-(trifluoromethyl)phenyl)methanimine ( 2 ), (3.23 g, 10 mmol, 1.0 EQ), triethylamine (3.48 mL, 25 mmol, 2.5 EQ) and 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride ( 18 ) (2.91 g, 13 mmol, 1.3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 7.71–7.60 (m, 4H), 7.44 (d, J = 8.2 Hz, 2H), 7.33 (d, J = 8.1 Hz, 2H), 7.15 (d, J = 8.3 Hz, 1H), 6.43 (dd, J = 8.3, 2.4 Hz, 1H), 6.37 (d, J = 2.3 Hz, 1H), 5.46 (d, J = 5.4 Hz, 1H), 4.90 (d, J = 14.3 Hz, 1H), 4.84 (d, J = 5.4 Hz, 1H), 4.30 (d, J = 14.3 Hz, 1H), 3.79 (s, 3H), 3.56 (s, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 166.77, 163.44, 161.18, 158.59, 138.33, 134.33, 131.59, 131.07, 127.60, 125.17, 125.13, 123.49, 115.03, 104.34, 98.33, 60.65, 59.54, 55.41, 55.00, 40.82. HRMS (ESI+) m/z calc. for C 27 H 21 F 3 N 2 O 5 510.1403, found [M + H] + 511.1470. Rf = 0.31 (EtOAc/n-Hex; 1:1, v / v ). 4-(1-(2,4-Dimethoxybenzyl)-3-(1,3-dioxoisoindolin-2-yl)-4-oxoazetidin-2-yl)benzonitrile ( 21 ), yield: 48%, colorless amorphous solid. The reaction was carried out according to General Procedure I using 4-(((2,4-dimethoxybenzyl)imino)methyl)benzonitrile ( 3 ) (0.32 g, 1.15 mmol, 1 EQ), triethylamine (0.40 mL, 2.87 mmol, 2.5 EQ) and 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride ( 18 ) (0.33 g, 1.50 mmol, 1.3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 2:1 ( v / v ) as eluent. 1 H NMR (400 MHz, DMSO- d 6 ) δ 7.80–7.72 (m, 4H), 7.63 (d, J = 8.4 Hz, 2H), 7.28 (d, J = 8.2 Hz, 2H), 7.21–7.17 (m, 1H), 6.49–6.46 (m, 2H), 5.58 (d, J = 5.5 Hz, 1H), 5.01 (d, J = 5.4 Hz, 1H), 4.61 (d, J = 14.5 Hz, 1H), 4.34 (d, J = 14.6 Hz, 1H), 3.74 (s, 3H), 3.57 (s, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ = 166.69, 163.32, 161.26, 158.57, 139.89, 134.49, 131.99, 131.65, 130.97, 127.88, 123.58, 118.41, 114.87, 111.87, 104.39, 98.39, 60.72, 59.57, 55.44, 55.02, 40.97. MS (ESI+, m/z), 468.4 ([M + H] + ). Rf = 0.24 (EtOAc/n-Hex; 1:1, v / v ). 2-(2-(4-Bromo-3-fluorophenyl)-1-(2,4-dimethoxybenzyl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 22 ), yield: 51%, pale yellow amorphous solid. The reaction was carried out according to General Procedure I using 1-(3-bromo-4-fluorophenyl)- N -(2,4-dimethoxybenzyl)methanimine ( 4 ) (0.53 g, 1.5 mmol, 1 EQ), triethylamine (0.52 mL, 3.75 mmol, 2.5 EQ) and 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride ( 18 ) (0.44 g, 1.95 mmol, 1.3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 2:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 7.73–7.64 (m, 4H), 7.37 (dd, J = 6.5, 2.2 Hz, 1H), 7.17–7.10 (m, 2H), 6.91 (t, J = 8.4 Hz, 1H), 6.43 (dd, J = 8.3, 2.4 Hz, 1H), 6.37 (d, J = 2.3 Hz, 1H), 5.40 (d, J = 5.3 Hz, 1H), 4.82 (d, J = 14.3 Hz, 1H), 4.74 (d, J = 5.4 Hz, 1H), 4.30 (d, J = 14.3 Hz, 1H), 3.80 (s, 3H), 3.61 (s, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 166.77, 163.36, 161.19, 158.57, 134.36, 132.52, 131.60, 131.51, 131.47, 131.14, 127.91, 127.84, 123.55, 116.33, 116.10, 115.03, 104.37, 98.37, 60.22, 59.61, 55.43, 55.10, 40.71. HRMS (ESI+) m/z calc. for C 26 H 20 BrFN 2 O 5 538.0540, found [M + H] + 539.0606. Rf = 0.56 (EtOAc/n-Hex; 2:1, v / v ). 2-(1-Benzyl-2-(4-nitrophenyl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 23 ), yield: 47%, pale yellow amorphous solid. The reaction was carried out according to General Procedure I using N -benzyl-1-(4-nitrophenyl)methanimine ( 5 ) (2.10 g, 10 mmol, 1 EQ), triethylamine (3.48 mL, 25 mmol, 2.5 EQ) and 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride ( 18 ) (2.91 g, 13 mmol, 1.3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 2:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 8.05 (d, J = 8.8 Hz, 2H), 7.74–7.60 (m, 4H), 7.39 (d, J = 8.7 Hz, 2H), 7.36–7.22 (m, 5H), 5.57 (d, J = 5.5 Hz, 1H), 5.06 (d, J = 14.8 Hz, 1H), 4.93 (d, J = 5.4 Hz, 1H), 4.26 (d, J = 14.8 Hz, 1H). 13 C NMR (100 MHz, CDCl 3 ) δ 166.64, 163.40, 147.84, 140.76, 134.59, 134.35, 130.93, 129.16, 128.66, 128.40, 128.33, 123.72, 123.70, 60.16, 59.88, 46.08. HRMS (ESI+) m/z calc. for C 24 H 17 N 3 O 5 427.1168, found [M + H] + 433.1386. Rf = 0.25 (EtOAc/n-Hex; 1:1, v / v ). 2-(1-(2,4-Dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 24 ), yield: 51%, pale yellow amorphous solid. The reaction was carried out according to General Procedure I using N -(2,4-dimethoxybenzyl)-1-(4-nitrophenyl)methanimine ( 6 ) (3.00 g, 10 mmol, 1 EQ), triethylamine (3.48 mL, 25 mmol, 2.5 EQ) and 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride ( 18 ) (2.91 g, 13 mmol, 1.3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 2:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 8.04 (d, J = 8.8 Hz, 2H), 7.72–7.60 (m 4H), 7.39 (d, J = 8.6 Hz, 2H), 7.16 (d, J = 8.3 Hz, 1H), 6.43 (dd, J = 8.3, 2.3 Hz, 1H), 6.36 (d, J = 2.3 Hz, 1H), 5.49 (d, J = 5.5 Hz, 1H), 4.91 (s, J = 14.3 Hz, 1H), 4.88 (d, J = 5.3 Hz, 1H), 4.33 (d, J = 14.3 Hz, 1H), 3.79 (s, 3H), 3.56 (s, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 166.67, 163.28, 161.29, 158.57, 147.57, 141.91, 134.50, 131.66, 130.96, 128.10, 123.64, 123.42, 114.82, 104.43, 98.41, 60.58, 59.58, 55.44, 55.06, 41.02. HRMS (ESI+) m/z calc. for C 26 H 21 N 3 O 7 487.1380, found [M + H] + 488.1443. Rf = 0.42 (EtOAc/n-Hex; 2:1, v / v ). 2-(1-(2,4-Dimethoxybenzyl)-2-(4-(methylsulfonyl)phenyl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 25 ), yield: 45%, pale yellow amorphous solid. The reaction was carried out according to General Procedure I using N -(2,4-dimethoxybenzyl)-1-(4-(methylsulfonyl)phenyl)methanimine ( 7 ) (1.00 g, 3.00 mmol, 1 EQ), triethylamine (1.04 mL, 7.50 mmol, 2.5 EQ) and 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride ( 18 ) (0.87 g, 3.90 mmol, 1.3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 2:1 ( v / v ) as eluent. 1 H NMR (400 MHz, DMSO- d 6 ) δ 7.86–7.69 (m, 4H), 7.68 (d, J = 8.2 Hz, 2H), 7.35 (d, J = 8.3 Hz, 2H), 7.20 (d, J = 8.8 Hz, 1H), 6.51–6.45 (m, 2H), 5.59 (d, J = 5.5 Hz, 1H), 5.03 (d, J = 5.4 Hz, 1H), 4.64 (d, J = 14.5 Hz, 1H), 4.34 (d, J = 14.6 Hz, 1H), 3.74 (s, 3H), 3.58 (s, 3H), 2.98 (s, 3H). 13 C NMR (100 MHz, DMSO- d 6 ) δ 166.80, 163.35, 161.07, 158.68, 141.06, 140.45, 135.42, 131.54, 130.78, 128.35, 126.90, 123.79, 115.39, 105.13, 98.72, 60.68, 59.76, 55.69, 55.64, 43.71, 41.04. HRMS (ESI+) m/z calc. for C 27 H 24 N 2 O 7 S 520.1304, found [M + H] + 521.1378. Rf = 0.36 (EtOAc/n-Hex; 2:1, v / v ). 2-(1-(2,4-Dimethoxybenzyl)-2-(furan-2-yl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 26 ), yield: 51%, colorless amorphous solid. The reaction was carried out according to General Procedure I using N -(2,4-dimethoxybenzyl)-1-(furan-2-yl)methanimine ( 10 ) (1.23 g, 5 mmol, 1 EQ), triethylamine (1.74 mL, 12.5 mmol, 2.5 EQ) and 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride ( 18 ) (1.45 g, 6.5 mmol, 1.3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 2:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 7.81–7.63 (m, 4H), 7.21–7.18 (m, 1H), 7.15 (d, J = 8.1 Hz, 1H), 6.48–6.40 (m, 2H), 6.29 (d, J = 3.3 Hz, 1H), 6.18 (dd, J = 3.3, 1.8 Hz, 1H), 5.42 (d, J = 5.0 Hz, 1H), 4.84 (d, J = 5.0 Hz, 1H), 4.80 (d, J = 14.5 Hz, 1H), 4.22 (d, J = 14.5 Hz, 1H), 3.80 (s, 3H), 3.71 (s, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 166.73, 163.17, 161.01, 158.61, 148.22, 142.85, 134.20, 131.45, 131.38, 123.51, 115.39, 110.56, 109.87, 104.20, 98.41, 59.03, 55.41, 55.29, 55.12, 40.18. HRMS (ESI+) m/z calc. for C 24 H 20 N 2 O 6 432.1321, found [M + H] + 433.1386. Rf = 0.23 (EtOAc/n-Hex; 2:1, v / v ). 2-(1-(2,4-Dimethoxybenzyl)-2-(1 H -imidazol-5-yl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 27 ), yield: 33%, light brown amorphous solid. The reaction was carried out according to General Procedure I using N -(2,4-dimethoxybenzyl)-1-(1 H -imidazol-5-yl)methanimine ( 11 ) (1.19 g, 7.7 mmol, 1 EQ), triethylamine (1.35 mL, 15.4 mmol, 2 EQ) and 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride ( 18 ) (2.23 g, 10 mmol, 1.3EQ). Product was purified by silica gel column chromatography using DKM: MeOH = 15:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 9.69 (s, 1H), 7.90–7.79 (m, 4H), 7.57–7.47 (m, 1H), 6.85 (d, J = 8.2 Hz, 1H), 6.38–6.23 (m, 2H), 4.92 (d, J = 16.7 Hz, 1H), 4.78 (d , J = 7.5 Hz, 2H), 4.74 (d, J = 7.3 Hz, 1H), 4.62 (d, J = 16.6 Hz, 1H), 3.81 (s, 3H), 3.75 (s, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 185.51, 184.92, 167.93, 161.04, 158.25, 140.99, 139.31, 136.01, 134.21, 132.14, 123.59, 104.24, 98.57, 63.70, 55.41, 55.38, 55.21, 39.94. HRMS (ESI+) m/z calc. for C 23 H 20 N 4 O 5 432.1434, found [M + H] + 433.1500. Rf = 0.42 (DKM/MeOH; 9:1, v / v ). 2-(1-Benzyl-2-isobutyl-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 28 ), yield: 31%, colorless amorphous solid. The reaction was carried out according to General Procedure I using N -Benzyl-3-methylbutan-1-imine ( 15 ) (2 g, 11.6 mmol, 1 EQ), triethylamine (3.21 mL, 23 mmol, 2 EQ) and 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride ( 18 ) (3.1 g, 15 mmol, 1.3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 7.95–7.66 (m, 4H), 7.34–7.14 (m, 5H), 6.41 (d, J = 13.8 Hz, 1H), 5.26 (dd, J = 13.8, 7.3 Hz, 1H), 4.84–4.44 (m, 4H), 2.38–2.22 (m, 1H), 0.98 (d, J = 6.7 Hz, 3H), 0.93 (d, J = 6.7 Hz, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 167.98, 164.64, 136.60, 134.07, 132.27, 128.52, 127.37, 125.63, 123.56, 60.38, 48.50, 39.81, 29.57, 22.85, 22.68. HRMS (ESI+) m/z calc. for C 22 H 22 N 2 O 3 362.1630, found [M + H] + 363.1694. Rf = 0.54 (EtOAc/n-Hex; 1:1, v / v ). 2-(1-(2,4-Dimethoxybenzyl)-2-isobutyl-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 29 ), yield: 25%, colorless amorphous solid. The reaction was carried out according to General Procedure I using N -(2,4-Dimethoxybenzyl)-3-methylbutan-1-imine ( 16 ) (2.73 g, 11.6 mmol, 1 EQ), triethylamine (3.21 mL, 23 mmol, 2 EQ) and 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride ( 18 ) (3.36 g, 15 mmol, 1.3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:1 ( v / v ) as eluent. 1 H NMR (400 MHz, DMSO- d 6 ) δ 7.99–7.82 (m, 4H), 7.01–6.93 (m, 1H), 6.84–6.74 (m, 1H), 6.63–6.51 (m, 1H), 6.50–6.44 (m, 1H), 5.14–5.00 (m, 1H), 4.83–4.55 (m, 4H), 3.90–3.66 (m, 6H), 2.37–2.13 (m, 1H), 0.95 (d, J = 6.7 Hz, 3H), 0.88 (d, J = 6.7 Hz, 3H). 13 C NMR (100 MHz, DMSO- d 6 ) δ 168.00, 165.35, 160.05, 157.83, 135.21, 132.07, 127.88, 123.80, 120.61, 116.68, 105.05, 98.63, 65.35, 55.86, 55.62, 43.67, 41.94, 29.24, 23.36. HRMS (ESI+) m/z calc. for C 24 H 26 N 2 O 5 422.1842, found [M + H] + 423.1910. Rf = 0.45 (EtOAc/n-Hex; 1:1, v / v ). 2-(1-(2,4-Dimethoxybenzyl)-2-hexyl-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 30 ), yield: 11%, brown oil. The reaction was carried out according to General Procedure I using N -(2,4-Dimethoxybenzyl)heptan-1-imine ( 17 ) (2.31 g, 8.76 mmol, 1EQ), triethylamine (3.05 mL, 21.9 mmol, 2.5 EQ) and 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride ( 18 ) (2.53 g, 11.3 mmol, 1.3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:1 ( v / v ) as eluent. 1H NMR (400 MHz, CDCl 3 ) δ 7.93–7.66 (m, 4H), 7.07–6.94 (m, 1H), 6.59–6.39 (m, 2H), 5.32–5.15 (m, 1H), 4.80–4.69 (m, 1H), 4.67–4.54 (m, 1H), 3.92–3.87 (m, 2H), 3.84–3.75 (m, 6H), 2.04–1.93 (m, 1H), 1.39–1.04 (m, 8H), 0.92–0.79 (m, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 168.03, 164.42, 160.42, 157.82, 134.02, 132.33, 123.53, 119.31, 113.91, 104.26, 98.29, 60.42, 55.39, 55.31, 42.61, 39.88, 31.14, 30.29, 29.37, 28.29, 22.46, 14.02. HRMS (ESI+) m/z calc. for C 26 H 30 N 2 O 5 450.2155, found [M + H] + 451.2229. Rf = 0.42 (EtOAc/n-Hex; 1:1, v / v ). 4.5. General Procedure for the Synthesis of Monocyclic Beta Lactam Core II ( 32 – 34 ) A solution of an appropriate Schiff base (2 EQ) and diazoketone (1 EQ) in 1,2 dimethoxyethane (3 mL) was stirred for 20–30 min at 180 °C in a microwave reactor. The volatiles were then removed in vacuo, and the crude product thus obtained was purified by silica gel column chromatography using EtOAc: Hexane (1:1) as an eluent. Benzyl ((1-benzyl-2-oxo-4-phenylazetidin-3-yl)methyl)carbamate ( 32 ), yield: 31%, brown oil. The reaction was carried out according to General Procedure II using N -benzyl-1-phenylmethanimine ( 9 ) (250 mg, 1.28 mmol, 2 EQ) and benzyl (3-diazo-2-oxopropyl)carbamate ( 31 ) (179 mg, 0.64 mmol, 1 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 7.42 (d, J = 1.6 Hz, 1H), 7.40–7.30 (m, 10H), 7.25–7.21 (m, 3H), 7.12 (dd, J = 6.9, 2.5 Hz, 2H), 5.18–5.13 (m, 1H), 5.11 (d, J = 12.3 Hz, 1H), 4.98 (d, J = 12.3 Hz, 1H), 4.83 (d, J = 15.0 Hz, 1H), 4.28 (d, J = 2.0 Hz, 1H), 3.76 (d, J = 14.7 Hz, 1H), 3.64–3.59 (m, 1H), 3.20–3.15 (m, 1H). 13 C NMR (100 MHz, CDCl 3 ) δ 166.69, 156.61, 137.00, 135.31, 134.51, 128.99, 128.79, 128.54, 128.39, 128.33, 128.17, 128.01, 127.73, 126.51, 69.10, 60.36, 57.72, 48.85, 44.57. HRMS (ESI+) m/z calc. for C 25 H 24 N 2 O 3 400.1787, found [M + H] + 401.1856. Rf = 0.27. (EtOAc/n-Hex; 1:1, v / v ). Benzyl ((2-(benzo[b]thiophen-2-yl)-1-benzyl-4-oxoazetidin-3-yl)methyl)carbamate ( 33 ), yield: 35%, yellow oil. The reaction was carried out according to General Procedure II using 1-(benzo[ b ]thiophen-2-yl)- N -benzylmethanimine ( 12 ) (250 mg, 1.00 mmol, 2 EQ) and benzyl (3-diazo-2-oxopropyl)carbamate ( 31 ) (117 mg, 0.50 mmol, 1 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:2 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 7.81 (d, J = 8.4 Hz, 1H), 7.70 (d, J = 8.6 Hz, 1H), 7.38–7.30 (m, 13H), 5.13 (s, 2H), 4.98 (d, J = 12.2 Hz, 1H), 4.87 (d, J = 15.1 Hz, 1H), 4.67 (d, J = 1.0 Hz, 1H), 4.33 (s, 1H), 3.86 (d, J = 15.1 Hz, 1H). 13 C NMR (100 MHz, CDCl 3 ) δ 167.35, 156.67, 141.88, 139.57, 139.44, 136.25, 135.16, 128.88,128.80, 128.78, 128.75, 128.56, 128.38, 128.21, 128.14, 128.02, 127.87, 124.82, 124.65, 123.68, 123.26, 122.60, 66.99, 60.92, 54.14, 52.36, 44.81. HRMS (ESI+) m/z calc. for C 27 H 24 N 2 O 3 S 456.1508, found [M + H] + 457.1579. Rf = 0.24 (EtOAc/n-Hex; 1:2, v / v ). Benzyl ((2-(benzo[d][1,3]dioxol-5-yl)-1-benzyl-4-oxoazetidin-3-yl)methyl)carbamate ( 34 ), yield: 29%, colorless amorphous solid. The reaction was carried out according to General Procedure II using 1-(benzo[d][1,3]dioxol-5-yl)- N -benzylmethanimine ( 13 ) (250 mg, 1.05 mmol, 2 EQ) and benzyl (3-diazo-2-oxopropyl)carbamate ( 31 ) (123 mg, 0.53 mmol, 1 EQ) Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:1 ( v / v ) as eluent. 1 H NMR (400 MHz, DMSO- d 6 ) δ 7.51 (t, J = 6.0 Hz, 1H), 7.38–7.25 (m, 9H), 7.16 (d, J = 6.7 Hz, 2H), 6.85 (d, J = 7.9 Hz, 1H), 6.79 (s, 1H), 6.67 (d, J = 7.9 Hz, 1H), 6.01 (d, J = 1.3 Hz, 2H), 5.02 (d, J = 12.6 Hz, 1H), 4.96 (d, J = 12.6 Hz, 1H), 4.65 (d, J = 15.6 Hz, 1H), 4.31 (d, J = 1.8 Hz, 1H), 3.81 (d, J = 15.6 Hz, 1H), 3.70–3.52 (m, 1H), 3.40–3.34 (m, 1H), 3.09 (ddd, J = 7.3, 5.1, 1.9 Hz, 1H). 13 C NMR (100 MHz, DMSO- d 6 ) δ 167.59, 156.79, 148.26, 147.60, 137.54, 136.50, 132.07, 129.07, 128.79, 128.23, 128.07, 127.81, 120.55, 108.85, 106.72, 101.59, 65.81, 60.66, 58.45, 51.93, 44.12. HRMS (ESI+) m/z calc. for C 26 H 24 N 2 O 5 444.1685, found [M + H] + 445.1752. Rf = 0.31 (EtOAc/n-Hex; 1:1, v / v ). 4.6. General Procedure for the Synthesis of Monocyclic Beta Lactam Core III ( 35 – 37 ) To a solution of an appropriate Schiff base (1 EQ) in dry dichloromethane (2 mL/mmol) was added dry tetrahydrofuran (5 mL/mmol), and the reaction mixture was cooled to −60 °C before mixed acid anhydride (1.3 EQ) was added slowly. Next, triethylamine (1.5 EQ) was added dropwise over a period of 30 min. Upon complete addition, the reaction mixture was allowed to warm to room temperature with stirring overnight. The volatiles were removed in vacuo, and the solid residue thus obtained dissolved in ethyl acetate. The organic phase was washed with 0.1 M HCl (aq), saturated NaHCO 3 (aq) and brine, dried (Na 2 SO 4 ), filtered, then concentrated in vacuo. The crude product thus obtained was then purified by silica gel column chromatography using EtOAc: Hex as eluent or recrystallized from methyl tert-butyl ether. tert -butyl (2-(4-cyanophenyl)-1-(2,4-dimethoxybenzyl)-4-oxoazetidin-3-yl)carbamate ( 35 ), yield: 11%, colorless amorphous solid. The reaction was carried out according to General Procedure III using 4-(((2,4-dimethoxybenzyl)imino)methyl)benzonitrile ( 3 ), (14.16 g, 50.5 mmol, 1 EQ), N -( tert -butoxycarbonyl)glycine (11.5 g, 65.6 mmol, 1.3 EQ) and triethylamine (10.5 mL, 75.7 mmol, 1.5 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:2 ( v / v ) as eluent. 1 H NMR (400 MHz, DMSO- d6 ) δ 7.74 (d, J = 8.3 Hz, 2H), 7.52 (d, J = 8.4 Hz, 1H), 7.33 (d, J = 8.2 Hz, 2H), 7.04 (d, J = 8.0 Hz, 1H), 6.46–6.39 (m, 2H), 4.93 (dd, J = 8.4, 5.0 Hz, 1H), 4.70 (d, J = 5.0 Hz, 1H), 4.43 (d, J = 14.5 Hz, 1H), 4.09–3.97 (m, 1H), 3.72 (s, 3H), 3.56 (s, 3H), 1.14 (s, 9H). 13 C NMR (100 MHz, DMSO- d 6 ) δ 166.08, 160.96, 158.59, 155.06, 141.81, 132.00, 131.42, 129.23, 119.33, 115.50, 110.46, 105.06, 98.63, 78.80, 62.76, 61.72, 60.22, 55.68, 55.60, 28.24. HRMS (ESI+) m/z calc. for C 24 H 27 N 3 O 5 437.1951, found [M + H] + 438.2015. Rf = 0.54 (EtOAc/n-Hex; 1:1, v / v ). Benzyl (1-(2,4-dimethoxybenzyl)-2-(4-(dimethylamino)phenyl)-4-oxoazetidin-3-yl)carbamate ( 36 ), yield: 31%, colorless amorphous solid. The reaction was carried out according to General Procedure III using 4-(((2,4-dimethoxybenzyl)imino)methyl)- N,N -dimethylaniline ( 8 ), (1.27 g, 4.25 mmol, 1 EQ), ((benzyloxy)carbonyl)glycine (1. 15 g, 5.52 mmol, 1.3 EQ,) and triethylamine (0.88 mL, 6.37 mmol, 1.5 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 2:3 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 7.39–7.28 (m, 4H), 7.24–6.93 (m, 5H), 6.76–6.60 (m, 2H), 6.48–6.36 (m, 2H), 5.71 (d, J = 47.4 Hz, 1H), 5.14 (dd, J = 21.9, 9.5 Hz, 1H), 4.94 (d, J = 12.0 Hz, 1H), 4.84 (d, J = 14.5 Hz, 1H), 4.18 (d, J = 20.5 Hz, 2H), 3.79 (app s, 6H), 2.98 (s, 6H). 13 C NMR (100 MHz, CDCl 3 ) δ 167.19, 160.87, 158.60, 153.11, 151.08, 131.78, 128.71, 128.52, 128.42, 128.22, 128.02, 127.72, 116.06, 112.03, 104.09, 98.45, 75.00, 67.19, 67.13, 55.41, 55.25, 48.33, 40.41, 38.28. HRMS (ESI+) m/z calc. for C 28 H 31 N 3 O 5 489.2264, found [M + H] + 490.2325. Rf = 0.45 (EtOAc/n-Hex; 1:1, v / v ). Benzyl (2-(benzo[d][1,3]dioxol-5-yl)-1-(2,4-dimethoxybenzyl)-4-oxoazetidin-3-yl)carbamate ( 37 ), yield: 33%, colorless amorphous solid. The reaction was carried out according to General Procedure III using 1-(benzo[d][1,3]dioxol-5-yl)- N -(2,4-dimethoxybenzyl)methanimine ( 14 ) (1.85 g, 6.18 mmol, 1 EQ), ((benzyloxy)carbonyl)glycine (1.68 g, 8.04 mmol, 1.3 EQ,) and triethylamine (1.29 mL, 9.27 mmol, 1.5 EQ). Product was precipitated from methyl tert- butyl ether. 1 H NMR (400 MHz, CDCl 3 ) δ 7.32–7.25 (m, 4H), 7.15 (d, J = 7.2 Hz, 1H), 7.02 (d, J = 8.1 Hz, 1H), 6.76 (d, J = 8.3 Hz, 1H), 6.62–6.58 (m, 2H), 6.39 (d, J = 7.2 Hz, 2H), 5.96 (s, 2H), 5.18 (dd, J = 9.4, 5.0 Hz, 1H), 4.98–4.94 (m, 2H), 4.70 (d, J = 14.4 Hz, 1H), 4.62 (d, J = 4.9 Hz, 1H), 3.99 (d, J = 14.4 Hz, 1H), 3.79 (s, 3H), 3.66 (s, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 166.42, 161.02, 158.56, 155.32, 148.15, 147.66, 135.99, 131.21, 128.46, 128.13, 127.86, 120.64, 115.31, 108.62, 107.02, 104.08, 101.27, 98.40, 77.35, 77.03, 76.72, 66.99, 61.82, 61.13, 55.40, 55.17, 45.81, 39.71, 8.63. HRMS (ESI+) m/z calc. for C 27 H 26 N 2 O 7 490.1740, found [M + H] + 491.1803. Rf = 0.28 (EtOAc/n-Hex; 1:1, v / v ). 4.7. General Procedure for the Synthesis of Phthalimide Deprotected ß Lactam ( 38 – 41 ) 4.7.1. Hydrazine Hydrate 3-Amino-4-substituted monocyclic ß-lactams with aromatic substituents: 2-(1-Benzyl-2-oxo-4-(4-(trifluoromethyl)phenyl)azetidin-3-yl)isoindoline-1,3-dione ( 19 ) (225 mg, 0.44 mmol, 1 EQ) was dissolved in dried methanol and placed under argon atmosphere. Hydrazine hydrate (0.046 mL, 0.76 mmol, 1.7 EQ) was added dropwise. The mixture was stirred for 2 h at room temperature. Then, the solvent was evaporated. Anhydrous methanol and 3 drops of concentrated aqueous HCl were added to the solid. After the solid was completely dissolved again, the solvent was evaporated. The solid was again dissolved in anhydrous methanol and stirred for 16 h at room temperature. The precipitate formed was filtered off and the solvent evaporated. The solid was dissolved in dichloromethane and washed with saturated aqueous NaHCO 3 . The aqueous phase was extracted three times with dichloromethane, the combined organic phases were dried over Na 2 SO 4 and the solvents evaporated. The deprotected amine was used directly in the next step without purification. 2-(2-((1-(2,4-Dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-yl)carbamoyl)benzoyl)hydrazin-1-ide ( 38 ), colorless amorphous solid. 1 H NMR (400 MHz, DMSO- d 6 ) δ 9.37 (s, 1H), 9.04 (d, J = 7.9 Hz, 1H), 8.15 (d, J = 8.8 Hz, 2H), 7.45 (d, J = 8.7 Hz, 2H), 7.39–7.27 (m, 2H), 7.20 (td, J = 7.5, 1.4 Hz, 1H), 7.09 (d, J = 8.0 Hz, 1H), 6.49–6.39 (m, 4H), 5.39 (dd, J = 7.8, 5.1 Hz, 1H), 4.91 (d, J = 5.0 Hz, 1H), 4.50 (d, J = 14.5 Hz, 1H), 4.30 (s, 1H), 4.17 (d, J = 14.4 Hz, 1H), 3.72 (s, 3H), 3.59 (s, 3H). MS (ESI + , m / z ) 518.1 ([M − H] − ). Rf = 0.05 (EtOAc). 1-Benzyl-2-oxo-4-(4-(trifluoromethyl)phenyl)azetidin-3-aminium chloride ( 39 ), yield: 91%, colorless amorphous solid. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.69 (s, 3H), 7.74 (d, J = 8.2 Hz, 2H), 7.59 (d, J = 8.2 Hz, 2H), 7.38–7.17 (m, 4H), 5.06 (d, J = 5.4 Hz, 1H), 4.92 (d, J = 5.4 Hz, 1H), 4.69 (d, J = 15.4 Hz, 1H), 4.17 (d, J = 15.4 Hz, 1H). 13 C NMR (100 MHz, DMSO- d 6 ) δ 163.03, 136.83, 135.32, 130.06, 129.63, 129.11, 128.75, 128.19, 125.84, 123.23, 58.90, 57.79, 45.10.HRMS (ESI+) m/z calc. for C 17 H 15 F 3 N 2 O 320.1136, found [M + H] + 321.1208. Rf = 0.61 (EtOAc). 1-(2,4-Dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-aminium chloride ( 40 ), yield: 94%, light brown oil. The reaction was carried out according to the General Procedure using 2-(1-(2,4-dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 24 ) (1.01 g, 2.1 mmol, 1 EQ) and hydrazine hydrate (0.213 mL, 3.5 mmol, 1.7 EQ). Product was used directly in the next step without purification. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.23–8.13 (m, 2H), 7.48–7.38 (m, 2H), 7.03 (dd, J = 11.5, 6.3 Hz, 1H), 6.48–6.37 (m, 2H), 4.68 (d, J = 5.1 Hz, 1H), 4.45 (d, J = 14.7 Hz, 1H), 4.43 (d, J = 5.3 Hz, 1H), 4.00 (d, J = 14.5 Hz, 1H), 3.71 (s, 3H), 3.56 (s, 3H). 13 C NMR (100 MHz, DMSO- d 6 ) δ 170.45, 160.89, 158.58, 147.23, 145.34, 131.29, 129.27, 123.55, 115.89, 105.07, 98.64, 71.70, 66.04, 62.52, 55.66, 55.61. HRMS (ESI+) m/z calc. for C 18 H 19 N 3 O 5 357.1325, found [M + H] + 358.1390. Rf = 0.51 (DCM/ iPrOH; 11:1, v / v ). 3-Amino-4-substituted monocyclic ß-lactams with aliphatic substituents: 2-(1-(2,4-Dimethoxybenzyl)-2-isobutyl-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 29 ) (200 mg, 0.47 mmol, 1 EQ) in was dissolved in dried methanol and placed under argon atmosphere. Hydrazine hydrate (0.088 mL, 1.42 mmol, 3 EQ) was added dropwise. The mixture was stirred for 2 h at room temperature. Then the solvent was evaporated, and the residue was redissolved in ethyl acetate. The organic phase was washed with saturated aqueous NaHCO 3 and brine, dried over Na 2 SO 4 and the solvent was evaporated. Mixture of isomers cis and trans 1-(2,4-dimethoxybenzyl)-2-isobutyl-4-oxoazetidin-3-aminium chloride ( 41 ), quantitative yield, colorless amorphous solid. 1 H NMR (400 MHz, CDCl 3 ) δ 7.34 (d, J = 14.6 Hz, 1H), 6.88 (d, J = 8.3 Hz, 1H), 6.80 (d, J = 8.4 Hz, 1H), 6.48–6.37 (m, 4H), 6.32 (d, J = 13.9 Hz, 1H), 5.09 (dd, J = 13.9, 7.3 Hz, 1H), 4.93 (dd, J = 14.6, 7.2 Hz, 1H), 4.78 (s, 2H), 4.55 (s, 2H), 3.83 (d, J = 6.3 Hz, 6H), 3.79 (d, J = 5.4 Hz, 6H), 3.64 (s, 2H), 3.43 (s, 2H), 2.34 (td, J = 13.7, 7.0 Hz, 1H), 2.26 (td, J = 13.5, 6.8 Hz, 1H), 1.02–0.95 (m, 6H), 0.95–0.89 (m, 6H). 13 C NMR (100 MHz, CDCl 3 ) δ 172.10, 171.86, 160.14, 159.82, 157.74, 157.36, 128.01, 126.41, 124.38, 123.70, 123.44, 119.97, 117.29, 116.10, 104.07, 104.01, 98.37, 98.22, 55.39, 55.34, 55.29, 55.26, 44.05, 43.74, 42.80, 42.13, 29.50, 29.45, 23.11, 22.92. MS (ESI+) m/z calc. for C 16 H 24 N 2 O 3 292.1787, found [M + H] + 293.1853. Rf = 0.07 (EtOAc). 4.7.2. Methylhydrazine 2-(1-(2,4-Dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-yl) isoindoline-1,3-dione ( 24 ) (250 mg, 0.51 mmol, 1 EQ) was dissolved in dry methanol and methylhydrazine (0.081 mL, 1.54 mmol, 3 EQ) was added. The reaction was stirred at room temperature. After 4 h, additional methylhydrazine (0.11 mL, 2.12 mmol, 4 EQ) was added and stirred overnight. As the reaction was not yet complete, further methylhydrazine (0.11 mL, 2.12 mmol, 4 EQ) was added, and the reaction was left at room temperature for an additional 72 h. The reaction was allowed to proceed to completion. The organic phase was washed with saturated NaHCO 3 solution and brine and dried over Na 2 SO 4 . The solvent was evaporated, and the product was purified by silica gel column chromatography using DCM:iPrOH = 11:1 as eluent. 4.7.3. Ethanolamine 2-(1-(2,4-Dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-yl) isoindoline-1,3-dione ( 24 ) (250 mg, 0.51 mmol, 1 EQ) was dissolved in ethyl acetate. Ethanolamine (0.46 mL, 7.7 mmol, 15 EQ) was added, and the reaction mixture was refluxed (80 °C) for 2 h. Then the reaction mixture was cooled to room temperature, and a saturated solution of NaHCO 3 and additional ethyl acetate were added. The organic phase was washed with brine and dried over Na 2 SO 4 . The solvent was evaporated, and the product was purified by column chromatography using DCM:iPrOH = 11:1 as eluent. 4.7.4. Ethylenediamine 2-(1-(2,4-Dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-yl) isoindoline-1,3-dione (24) (250 mg, 0.51 mmol, 1 EQ) was dissolved in ethyl acetate. Ethylendiamine solution (1 M in ethyl acetate; 0.67 mL in 10 mL of ethyl acetate, 10 mmol, 19.5 EQ) was added, and the reaction mixture was stirred overnight at room temperature. After 16 h saturated solution of NaHCO3 and additional ethyl acetate were added. The organic phase was washed with brine and dried over Na 2 SO 4 . The solvent was evaporated and the product purified by silica gel column chromatography using DCM:iPrOH = 11:1 as eluent. 4.8. General Procedure for the Synthesis of tert-Butyloxycarbonyl Protected ß Lactam ( 43 , 51 – 53 ) In a flame-dried flask, 3-amino ß-lactam (1 EQ) was dissolved in dry dichloromethane. Triethylamine (1.1 EQ), di- tert -butyl dicarbonate (1.5 EQ) and 4-(dimethylamino)pyridine (catalytic amount) were added, and the solution was stirred overnight at room temperature. The solvent was evaporated, and the crude product was purified by silica gel column chromatography using EtOAc: Hex as eluent. tert -butyl (1-benzyl-2-isobutyl-4-oxoazetidin-3-yl)carbamate ( 43 ), yield: 51%, colorless amorphous solid. The reaction was carried out according to General Procedure using 3-amino-1-(benzyl)-4-isobutylazetidin-2-one (160 mg, 0.69 mmol, 1 EQ), triethylamine (0.11 mL, 0.76 mmol, 1.1 EQ), di- tert -butyl dicarbonate (226 mg, 1.03 mmol, 1.5 EQ) and 4-(dimethylamino)pyridine (catalytic amount). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:2 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 7.37–7.08 (m, 5H), 5.60–5.41 (m, 1H), 5.21–4.93 (m, 1H), 4.88–4.65 (m, 2H), 4.17–3.91 (m, 2H), 2.39–2.19 (m, 1H), 1.52–1.38 (m, 9H), 0.98–0.90 (m, 6H). 13 C NMR (100 MHz, CDCl 3 ) δ 167.47, 155.82, 136.71, 128.52, 127.10, 123.16, 79.76, 47.86, 42.96, 29.54, 28.35, 27.91, 23.01, 22.74. HRMS (ESI+) m/z calc. for C 19 H 28 N 2 O 3 332.2100, found [M + H] + 333.2166. Rf = 0.63 (EtOAc/n-Hex; 1:1, v / v ). tert -butyl (1-benzyl-2-oxo-4-(4-(trifluoromethyl)phenyl)azetidin-3-yl)carbamate ( 51 ), yield: 47%, colorless amorphous solid. The reaction was carried out according to General Procedure using 1-benzyl-2-oxo-4-(4-(trifluoromethyl)phenyl)azetidin-3-aminium chloride ( 39 ) (143 mg, 0.4 mmol, 1 EQ), triethylamine (0.061 mL, 0.44 mmol, 1 EQ), di- tert -butyl dicarbonate (130 mg, 0.6 mmol, 1.5 EQ) and 4-(dimethylamino)pyridine (catalytic amount). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:2 ( v / v ) as eluent. 1 H NMR (400 MHz, DMSO- d 6 ) δ 7.64 (d, J = 8.1 Hz, 2H), 7.55 (d, J = 8.3 Hz, 1H), 7.39 (d, J = 8.1 Hz, 2H), 7.34–7.25 (m, 3H), 7.22 (d, J = 6.6 Hz, 2H), 5.05 (dd, J = 8.3, 5.0 Hz, 1H), 4.85 (d, J = 4.9 Hz, 1H), 4.66 (d, J = 15.4 Hz, 1H), 4.13 (d, J = 15.4 Hz, 1H), 1.39 (s, 9H). 13 C NMR (100 MHz, CDCl 3 ) δ 166.14, 154.27, 138.46, 134.44, 128.99, 128.53, 128.20, 127.83, 125.56, 125.54, 80.39, 62.46, 60.92, 45.15, 27.86. HRMS (ESI+) m/z calc. for C 22 H 23 F 3 N 2 O 3 420.1661, found [M + Na] + 443.1550. Rf = 0.58 (EtOAc/n-Hex; 1:1, v / v ). tert -butyl (1-(2,4-dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-yl)carbamate ( 52 ), yield: 60%, colorless amorphous solid. The reaction was carried out according to General Procedure using 1-(2,4-dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-aminium chloride ( 40 ) (143 mg, 0.4 mmol, 1 EQ), triethylamine (0.061 mL, 0.44 mmol, 1 EQ), di- tert -butyl dicarbonate (130 mg, 0.6 mmol, 1.5 EQ) and 4-(dimethylamino)pyridine (catalytic amount). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:2 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 8.18 (d, J = 8.6 Hz, 2H), 7.32 (d, J = 8.7 Hz, 2H), 7.04 (d, J = 8.3 Hz, 1H), 6.39 (dd, J = 8.3, 2.3 Hz, 1H), 6.33 (d, J = 2.3 Hz, 1H), 5.15 (dd, J = 8.1, 5.0 Hz, 1H), 4.86 (d, J = 8.1 Hz, 1H), 4.77 (d, J = 4.9 Hz, 1H), 4.69 (d, J = 14.3 Hz, 1H), 4.13 (d, J = 14.3 Hz, 1H), 3.78 (s, 3H), 3.57 (s, 3H), 1.19 (s, 9H). 13 C NMR (100 MHz, CDCl 3 ) δ 165.89, 161.27, 158.50, 154.33, 147.63, 143.08, 131.45, 128.10, 123.40, 114.79, 104.28, 98.36, 80.46, 62.34, 61.42, 55.42, 55.03, 40.41, 27.90. HRMS (ESI+) m/z calc. for C 23 H 27 N 3 O 7 457.1849, found [M + H] + 458.1917. Rf = 0.57 (EtOAc/n-Hex; 1:1, v / v ). tert -butyl (1-(2,4-dimethoxybenzyl)-2-isobutyl-4-oxoazetidin-3-yl)carbamate ( 53 ), yield: 54%, colorless amorphous solid. The reaction was carried out according to General Procedure using 3-amino-1-(2,4-dimethoxybenzyl)-4-isobutylazetidin-2-one ( 41 ) (138 mg, 0.47 mmol, 1 EQ), triethylamine (0.072 mL, 0.52 mmol, 1.1 EQ), di- tert -butyl dicarbonate (155 mg, 0.71 mmol, 1.5 EQ) and 4-(dimethylamino)pyridine (catalytic amount). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 6.88 (d, J = 8.3 Hz, 1H), 6.47–6.22 (m, 3H), 5.52 (br s, 1H), 5.20–4.90 (m, 1H), 4.80–4.50 (m, 2H), 4.17–3.92 (m, 2H), 3.85–3.75 (m, 6H), 2.38–2.17 (m, 1H), 1.49–1.44 (m, 9H), 0.96 (d, J = 6.7 Hz, 3H), 0.93 (d, J = 6.7 Hz, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 167.35, 159.93, 157.42, 154.13, 128.19, 123.12, 104.07, 104.01, 98.26, 83.65, 55.26, 43.29, 43.01, 42.77, 42.40, 29.47, 27.94, 22.79. HRMS (ESI+) m/z calc. for C 21 H 32 N 2 O 5 392.2311, found [M + H] + 393.2384. Rf = 0.63 (EtOAc/n-Hex; 1:1, v / v ). 4.9. General Procedure for the Synthesis of N1-Benzyl Deprotected ß Lactam ( 42 – 44 ) Birch Reduction In a flame-dried flask, Na dispersion in mineral oil (25 wt%, TCI, 6 EQ) and 15-crown-5 (6 EQ) were dissolved in dry tetrahydrofuran. The solution was warmed to room temperature under argon and stirred vigorously for 5 min. Then, the reaction mixture was cooled to 0 °C before a solution of ß-lactam (1 EQ), and isopropanol (6 EQ) in tetrahydrofuran was slowly added. After 15 min, the reaction was stopped by the addition of a saturated aqueous solution of NaHCO 3 and diethyl ether. The aqueous phase was extracted with diethyl ether (2 × 30 mL). The combined organic phases were dried (Na 2 SO 4 ), filtered, then concentrated in vacuo. The crude product thus obtained was purified by silica gel column chromatography using EtOAc:Hex as eluent. N-(2,4-dimethoxybenzyl)-3-(p-tolyl)propanamide ( 42 ), colorless amorphous solid. 1 H NMR (400 MHz, CDCl 3 ) δ 7.10 (d, J = 8.0 Hz, 1H), 7.05 (s, 4H), 6.46–6.37 (m, 2H), 5.76 (br s, 1H), 4.32 (d, J = 5.7 Hz, 2H), 3.80 (s, 3H), 3.77 (s, 3H), 2.94–2.85 (m, 2H), 2.47–2.38 (m, 2H), 2.30 (s, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 171.69, 160.48, 158.52, 137.86, 135.54, 130.53, 129.12, 128.21, 118.88, 103.88, 98.55, 55.41, 55.28, 38.86, 38.72, 31.29, 21.00. HRMS (ESI+) m/z calc. for C 19 H 23 NO 3 313.1678, found [M + H] + 314.1745. Rf = 0.51 (EtOAc/n-Hex; 1:1, v / v ). tert -butyl (2-isobutyl-4-oxoazetidin-3-yl)carbamate ( 44 ), yield: 89%, transparent oil. The reaction was carried out according to General Procedure using Na dispersion in mineral oil (25 wt%, TCI, 131 mg, 13.5 mmol, 6 EQ),15-crown-5 were (0.283 mL, 13.5 mmol, 6 EQ), tert -butyl (1-benzyl-2-isobutyl-4-oxoazetidin-3-yl) carbamate ( 43 ) (75 mg, 2.25 mmol, 1 EQ) and isopropanol (0,109 mL, 13.5 mmol, 6 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:1 ( v / v ) as eluent.. 1 H NMR (400 MHz, CDCl 3 ) δ 7.64 (br s, 1H), 6.69 (ddd, J = 14.3, 10.5, 1.2 Hz, 1H), 5.19 (dd, J = 14.3, 7.0 Hz, 1H), 5.15–5.07 (m, 1H), 3.82 (d, J = 5.9 Hz, 2H), 2.33 (dqd, J = 13.5, 6.8, 1.3 Hz, 1H), 1.47 (s, 9H), 1.01 (s, 3H), 1.00 (s, 3H). HRMS (ESI+) m/z calc. for C 12 H 22 N 2 O 3 242.1630, found [M + H] + 243.1685. Rf = 0.36 (EtOAc/n-Hex; 1:1, v / v ). 4.10. General Procedure for the Synthesis of N1-Dimethoxybenzyl Deprotected ß Lactam ( 46 – 50 ) Cerium Ammonium Nitrate ß-Lactam (1 EQ) was dissolved in acetonitrile (25 mL/mmol) and distilled water (20 mL/mmol) and placed under argon. The solution was cooled to −10 °C with a sodium chloride ice bath. Cerium ammonium nitrate (3 EQ) was dissolved in distilled water and added dropwise to the vigorously stirring reaction mixture. The reaction was stirred at −10 °C for 1–2 h and then transferred to a separation funnel containing diethyl ether and saturated aqueous NaHCO 3 . The aqueous phase was washed with diethyl ether. The combined organic phases were dried over Na 2 SO 4, and the solvents were evaporated. The solid was purified by silica gel column chromatography, using the gradient EtOAc: Hex as eluent. tert -butyl (1-(2,4-dimethoxybenzoyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-yl)carbamate ( 44 ), light orange amorphous solid. 1 H NMR (400 MHz, CDCl 3 ) δ 8.25 (d, J = 8.7 Hz, 2H), 7.55–7.51 (m, 1H), 7.52 (d, J = 8.4, 2H), 6.61–6.51 (m, 2H), 5.59 (d, J = 6.4 Hz, 1H), 5.36–5.27 (m, 1H), 4.70 (d, J = 8.4 Hz, 1H), 3.91 (s, 3H), 3.88 (s, 3H), 1.28 (s, 9H). 13 C NMR (100 MHz, CDCl 3 ) δ 164.61, 163.87, 160.10, 154.18, 147.80, 141.44, 132.04, 127.90, 123.80, 115.51, 105.18, 98.65, 81.20, 60.79, 59.56, 55.93, 55.58, 27.95. HRMS (ESI+) m/z calc. for C 23 H 25 N 3 O 8 471.1642, found [M + Na] + 494.1532. Rf = 0.43 (EtOAc/n-Hex; 2:1, v / v ). 2-(2-Oxo-4-(4-(trifluoromethyl)phenyl)azetidin-3-yl)isoindoline-1,3-dione ( 46 ), yield: 65%, colorless amorphous solid. The reaction was carried out according to General Procedure using 4-(1-(2,4-dimethoxybenzyl)-3-(1,3-dioxoisoindolin-2-yl)-4-oxoazetidin-2-yl)benzonitrile ( 21 ) (156 mg, 0.31 mmol, 1 EQ) and cerium ammonium nitrate (502 mg, 0.92 mmol, 3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 7.72–7.60 (m, 4H), 7.54–7.41 (m, 4H), 7.28 (br s, 1H), 5.71 (dd, J = 5.4, 1.8 Hz, 1H), 5.26 (d, J = 5.4 Hz, 1H). 13 C NMR (100 MHz, CDCl 3 ) δ = 166.66, 164.59, 138.82, 134.48, 131.00, 130.45, 130.13, 127.11, 125.43, 123.62, 60.37, 57.23. HRMS (ESI+) m/z calc. for C 18 H 11 F 3 N 2 O 3 360.0722, found [M + H] + 361.0791. Rf = 0.43 (EtOAc/n-Hex; 1:1, v / v ). 2-(2-(3-Bromo-4-fluorophenyl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 47 ), yield: 40%, light yellow amorphous solid. The reaction was carried out according to General Procedure using 2-(2-(4-bromo-3-fluorophenyl)-1-(2,4-dimethoxybenzyl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 22 ) (151 g, 0,29 mmol, 1 EQ) and cerium ammonium nitrate (473 mg, 0.86 mmol, 3EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 2:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 7.76–7.65 (m, 4H), 7.53 (dd, J = 6.4, 2.0 Hz, 1H), 7.29–7.23 (m, 1H), 7.03 (br s, 1H), 6.99 (t, J = 8.4 Hz, 1H), 5.63 (dd, J = 5.3, 1.9 Hz, 1H), 5.14 (d, J = 5.3 Hz, 1H). 13 C NMR (100 MHz, CDCl 3 ) δ 166.65, 164.37, 159.93, 157.46, 134.52, 132.15, 132.12, 131.98, 131.04, 127.46, 127.38, 123.69, 116.66, 116.44, 109.29, 109.07, 60.45, 56.58. HRMS (ESI+) m/z calc. for C 17 H 10 BrFN 2 O 3 387.9859, found [M + H] + 388.9936. Rf = 0.43 (EtOAc/n-Hex; 2:1, v / v ). tert -butyl (2-(4-nitrophenyl)-4-oxoazetidin-3-yl)carbamate ( 48 ), yield: 51%, light red amorphous solid. The reaction was carried out according to General Procedure using tert -butyl (1-(2,4-dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-yl)carbamate ( 52 ) (2.5 g, 5.4 mmol, 1 EQ) and cerium ammonium nitrate (9 g, 16.4 mmol, 3 EQ). Product was purified by silica gel column chromatography using the gradient EtOAc: Hex = 1:1 to 4:1 as eluent. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.71 (br s, 1H), 8.22 (d, J = 8.7 Hz, 2H), 7.46 (d, J = 8.6 Hz, 2H), 7.42 (d, J = 8.4 Hz, 1H), 5. 41–5.36 (m, 1H), 5.05–4.93 (m, 1H), 1.18 (s, J = 12.9 Hz, 9H). 13 C NMR (100 MHz, DMSO- d 6 ) δ 167.10, 155.14, 147.19, 145.99, 128.87, 123.40, 78.84, 63.68, 56.77, 28.27. HRMS (ESI+) m/z calc. for C 14 H 17 N 3 O 5 307.1168, found [M + H] + 308.1240. Rf = 0.23 (EtOAc/n-Hex; 2:1, v / v ). 2-(2-(4-(Methylsulfonyl)phenyl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 49 ), yield: 43%, yellow amorphous solid. The reaction was carried out according to General Procedure using 2-(1-(2,4-Dimethoxybenzyl)-2-(4-(methylsulfonyl)phenyl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 25 ) (72 mg, 0.14 mmol, 1 EQ) and cerium ammonium nitrate (228 mg, 0.42 mmol, 3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 2:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 7.81 (d, J = 8.4 Hz, 2H), 7.74–7.64 (m, 4H), 7.54 (d, J = 8.2 Hz, 2H), 7.14 (br s, 1H), 5.72 (dd, J = 5.5, 1.9 Hz, 1H), 5.27 (d, J = 5.5 Hz, 1H), 2.90 (s, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 166.58, 164.18, 141.21, 140.20, 134.62, 130.95, 127.79, 127.56, 123.74, 60.48, 57.18, 44.33. HRMS (ESI+) m/z calc. for C 18 H 14 N 2 O 5 S 370.0623, found [M + Na] + 393.0530. Rf = 0.33 (EtOA). 2-(2-(Furan-2-yl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 50 ), yield: 15%, light brown oil. The reaction was carried out according to General Procedure 2-(1-(2,4-dimethoxybenzyl)-2-(furan-2-yl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 26 ) (206 mg, 0.48 mmol, 1 EQ) and cerium ammonium nitrate (0.84 g, 1.43 mmol, 3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 2:1 ( v / v ) as eluent. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.04–7.80 (m, 4H), 7.02 (dd, J = 10.6, 3.7 Hz, 1H), 6.91 (d, J = 7.3 Hz, 1H), 6.08 (dd, J = 10.6, 1.4 Hz, 1H), 5.87 (d, J = 6.0 Hz, 1H), 5.64 (dd, J = 6.0, 3.7 Hz, 1H), 4.55 (d, J = 6.0 Hz, 1H). 13 C NMR (100 MHz, DMSO- d 6 ) δ 191.40, 167.88, 165.11, 146.54, 135.84, 131.53, 128.34, 124.56, 67.63, 56.54, 53.78. HRMS (ESI+) m/z calc. for C 15 H 10 N 2 O 4 282.0641, found [M + H] + 283.0721. Rf = 0.33 (EtOAc/n-Hex; 2:1, v / v ). 4.11. General Procedure for the Synthesis of tert-Butyloxycarbonyl Deprotected ß Lactam ( 54 ) With Use of Trifluoroacetic Acid tert -butyl (2-(4-nitrophenyl)-4-oxoazetidin-3-yl)carbamate ( 48 ) (150 mg, 0.5 mmol, 1 EQ) was dissolved in dry dichloromethane (2 mL), anisole (0.49 mL, 4.5 mmol, 9 EQ) was added, and the solution was cooled to −5 °C on a sodium chloride ice bath. Trifluoroacetic acid (1.53 mL, 20 mmol, 40 EQ) was added, and the solution was slowly warmed to room temperature. After stirring for 1.5 h, the solvent and the excess trifluoroacetic acid were evaporated. The residue was precipitated from methyl tert -butyl ether. The solid was used in the next step without further purification. 2-(4-Nitrophenyl)-4-oxoazetidin-3-aminium trifluoroacetate ( 54 ), yield: 83%, brown oil. 1 H NMR (400 MHz, DMSO- d 6 ) δ 9.24 (s, 1H), 8.46 (br s, 3H), 8.30 (d, J = 8.8 Hz, 2H), 7.67 (d, J = 8.6 Hz, 2H), 5.19 (d, J = 5.4 Hz, 1H), 4.86 (dd, J = 5.4, 1.8 Hz, 1H). 13 C NMR (100 MHz, DMSO- d 6 ) δ 163.43, 148.06, 142.37, 129.63, 123.92, 59.90, 49.18. HRMS (ESI+) m/z calc. for C 9 H 9 N 3 O 3 207.0644, found [M + H] + 208.0716. Rf = 0.08 (EtOAc/n-Hex; 9:1, v / v ).
Unless otherwise stated, all reactions were carried out under argon atmosphere in flame-dried glassware. Chemicals and solvents were obtained from commercial sources (Sigma-Aldrich, Acros Organics, TCI Europe, fluorochem, and Apollo Sci) and were used as supplied. Dry solvents were prepared by distillation from CaH 2 (CH 2 Cl 2 ) or from a mixture of sodium and benzophenone (tetrahydrofuran). Other solvents (dimethylformamide, toluene, methanol, and CH 3 CN) were used directly from anhydrous Aldrich Sure/Seal bottles. Evaporation of the solvent was carried out under reduced pressure. Reactions were monitored by thin-layer chromatography (TLC) on silica gel aluminum plates (Merck DC Fertigplatten Kieselgel 60 GF254), visualized under UV light (254 nm), and stained with appropriate TLC stains for detection (ninhydrin, dinitrophenylhydrazine, and phospho-molybdic acid). The products were purified by flash column chromatography performed on Merck silica gel 60 (mesh size, 70–230) using the indicated solvents. Yields are reported for the purified products. 1 H NMR and 13 C NMR spectra were recorded at 295 K using a Bruker Avance III NMR spectrometer equipped with a Broadband decoupling inverse 1 H probe, at 400 MHz and 100 MHz, respectively. Chemical shifts (δ) are given in parts per million (ppm) and refer to tetramethylsilane (TMS) as an internal standard. The coupling constants ( J ) are given in Hertz (Hz), and the splitting patterns are reported as: s, singlet; br s, broad singlet; d, doublet; dd, double doublet; t, triplet, and m, multiplet. Mass spectra were recorded using an ADVION Expres-sion CMSL mass spectrometer (Advion Inc., Ithaca, NY, USA). High-resolution, accurate mass measurements were performed using the ExactiveTM Plus Orbitrap mass spectrometer (Thermo Fisher Scientific Inc., Waltham, MA, USA).
1 – 17 ) To a solution of an appropriate aldehyde (1 EQ) in dry dichloromethane or dry methanol was added an amine (1 EQ). The resultant solution was stirred for 15 min before Na 2 SO 4 (4 EQ) was added. The reaction mixture was then stirred at room temperature until TLC showed complete consumption of the starting material (30 min to 16 h). Next, the drying agent was removed by filtration, and the volatiles were removed under reduced pressure to afford the desired products, which were used in the next step without further purification. N -Benzyl-1-(4-(trifluoromethyl)phenyl)methanimine ( 1 ), quantitative yield, brown oil. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.62 (s, 1H), 8.01 (d, J = 8.7 Hz, 2H), 7.83 (d, J = 8.7 Hz, 2H), 7.33 (m, 5H), 4.84 (s, 2H); Rf = 0.66 (EtOAc/Hexane = 1:1, v / v ) as reported . N -(2,4-Dimethoxybenzyl)-1-(4-(trifluoromethyl)phenyl)methanimine ( 2 ), quantitative yield, brown oil. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.50 (s, 1H), 7.97 (d, J = 8.0 Hz, 2H), 7.81 (d, J = 8.2 Hz, 2H), 7.16 (d, J = 8.3 Hz, 1H), 6.58 (d, J = 2.4 Hz, 1H), 6.51 (dd, J = 8.3, 2.4 Hz, 1H), 4.71 (s, 2H), 3.79 (s, J = 4.7 Hz, 3H), 3.76 (s, J = 3.6 Hz, 3H); Rf = 0.60 (EtOAc/Hexane = 1:1, v / v ) as reported . 4-(((2,4-Dimethoxybenzyl)imino)methyl)benzonitrile ( 3 ), quantitative yield, colorless amorphous solid. 1 H NMR (400 MHz, CDCl 3 ) δ 8.33 (s, 1H), 7.89–7.79 (m, 2H), 7.74–7.61 (m, 2H), 7.21–7.13 (m, 1H), 6.53–6.42 (m, 2H), 4.80 (s, 2H), 3.81 (app s, 3H); Rf = 0.54 (EtOAc/Hexane = 1:1, v / v ) as reported . 1-(3-Bromo-4-fluorophenyl)- N -(2,4-dimethoxybenzyl)methanimine ( 4 ), quantitative yield, yellow oil. 1 H NMR (400 MHz, CDCl 3 ) δ 8.21 (s, 1H), 8.00 (dd, J = 6.8, 2.1 Hz, 1H), 7.68–7.58 (m, 1H), 7.21–7.14 (m, 1H), 7.15–7.05 (m, 1H), 6.50–6.46 (m, 2H), 4.74 (s, 2H), 3.80 (app s, 6H). 13 C NMR (100 MHz, CDCl 3 ) δ 161.61, 160.32, 158.86, 158.37, 133.10, 130.33, 128.99, 128.91, 119.29, 116.64, 116.41, 104.16, 98.56, 58.78, 55.40, 55.40. HRMS (ESI+) m/z calc. for C 16 H 15 BrFNO 2 351.0270, found [M + H] + 352.0338. Rf = 0.86 (EtOAc/Hexane = 1:1 v / v ). N -Benzyl-1-(4-nitrophenyl)methanimine ( 5 ), quantitative yield, yellow amorphous solid. 1 H NMR (400 MHz, CDCl 3 ) δ 8.46 (s, 1H), 8.26 (d, J = 8.7 Hz, 2H), 7.94 (d, J = 8.7 Hz, 2H), 7.43–7.15 (m, 5H), 4.88 (s, 2H); Rf = 0.66 (EtOAc/Hexane = 1:1, v / v ) as reported . N -(2,4-Dimethoxybenzyl)-1-(4-nitrophenyl)methanimine ( 6 ), quantitative yield, yellow amorphous solid. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.54 (s, 1H), 8.31–8.26 (m, 2H), 8.04–7.98 (m, 2H), 7.16 (d, J = 8.3 Hz, 1H), 6.58 (d, J = 2.4 Hz, 1H), 6.51 (dd, J = 8.3, 2.4 Hz, 1H), 4.73 (s, 2H), 3.79 (s, 3H), 3.76 (s, 3H); Rf = 0.46 (EtOAc/Hexane = 1:1, v / v ) as reported . N -(2,4-Dimethoxybenzyl)-1-(4-(methylsulfonyl)phenyl)methanimine ( 7 ), quantitative yield, pale yellow amorphous solid. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.51 (s, 1H), 8.00 (m, 4H), 7.16 (d, J = 8.3 Hz, 1H), 6.58 (d, J = 2.4 Hz, 1H), 6.51 (dd, J = 8.3, 2.4 Hz, 1H), 4.72 (s, 2H), 3.79 (s, 3H), 3.76 (s, 3H), 3.25 (s, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 160.40, 159.59, 158.40, 141.76, 141.24, 130.37, 128.93, 127.63, 119.01, 104.21, 98.57, 59.13, 55.41, 44.46. HRMS (ESI+) m/z calc. for C 17 H 19 NO 4 S 333.1035, found [M + H] + 334.1104. Rf = 0.25 (EtOAc:Hex = 1:1, v / v ). 4-(((2,4-Dimethoxybenzyl)imino)methyl)- N,N -dimethylaniline ( 8 ), quantitative yield, colorless amorphous solid. 1 H NMR (400 MHz, CDCl 3 ) δ 8.19 (d, J = 1.4 Hz, 1H), 7.63 (d, J = 8.9 Hz, 2H), 7.18 (d, J = 8.9 Hz, 2H), 6.68 (d, J = 8.9 Hz, 1H), 6.48–6.42 (m, 2H), 4.68 (s, 2H), 3.79 (s, 3H), 3.78 (s, 3H), 2.98 (s, 6H). 13 C NMR (100 MHz, CDCl 3 ) δ 162.19, 159.96, 158.22, 152.06, 129.96, 129.70, 124.44, 120.58, 111.61, 111.01, 104.01, 98.42, 58.71, 55.37, 50.39, 40.21, 40.06. HRMS (ESI+) m/z calc. for C 18 H 22 N 2 O 2 298.1681, found [M + H] + 299.1751; Rf = 0.63 (EtOAc/Hexane = 1:1, v / v ). N -Benzyl-1-phenylmethanimine ( 9 ), quantitative yield, brown oil. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.49 (s, 1H), 7.84–7.75 (m, 2H), 7.50–7.22 (m, 8H), 4.77 (s, J = 1.2 Hz, 2H); Rf = 0.65 (EtOAc/Hex = 1:1 v / v ) as reported . N -(2,4-Dimethoxybenzyl)-1-(furan-2-yl)methanimine ( 10 ), quantitative yield, dark brown oil. 1 H NMR (400 MHz, CDCl 3 ) δ 8.08 (s, 1H), 7.49 (s, 1H), 7.21–7.16 (m, 1H), 6.73 (d, J = 3.4 Hz, 1H), 6.49–6.44 (m, 3H), 4.74 (s, 2H), 3.80 (s, 3H), 3.79 (s, 3H); Rf = 0.36 (EtOAc/Hexane = 1:1, v / v ) as reported . N -(2,4-Dimethoxybenzyl)-1-(1 H -imidazol-5-yl)methanimine ( 11 ), quantitative yield, colorless amorphous solid. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.22 (s, 1H), 7.72–7.69 (m, 1H), 7.41 (s, 1H), 7.13 (d, J = 8.3 Hz, 1H), 6.56 (d, J = 2.4 Hz, 1H), 6.49 (dd, J = 8.3, 2.4 Hz, 1H), 4.57 (s, 2H), 3.77 (s, 3H), 3.75 (s, 3H). 13 C NMR (100 MHz, DMSO- d 6 ) δ 160.21, 158.35, 130.56, 120.00, 104.92, 98.69, 58.39, 55.82, 55.63. HRMS (ESI+) m/z calc. for C 13 H 15 N 3 O 2 245.1164, found [M + H] + 246.1234. Rf = 0.1 (EtOAc). 1-(Benzo[ b ]thiophen-2-yl)- N -benzylmethanimine ( 12 ), quantitative yield, yellow amorphous solid. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.76 (s, 1H), 7.98–7.8 (m 2H), 7.87 (s, 1H), 7.44–7.40 (m, 2H), 7.38–7.21 (m, 5H), 4.80 (s, 2H); Rf = 0.67 (EtOAc/Hexane = 1:1, v / v ) as reported . 1-(Benzo[ d ][1,3]dioxol-5-yl)- N -benzylmethanimine ( 13 ), quantitative yield, colorless amorphous solid. 1 H NMR (400 MHz, CDCl 3 ) δ 8.26 (t, J = 1.4 Hz, 1H), 7.41 (d, J = 1.4 Hz, 1H), 7.35–7.22 (m, 5H), 7.14 (dd, J = 8.0, 1.6 Hz, 1H), 6.82 (d, J = 7.9 Hz, 1H), 5.98 (s, 2H), 4.77 (s, 2H); Rf = 0.60 (EtOAc/Hexane = 1:1, v / v ) as reported . 1-(Benzo[d][1,3]dioxol-5-yl)- N -(2,4-dimethoxybenzyl)methanimine ( 14 ), quantitative yield, colorless amorphous solid. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.27 (s, 1H), 7.30 (d, J = 1.5 Hz, 1H), 7.20 (dd, J = 7.9, 1.5 Hz, 1H), 7.12 (d, J = 8.3 Hz, 1H), 6.97 (d, J = 8.0 Hz, 1H), 6.56 (d, J = 2.4 Hz, 1H), 6.49 (dd, J = 8.3, 2.4 Hz, 1H), 6.07 (s, 2H), 4.60 (s, 2H), 3.78 (s, 3H), 3.75 (s, 3H); Rf = 0.65 (EtOAc/Hexane = 1:1, v / v ) as reported . N -Benzyl-3-methylbutan-1-imine ( 15 ), quantitative yield, light orange oil. 1 H NMR (400 MHz, CDCl 3 ) δ 7.38–7.25 (m, 5H), 6.32 (s, 1H), 3.90 (s, 2H), 2.03–1.81 (m, 3H), 0.99–0.76 (m, 6H). 13 C NMR (100 MHz, CDCl 3 ) δ 179.39, 138.07, 128.76, 127.95, 127.81, 45.70, 44.53, 25.90, 22.58. HRMS (ESI+) m/z calc. for C 12 H 17 N 175.1361, found [M + H] + 176.1435. Rf = 0.65 (EtOAc/Hex = 1:1, v / v ). N -(2,4-Dimethoxybenzyl)-3-methylbutan-1-imine ( 16 ), quantitative yield, yellow oil. 1 H NMR (400 MHz, CDCl 3 ) δ 7.15 (d, J = 8.1 Hz, 1H), 6.73 (s, 1H), 6.46–6.38 (m, 2H), 3.86 (s, 2H), 3.81 (s, 3H), 3.79 (s, 3H), 2.00–1.88 (m, 3H), 0.88 (s, 3H), 0.87 (s, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 179.20, 160.92, 158.55, 130.46, 118.49, 104.03, 98.55, 55.38, 55.29, 45.95, 40.20, 25.93, 22.64. Rf = 0.63 (EtOAc/Hex = 1:1, v / v ). N -(2,4-Dimethoxybenzyl)heptan-1-imine ( 17 ), quantitative yield, orange oil. 1 H NMR (400 MHz, CDCl 3 ) δ 9.48 (s, 1H), 7.18–7.15 (m, 1H), 6.46–6.39 (m, 2H), 3.92 (s, 2H), 3.84–3.73 (m, 6H), 1.93–1.79 (m, 4H), 1.51–1.20 (m, 6H), 0.95–0.82 (m, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 176.84, 161.52, 158.65, 131.30, 114.65, 104.24, 98.38, 55.32, 55.27, 39.00, 22.73, 14.03. HRMS (ESI+) m/z calc. for C 16 H 25 NO 2 263.1885, found [M + H] + 264.1954. Rf = 0.85 (EtOAc/Hex = 1:1, v / v ).
4.3.1. General Procedure for the Synthesis of Acid Chloride ( 18 ) N -phthaloylglycine (2.00 g, 9.75 mmol, 1 EQ) was dissolved in dry dichloromethane (10 mL), and the solution was cooled to 0 °C using an ice bath before oxalyl chloride (0.95 mL, 10.73 mmol, 1.1 EQ) was added dropwise over 30 min. Upon complete addition, the reaction mixture was stirred at 0 °C for an additional 2 h, and the solvent was removed under reduced pressure without heating. The acyl chlorides thus obtained were used in the subsequent step without further purification. 2-(1,3-Dioxoisoindolin-2-yl)acetyl chloride ( 18 ), quantitative yield, yellow amorphous solid. 1 H NMR (400 MHz, CDCl 3 ) 7.95–7.89 (m, 1H), 7.82–7.76 (m, 1H), 4.83 (s, 1H); as reported . 4.3.2. General Procedure for the Synthesis of Diazoketone ( 31 ) N -Benzyloxycarbonylglycine (2.09 g, 10.0 mmol, 1 EQ) was dissolved in dry tetrahydrofuran (20 mL), and the resultant solution was cooled to −20 °C using a sodium chloride ice bath before triethylamine (1.39 mL, 10.0 mmol, 1 EQ) was added in one portion. Ethyl chloroformate (1.91 mL, 10.0 mmol, 1 EQ) was then added dropwise, and the reaction mixture was stirred for another 1 h. The white precipitate formed was removed by filtration. To the filtrate were slowly added dry acetonitrile (80 mL) (4:1 solution in THF) and (trimethylsilyl)diazomethane (2.0 M solution in hexane, 10 mL, 20.0 mmol, 2 EQ). The resultant reaction mixture was then stirred at 4 °C for 24–48 h. The reaction was quenched by the addition of diethyl ether and 10% (m/m) aqueous citric acid. The organic phase was then washed with saturated aqueous NaHCO 3 and brine. The organic layer was dried over Na 2 SO 4, and the solvents were evaporated. The diazoketone was purified by silica gel column chromatography using EtOAc:Hex = 1:1, v / v as eluent. Benzyl (3-diazo-2-oxopropyl)carbamate ( 31 ), quantitative yield, transparent amorphous solid. 1 H NMR (400 MHz, CDCl 3 ) δ (ppm) = 7.42–7.31 (m, 5H), 5.46 (s, 1H), 5.39 (s, 1H), 5.13 (s, 2H), 3.99 (s, 1H); as reported . 4.3.3. General Procedure for the Synthesis of Mixed Anhydride In a flame-dried flask, N -( tert -butoxycarbonyl)glycine (3.00 g, 17.13 mmol, 1 EQ) was dissolved in dry tetrahydrofuran (20 mL) and placed under an argon atmosphere. The solution was cooled to −60 °C, and triethylamine (2.62 mL, 18.84 mmol, 1.1 EQ) was added in one portion. Then ethyl chloroformate (2.13 mL, 22.27 mmol, 1.3EQ) was added dropwise over a period of 30 min. After the complete addition of the reagent, the reaction mixture was stirred at −40 °C for another 2 h. The resultant reaction mixture was then directly used in the next step without any further purification. The same reaction conditions were used for the synthesis of 2-(((benzyloxy)carbonyl)amino)acetic anhydride from ((benzyloxy)carbonyl)glycine.
18 ) N -phthaloylglycine (2.00 g, 9.75 mmol, 1 EQ) was dissolved in dry dichloromethane (10 mL), and the solution was cooled to 0 °C using an ice bath before oxalyl chloride (0.95 mL, 10.73 mmol, 1.1 EQ) was added dropwise over 30 min. Upon complete addition, the reaction mixture was stirred at 0 °C for an additional 2 h, and the solvent was removed under reduced pressure without heating. The acyl chlorides thus obtained were used in the subsequent step without further purification. 2-(1,3-Dioxoisoindolin-2-yl)acetyl chloride ( 18 ), quantitative yield, yellow amorphous solid. 1 H NMR (400 MHz, CDCl 3 ) 7.95–7.89 (m, 1H), 7.82–7.76 (m, 1H), 4.83 (s, 1H); as reported .
31 ) N -Benzyloxycarbonylglycine (2.09 g, 10.0 mmol, 1 EQ) was dissolved in dry tetrahydrofuran (20 mL), and the resultant solution was cooled to −20 °C using a sodium chloride ice bath before triethylamine (1.39 mL, 10.0 mmol, 1 EQ) was added in one portion. Ethyl chloroformate (1.91 mL, 10.0 mmol, 1 EQ) was then added dropwise, and the reaction mixture was stirred for another 1 h. The white precipitate formed was removed by filtration. To the filtrate were slowly added dry acetonitrile (80 mL) (4:1 solution in THF) and (trimethylsilyl)diazomethane (2.0 M solution in hexane, 10 mL, 20.0 mmol, 2 EQ). The resultant reaction mixture was then stirred at 4 °C for 24–48 h. The reaction was quenched by the addition of diethyl ether and 10% (m/m) aqueous citric acid. The organic phase was then washed with saturated aqueous NaHCO 3 and brine. The organic layer was dried over Na 2 SO 4, and the solvents were evaporated. The diazoketone was purified by silica gel column chromatography using EtOAc:Hex = 1:1, v / v as eluent. Benzyl (3-diazo-2-oxopropyl)carbamate ( 31 ), quantitative yield, transparent amorphous solid. 1 H NMR (400 MHz, CDCl 3 ) δ (ppm) = 7.42–7.31 (m, 5H), 5.46 (s, 1H), 5.39 (s, 1H), 5.13 (s, 2H), 3.99 (s, 1H); as reported .
In a flame-dried flask, N -( tert -butoxycarbonyl)glycine (3.00 g, 17.13 mmol, 1 EQ) was dissolved in dry tetrahydrofuran (20 mL) and placed under an argon atmosphere. The solution was cooled to −60 °C, and triethylamine (2.62 mL, 18.84 mmol, 1.1 EQ) was added in one portion. Then ethyl chloroformate (2.13 mL, 22.27 mmol, 1.3EQ) was added dropwise over a period of 30 min. After the complete addition of the reagent, the reaction mixture was stirred at −40 °C for another 2 h. The resultant reaction mixture was then directly used in the next step without any further purification. The same reaction conditions were used for the synthesis of 2-(((benzyloxy)carbonyl)amino)acetic anhydride from ((benzyloxy)carbonyl)glycine.
19 – 30 ) Schiff base (1 EQ) was dissolved in dry toluene (0.1–0.2 mmol/mL) in a flame-dried flask and placed under an argon atmosphere. Triethylamine (2.5 EQ) was then added in one portion, and the resultant solution was heated to 80 °C, before 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride (1.3 EQ), dissolved in in dry toluene, was added dropwise over a period of 30 min. Upon complete addition, the reaction was stirred at 80 °C for a further 1.5–3.5 h. The reaction mixture was then cooled to room temperature, and the volatiles were removed in vacuo. The solid residue thus obtained was redissolved in ethyl acetate. The organic phase was washed with 10% aq. citric acid solution, saturated NaHCO 3, and brine. The organic phase was dried (Na 2 SO 4 ), filtered, then concentrated in vacuo. Some cyclized ß-lactams were purified by silica gel column chromatography using EtOAc: Hex as eluent. 2-(1-Benzyl-2-oxo-4-(4-(trifluoromethyl)phenyl)azetidin-3-yl)isoindoline-1,3-dione ( 19 ), yield: 51%, colorless amorphous solid. The reaction was carried out according to General Procedure I using N -benzyl-1-(4-(trifluoromethyl)phenyl)methanimine ( 1 ), (1.32 g, 5 mmol, 1.0 EQ), triethylamine (1.74 mL, 12.5 mmol, 2.5 EQ) and 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride (1.45 g, 6.5 mmol, 1.3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 2:1 ( v / v ) as eluent. 1 H NMR (400 MHz, DMSO- d 6 ) δ 7.80–7.70 (m, 4H), 7.51 (d, J = 8.2 Hz, 2H), 7.39–7.28 (m, 7H), 5.77 (d, J = 5.4 Hz, 1H), 5.18 (d, J = 5.4 Hz, 1H), 4.82 (d, J = 15.4 Hz, 1H), 4.47 (d, J = 15.4 Hz, 1H). 13 C NMR (100 MHz, CDCl 3 ) δ 166.73, 163.56, 137.29, 134.53, 134.42, 131.04, 129.09, 128.64, 128.26, 127.79, 125.50, 125.47, 123.59, 60.16, 59.84, 45.83. HRMS (ESI+) m/z calc. for C 25 H 17 F 3 N 2 O 3 450.1191, found [M + H] + 451.1260. Rf = 0.42 (EtOAc/n-Hex; 2:1, v / v ). 2-(1-(2,4-Dimethoxybenzyl)-2-oxo-4-(4-(trifluoromethyl)phenyl)azetidin-3-yl)isoindoline-1,3-dione ( 20 ), yield: 55%, light brown amorphous solid. The reaction was carried out according to General Procedure I using N -(2,4-dimethoxybenzyl)-1-(4-(trifluoromethyl)phenyl)methanimine ( 2 ), (3.23 g, 10 mmol, 1.0 EQ), triethylamine (3.48 mL, 25 mmol, 2.5 EQ) and 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride ( 18 ) (2.91 g, 13 mmol, 1.3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 7.71–7.60 (m, 4H), 7.44 (d, J = 8.2 Hz, 2H), 7.33 (d, J = 8.1 Hz, 2H), 7.15 (d, J = 8.3 Hz, 1H), 6.43 (dd, J = 8.3, 2.4 Hz, 1H), 6.37 (d, J = 2.3 Hz, 1H), 5.46 (d, J = 5.4 Hz, 1H), 4.90 (d, J = 14.3 Hz, 1H), 4.84 (d, J = 5.4 Hz, 1H), 4.30 (d, J = 14.3 Hz, 1H), 3.79 (s, 3H), 3.56 (s, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 166.77, 163.44, 161.18, 158.59, 138.33, 134.33, 131.59, 131.07, 127.60, 125.17, 125.13, 123.49, 115.03, 104.34, 98.33, 60.65, 59.54, 55.41, 55.00, 40.82. HRMS (ESI+) m/z calc. for C 27 H 21 F 3 N 2 O 5 510.1403, found [M + H] + 511.1470. Rf = 0.31 (EtOAc/n-Hex; 1:1, v / v ). 4-(1-(2,4-Dimethoxybenzyl)-3-(1,3-dioxoisoindolin-2-yl)-4-oxoazetidin-2-yl)benzonitrile ( 21 ), yield: 48%, colorless amorphous solid. The reaction was carried out according to General Procedure I using 4-(((2,4-dimethoxybenzyl)imino)methyl)benzonitrile ( 3 ) (0.32 g, 1.15 mmol, 1 EQ), triethylamine (0.40 mL, 2.87 mmol, 2.5 EQ) and 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride ( 18 ) (0.33 g, 1.50 mmol, 1.3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 2:1 ( v / v ) as eluent. 1 H NMR (400 MHz, DMSO- d 6 ) δ 7.80–7.72 (m, 4H), 7.63 (d, J = 8.4 Hz, 2H), 7.28 (d, J = 8.2 Hz, 2H), 7.21–7.17 (m, 1H), 6.49–6.46 (m, 2H), 5.58 (d, J = 5.5 Hz, 1H), 5.01 (d, J = 5.4 Hz, 1H), 4.61 (d, J = 14.5 Hz, 1H), 4.34 (d, J = 14.6 Hz, 1H), 3.74 (s, 3H), 3.57 (s, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ = 166.69, 163.32, 161.26, 158.57, 139.89, 134.49, 131.99, 131.65, 130.97, 127.88, 123.58, 118.41, 114.87, 111.87, 104.39, 98.39, 60.72, 59.57, 55.44, 55.02, 40.97. MS (ESI+, m/z), 468.4 ([M + H] + ). Rf = 0.24 (EtOAc/n-Hex; 1:1, v / v ). 2-(2-(4-Bromo-3-fluorophenyl)-1-(2,4-dimethoxybenzyl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 22 ), yield: 51%, pale yellow amorphous solid. The reaction was carried out according to General Procedure I using 1-(3-bromo-4-fluorophenyl)- N -(2,4-dimethoxybenzyl)methanimine ( 4 ) (0.53 g, 1.5 mmol, 1 EQ), triethylamine (0.52 mL, 3.75 mmol, 2.5 EQ) and 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride ( 18 ) (0.44 g, 1.95 mmol, 1.3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 2:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 7.73–7.64 (m, 4H), 7.37 (dd, J = 6.5, 2.2 Hz, 1H), 7.17–7.10 (m, 2H), 6.91 (t, J = 8.4 Hz, 1H), 6.43 (dd, J = 8.3, 2.4 Hz, 1H), 6.37 (d, J = 2.3 Hz, 1H), 5.40 (d, J = 5.3 Hz, 1H), 4.82 (d, J = 14.3 Hz, 1H), 4.74 (d, J = 5.4 Hz, 1H), 4.30 (d, J = 14.3 Hz, 1H), 3.80 (s, 3H), 3.61 (s, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 166.77, 163.36, 161.19, 158.57, 134.36, 132.52, 131.60, 131.51, 131.47, 131.14, 127.91, 127.84, 123.55, 116.33, 116.10, 115.03, 104.37, 98.37, 60.22, 59.61, 55.43, 55.10, 40.71. HRMS (ESI+) m/z calc. for C 26 H 20 BrFN 2 O 5 538.0540, found [M + H] + 539.0606. Rf = 0.56 (EtOAc/n-Hex; 2:1, v / v ). 2-(1-Benzyl-2-(4-nitrophenyl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 23 ), yield: 47%, pale yellow amorphous solid. The reaction was carried out according to General Procedure I using N -benzyl-1-(4-nitrophenyl)methanimine ( 5 ) (2.10 g, 10 mmol, 1 EQ), triethylamine (3.48 mL, 25 mmol, 2.5 EQ) and 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride ( 18 ) (2.91 g, 13 mmol, 1.3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 2:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 8.05 (d, J = 8.8 Hz, 2H), 7.74–7.60 (m, 4H), 7.39 (d, J = 8.7 Hz, 2H), 7.36–7.22 (m, 5H), 5.57 (d, J = 5.5 Hz, 1H), 5.06 (d, J = 14.8 Hz, 1H), 4.93 (d, J = 5.4 Hz, 1H), 4.26 (d, J = 14.8 Hz, 1H). 13 C NMR (100 MHz, CDCl 3 ) δ 166.64, 163.40, 147.84, 140.76, 134.59, 134.35, 130.93, 129.16, 128.66, 128.40, 128.33, 123.72, 123.70, 60.16, 59.88, 46.08. HRMS (ESI+) m/z calc. for C 24 H 17 N 3 O 5 427.1168, found [M + H] + 433.1386. Rf = 0.25 (EtOAc/n-Hex; 1:1, v / v ). 2-(1-(2,4-Dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 24 ), yield: 51%, pale yellow amorphous solid. The reaction was carried out according to General Procedure I using N -(2,4-dimethoxybenzyl)-1-(4-nitrophenyl)methanimine ( 6 ) (3.00 g, 10 mmol, 1 EQ), triethylamine (3.48 mL, 25 mmol, 2.5 EQ) and 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride ( 18 ) (2.91 g, 13 mmol, 1.3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 2:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 8.04 (d, J = 8.8 Hz, 2H), 7.72–7.60 (m 4H), 7.39 (d, J = 8.6 Hz, 2H), 7.16 (d, J = 8.3 Hz, 1H), 6.43 (dd, J = 8.3, 2.3 Hz, 1H), 6.36 (d, J = 2.3 Hz, 1H), 5.49 (d, J = 5.5 Hz, 1H), 4.91 (s, J = 14.3 Hz, 1H), 4.88 (d, J = 5.3 Hz, 1H), 4.33 (d, J = 14.3 Hz, 1H), 3.79 (s, 3H), 3.56 (s, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 166.67, 163.28, 161.29, 158.57, 147.57, 141.91, 134.50, 131.66, 130.96, 128.10, 123.64, 123.42, 114.82, 104.43, 98.41, 60.58, 59.58, 55.44, 55.06, 41.02. HRMS (ESI+) m/z calc. for C 26 H 21 N 3 O 7 487.1380, found [M + H] + 488.1443. Rf = 0.42 (EtOAc/n-Hex; 2:1, v / v ). 2-(1-(2,4-Dimethoxybenzyl)-2-(4-(methylsulfonyl)phenyl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 25 ), yield: 45%, pale yellow amorphous solid. The reaction was carried out according to General Procedure I using N -(2,4-dimethoxybenzyl)-1-(4-(methylsulfonyl)phenyl)methanimine ( 7 ) (1.00 g, 3.00 mmol, 1 EQ), triethylamine (1.04 mL, 7.50 mmol, 2.5 EQ) and 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride ( 18 ) (0.87 g, 3.90 mmol, 1.3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 2:1 ( v / v ) as eluent. 1 H NMR (400 MHz, DMSO- d 6 ) δ 7.86–7.69 (m, 4H), 7.68 (d, J = 8.2 Hz, 2H), 7.35 (d, J = 8.3 Hz, 2H), 7.20 (d, J = 8.8 Hz, 1H), 6.51–6.45 (m, 2H), 5.59 (d, J = 5.5 Hz, 1H), 5.03 (d, J = 5.4 Hz, 1H), 4.64 (d, J = 14.5 Hz, 1H), 4.34 (d, J = 14.6 Hz, 1H), 3.74 (s, 3H), 3.58 (s, 3H), 2.98 (s, 3H). 13 C NMR (100 MHz, DMSO- d 6 ) δ 166.80, 163.35, 161.07, 158.68, 141.06, 140.45, 135.42, 131.54, 130.78, 128.35, 126.90, 123.79, 115.39, 105.13, 98.72, 60.68, 59.76, 55.69, 55.64, 43.71, 41.04. HRMS (ESI+) m/z calc. for C 27 H 24 N 2 O 7 S 520.1304, found [M + H] + 521.1378. Rf = 0.36 (EtOAc/n-Hex; 2:1, v / v ). 2-(1-(2,4-Dimethoxybenzyl)-2-(furan-2-yl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 26 ), yield: 51%, colorless amorphous solid. The reaction was carried out according to General Procedure I using N -(2,4-dimethoxybenzyl)-1-(furan-2-yl)methanimine ( 10 ) (1.23 g, 5 mmol, 1 EQ), triethylamine (1.74 mL, 12.5 mmol, 2.5 EQ) and 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride ( 18 ) (1.45 g, 6.5 mmol, 1.3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 2:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 7.81–7.63 (m, 4H), 7.21–7.18 (m, 1H), 7.15 (d, J = 8.1 Hz, 1H), 6.48–6.40 (m, 2H), 6.29 (d, J = 3.3 Hz, 1H), 6.18 (dd, J = 3.3, 1.8 Hz, 1H), 5.42 (d, J = 5.0 Hz, 1H), 4.84 (d, J = 5.0 Hz, 1H), 4.80 (d, J = 14.5 Hz, 1H), 4.22 (d, J = 14.5 Hz, 1H), 3.80 (s, 3H), 3.71 (s, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 166.73, 163.17, 161.01, 158.61, 148.22, 142.85, 134.20, 131.45, 131.38, 123.51, 115.39, 110.56, 109.87, 104.20, 98.41, 59.03, 55.41, 55.29, 55.12, 40.18. HRMS (ESI+) m/z calc. for C 24 H 20 N 2 O 6 432.1321, found [M + H] + 433.1386. Rf = 0.23 (EtOAc/n-Hex; 2:1, v / v ). 2-(1-(2,4-Dimethoxybenzyl)-2-(1 H -imidazol-5-yl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 27 ), yield: 33%, light brown amorphous solid. The reaction was carried out according to General Procedure I using N -(2,4-dimethoxybenzyl)-1-(1 H -imidazol-5-yl)methanimine ( 11 ) (1.19 g, 7.7 mmol, 1 EQ), triethylamine (1.35 mL, 15.4 mmol, 2 EQ) and 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride ( 18 ) (2.23 g, 10 mmol, 1.3EQ). Product was purified by silica gel column chromatography using DKM: MeOH = 15:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 9.69 (s, 1H), 7.90–7.79 (m, 4H), 7.57–7.47 (m, 1H), 6.85 (d, J = 8.2 Hz, 1H), 6.38–6.23 (m, 2H), 4.92 (d, J = 16.7 Hz, 1H), 4.78 (d , J = 7.5 Hz, 2H), 4.74 (d, J = 7.3 Hz, 1H), 4.62 (d, J = 16.6 Hz, 1H), 3.81 (s, 3H), 3.75 (s, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 185.51, 184.92, 167.93, 161.04, 158.25, 140.99, 139.31, 136.01, 134.21, 132.14, 123.59, 104.24, 98.57, 63.70, 55.41, 55.38, 55.21, 39.94. HRMS (ESI+) m/z calc. for C 23 H 20 N 4 O 5 432.1434, found [M + H] + 433.1500. Rf = 0.42 (DKM/MeOH; 9:1, v / v ). 2-(1-Benzyl-2-isobutyl-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 28 ), yield: 31%, colorless amorphous solid. The reaction was carried out according to General Procedure I using N -Benzyl-3-methylbutan-1-imine ( 15 ) (2 g, 11.6 mmol, 1 EQ), triethylamine (3.21 mL, 23 mmol, 2 EQ) and 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride ( 18 ) (3.1 g, 15 mmol, 1.3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 7.95–7.66 (m, 4H), 7.34–7.14 (m, 5H), 6.41 (d, J = 13.8 Hz, 1H), 5.26 (dd, J = 13.8, 7.3 Hz, 1H), 4.84–4.44 (m, 4H), 2.38–2.22 (m, 1H), 0.98 (d, J = 6.7 Hz, 3H), 0.93 (d, J = 6.7 Hz, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 167.98, 164.64, 136.60, 134.07, 132.27, 128.52, 127.37, 125.63, 123.56, 60.38, 48.50, 39.81, 29.57, 22.85, 22.68. HRMS (ESI+) m/z calc. for C 22 H 22 N 2 O 3 362.1630, found [M + H] + 363.1694. Rf = 0.54 (EtOAc/n-Hex; 1:1, v / v ). 2-(1-(2,4-Dimethoxybenzyl)-2-isobutyl-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 29 ), yield: 25%, colorless amorphous solid. The reaction was carried out according to General Procedure I using N -(2,4-Dimethoxybenzyl)-3-methylbutan-1-imine ( 16 ) (2.73 g, 11.6 mmol, 1 EQ), triethylamine (3.21 mL, 23 mmol, 2 EQ) and 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride ( 18 ) (3.36 g, 15 mmol, 1.3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:1 ( v / v ) as eluent. 1 H NMR (400 MHz, DMSO- d 6 ) δ 7.99–7.82 (m, 4H), 7.01–6.93 (m, 1H), 6.84–6.74 (m, 1H), 6.63–6.51 (m, 1H), 6.50–6.44 (m, 1H), 5.14–5.00 (m, 1H), 4.83–4.55 (m, 4H), 3.90–3.66 (m, 6H), 2.37–2.13 (m, 1H), 0.95 (d, J = 6.7 Hz, 3H), 0.88 (d, J = 6.7 Hz, 3H). 13 C NMR (100 MHz, DMSO- d 6 ) δ 168.00, 165.35, 160.05, 157.83, 135.21, 132.07, 127.88, 123.80, 120.61, 116.68, 105.05, 98.63, 65.35, 55.86, 55.62, 43.67, 41.94, 29.24, 23.36. HRMS (ESI+) m/z calc. for C 24 H 26 N 2 O 5 422.1842, found [M + H] + 423.1910. Rf = 0.45 (EtOAc/n-Hex; 1:1, v / v ). 2-(1-(2,4-Dimethoxybenzyl)-2-hexyl-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 30 ), yield: 11%, brown oil. The reaction was carried out according to General Procedure I using N -(2,4-Dimethoxybenzyl)heptan-1-imine ( 17 ) (2.31 g, 8.76 mmol, 1EQ), triethylamine (3.05 mL, 21.9 mmol, 2.5 EQ) and 2-(1,3-dioxoisoindolin-2-yl)acetyl chloride ( 18 ) (2.53 g, 11.3 mmol, 1.3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:1 ( v / v ) as eluent. 1H NMR (400 MHz, CDCl 3 ) δ 7.93–7.66 (m, 4H), 7.07–6.94 (m, 1H), 6.59–6.39 (m, 2H), 5.32–5.15 (m, 1H), 4.80–4.69 (m, 1H), 4.67–4.54 (m, 1H), 3.92–3.87 (m, 2H), 3.84–3.75 (m, 6H), 2.04–1.93 (m, 1H), 1.39–1.04 (m, 8H), 0.92–0.79 (m, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 168.03, 164.42, 160.42, 157.82, 134.02, 132.33, 123.53, 119.31, 113.91, 104.26, 98.29, 60.42, 55.39, 55.31, 42.61, 39.88, 31.14, 30.29, 29.37, 28.29, 22.46, 14.02. HRMS (ESI+) m/z calc. for C 26 H 30 N 2 O 5 450.2155, found [M + H] + 451.2229. Rf = 0.42 (EtOAc/n-Hex; 1:1, v / v ).
32 – 34 ) A solution of an appropriate Schiff base (2 EQ) and diazoketone (1 EQ) in 1,2 dimethoxyethane (3 mL) was stirred for 20–30 min at 180 °C in a microwave reactor. The volatiles were then removed in vacuo, and the crude product thus obtained was purified by silica gel column chromatography using EtOAc: Hexane (1:1) as an eluent. Benzyl ((1-benzyl-2-oxo-4-phenylazetidin-3-yl)methyl)carbamate ( 32 ), yield: 31%, brown oil. The reaction was carried out according to General Procedure II using N -benzyl-1-phenylmethanimine ( 9 ) (250 mg, 1.28 mmol, 2 EQ) and benzyl (3-diazo-2-oxopropyl)carbamate ( 31 ) (179 mg, 0.64 mmol, 1 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 7.42 (d, J = 1.6 Hz, 1H), 7.40–7.30 (m, 10H), 7.25–7.21 (m, 3H), 7.12 (dd, J = 6.9, 2.5 Hz, 2H), 5.18–5.13 (m, 1H), 5.11 (d, J = 12.3 Hz, 1H), 4.98 (d, J = 12.3 Hz, 1H), 4.83 (d, J = 15.0 Hz, 1H), 4.28 (d, J = 2.0 Hz, 1H), 3.76 (d, J = 14.7 Hz, 1H), 3.64–3.59 (m, 1H), 3.20–3.15 (m, 1H). 13 C NMR (100 MHz, CDCl 3 ) δ 166.69, 156.61, 137.00, 135.31, 134.51, 128.99, 128.79, 128.54, 128.39, 128.33, 128.17, 128.01, 127.73, 126.51, 69.10, 60.36, 57.72, 48.85, 44.57. HRMS (ESI+) m/z calc. for C 25 H 24 N 2 O 3 400.1787, found [M + H] + 401.1856. Rf = 0.27. (EtOAc/n-Hex; 1:1, v / v ). Benzyl ((2-(benzo[b]thiophen-2-yl)-1-benzyl-4-oxoazetidin-3-yl)methyl)carbamate ( 33 ), yield: 35%, yellow oil. The reaction was carried out according to General Procedure II using 1-(benzo[ b ]thiophen-2-yl)- N -benzylmethanimine ( 12 ) (250 mg, 1.00 mmol, 2 EQ) and benzyl (3-diazo-2-oxopropyl)carbamate ( 31 ) (117 mg, 0.50 mmol, 1 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:2 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 7.81 (d, J = 8.4 Hz, 1H), 7.70 (d, J = 8.6 Hz, 1H), 7.38–7.30 (m, 13H), 5.13 (s, 2H), 4.98 (d, J = 12.2 Hz, 1H), 4.87 (d, J = 15.1 Hz, 1H), 4.67 (d, J = 1.0 Hz, 1H), 4.33 (s, 1H), 3.86 (d, J = 15.1 Hz, 1H). 13 C NMR (100 MHz, CDCl 3 ) δ 167.35, 156.67, 141.88, 139.57, 139.44, 136.25, 135.16, 128.88,128.80, 128.78, 128.75, 128.56, 128.38, 128.21, 128.14, 128.02, 127.87, 124.82, 124.65, 123.68, 123.26, 122.60, 66.99, 60.92, 54.14, 52.36, 44.81. HRMS (ESI+) m/z calc. for C 27 H 24 N 2 O 3 S 456.1508, found [M + H] + 457.1579. Rf = 0.24 (EtOAc/n-Hex; 1:2, v / v ). Benzyl ((2-(benzo[d][1,3]dioxol-5-yl)-1-benzyl-4-oxoazetidin-3-yl)methyl)carbamate ( 34 ), yield: 29%, colorless amorphous solid. The reaction was carried out according to General Procedure II using 1-(benzo[d][1,3]dioxol-5-yl)- N -benzylmethanimine ( 13 ) (250 mg, 1.05 mmol, 2 EQ) and benzyl (3-diazo-2-oxopropyl)carbamate ( 31 ) (123 mg, 0.53 mmol, 1 EQ) Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:1 ( v / v ) as eluent. 1 H NMR (400 MHz, DMSO- d 6 ) δ 7.51 (t, J = 6.0 Hz, 1H), 7.38–7.25 (m, 9H), 7.16 (d, J = 6.7 Hz, 2H), 6.85 (d, J = 7.9 Hz, 1H), 6.79 (s, 1H), 6.67 (d, J = 7.9 Hz, 1H), 6.01 (d, J = 1.3 Hz, 2H), 5.02 (d, J = 12.6 Hz, 1H), 4.96 (d, J = 12.6 Hz, 1H), 4.65 (d, J = 15.6 Hz, 1H), 4.31 (d, J = 1.8 Hz, 1H), 3.81 (d, J = 15.6 Hz, 1H), 3.70–3.52 (m, 1H), 3.40–3.34 (m, 1H), 3.09 (ddd, J = 7.3, 5.1, 1.9 Hz, 1H). 13 C NMR (100 MHz, DMSO- d 6 ) δ 167.59, 156.79, 148.26, 147.60, 137.54, 136.50, 132.07, 129.07, 128.79, 128.23, 128.07, 127.81, 120.55, 108.85, 106.72, 101.59, 65.81, 60.66, 58.45, 51.93, 44.12. HRMS (ESI+) m/z calc. for C 26 H 24 N 2 O 5 444.1685, found [M + H] + 445.1752. Rf = 0.31 (EtOAc/n-Hex; 1:1, v / v ).
35 – 37 ) To a solution of an appropriate Schiff base (1 EQ) in dry dichloromethane (2 mL/mmol) was added dry tetrahydrofuran (5 mL/mmol), and the reaction mixture was cooled to −60 °C before mixed acid anhydride (1.3 EQ) was added slowly. Next, triethylamine (1.5 EQ) was added dropwise over a period of 30 min. Upon complete addition, the reaction mixture was allowed to warm to room temperature with stirring overnight. The volatiles were removed in vacuo, and the solid residue thus obtained dissolved in ethyl acetate. The organic phase was washed with 0.1 M HCl (aq), saturated NaHCO 3 (aq) and brine, dried (Na 2 SO 4 ), filtered, then concentrated in vacuo. The crude product thus obtained was then purified by silica gel column chromatography using EtOAc: Hex as eluent or recrystallized from methyl tert-butyl ether. tert -butyl (2-(4-cyanophenyl)-1-(2,4-dimethoxybenzyl)-4-oxoazetidin-3-yl)carbamate ( 35 ), yield: 11%, colorless amorphous solid. The reaction was carried out according to General Procedure III using 4-(((2,4-dimethoxybenzyl)imino)methyl)benzonitrile ( 3 ), (14.16 g, 50.5 mmol, 1 EQ), N -( tert -butoxycarbonyl)glycine (11.5 g, 65.6 mmol, 1.3 EQ) and triethylamine (10.5 mL, 75.7 mmol, 1.5 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:2 ( v / v ) as eluent. 1 H NMR (400 MHz, DMSO- d6 ) δ 7.74 (d, J = 8.3 Hz, 2H), 7.52 (d, J = 8.4 Hz, 1H), 7.33 (d, J = 8.2 Hz, 2H), 7.04 (d, J = 8.0 Hz, 1H), 6.46–6.39 (m, 2H), 4.93 (dd, J = 8.4, 5.0 Hz, 1H), 4.70 (d, J = 5.0 Hz, 1H), 4.43 (d, J = 14.5 Hz, 1H), 4.09–3.97 (m, 1H), 3.72 (s, 3H), 3.56 (s, 3H), 1.14 (s, 9H). 13 C NMR (100 MHz, DMSO- d 6 ) δ 166.08, 160.96, 158.59, 155.06, 141.81, 132.00, 131.42, 129.23, 119.33, 115.50, 110.46, 105.06, 98.63, 78.80, 62.76, 61.72, 60.22, 55.68, 55.60, 28.24. HRMS (ESI+) m/z calc. for C 24 H 27 N 3 O 5 437.1951, found [M + H] + 438.2015. Rf = 0.54 (EtOAc/n-Hex; 1:1, v / v ). Benzyl (1-(2,4-dimethoxybenzyl)-2-(4-(dimethylamino)phenyl)-4-oxoazetidin-3-yl)carbamate ( 36 ), yield: 31%, colorless amorphous solid. The reaction was carried out according to General Procedure III using 4-(((2,4-dimethoxybenzyl)imino)methyl)- N,N -dimethylaniline ( 8 ), (1.27 g, 4.25 mmol, 1 EQ), ((benzyloxy)carbonyl)glycine (1. 15 g, 5.52 mmol, 1.3 EQ,) and triethylamine (0.88 mL, 6.37 mmol, 1.5 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 2:3 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 7.39–7.28 (m, 4H), 7.24–6.93 (m, 5H), 6.76–6.60 (m, 2H), 6.48–6.36 (m, 2H), 5.71 (d, J = 47.4 Hz, 1H), 5.14 (dd, J = 21.9, 9.5 Hz, 1H), 4.94 (d, J = 12.0 Hz, 1H), 4.84 (d, J = 14.5 Hz, 1H), 4.18 (d, J = 20.5 Hz, 2H), 3.79 (app s, 6H), 2.98 (s, 6H). 13 C NMR (100 MHz, CDCl 3 ) δ 167.19, 160.87, 158.60, 153.11, 151.08, 131.78, 128.71, 128.52, 128.42, 128.22, 128.02, 127.72, 116.06, 112.03, 104.09, 98.45, 75.00, 67.19, 67.13, 55.41, 55.25, 48.33, 40.41, 38.28. HRMS (ESI+) m/z calc. for C 28 H 31 N 3 O 5 489.2264, found [M + H] + 490.2325. Rf = 0.45 (EtOAc/n-Hex; 1:1, v / v ). Benzyl (2-(benzo[d][1,3]dioxol-5-yl)-1-(2,4-dimethoxybenzyl)-4-oxoazetidin-3-yl)carbamate ( 37 ), yield: 33%, colorless amorphous solid. The reaction was carried out according to General Procedure III using 1-(benzo[d][1,3]dioxol-5-yl)- N -(2,4-dimethoxybenzyl)methanimine ( 14 ) (1.85 g, 6.18 mmol, 1 EQ), ((benzyloxy)carbonyl)glycine (1.68 g, 8.04 mmol, 1.3 EQ,) and triethylamine (1.29 mL, 9.27 mmol, 1.5 EQ). Product was precipitated from methyl tert- butyl ether. 1 H NMR (400 MHz, CDCl 3 ) δ 7.32–7.25 (m, 4H), 7.15 (d, J = 7.2 Hz, 1H), 7.02 (d, J = 8.1 Hz, 1H), 6.76 (d, J = 8.3 Hz, 1H), 6.62–6.58 (m, 2H), 6.39 (d, J = 7.2 Hz, 2H), 5.96 (s, 2H), 5.18 (dd, J = 9.4, 5.0 Hz, 1H), 4.98–4.94 (m, 2H), 4.70 (d, J = 14.4 Hz, 1H), 4.62 (d, J = 4.9 Hz, 1H), 3.99 (d, J = 14.4 Hz, 1H), 3.79 (s, 3H), 3.66 (s, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 166.42, 161.02, 158.56, 155.32, 148.15, 147.66, 135.99, 131.21, 128.46, 128.13, 127.86, 120.64, 115.31, 108.62, 107.02, 104.08, 101.27, 98.40, 77.35, 77.03, 76.72, 66.99, 61.82, 61.13, 55.40, 55.17, 45.81, 39.71, 8.63. HRMS (ESI+) m/z calc. for C 27 H 26 N 2 O 7 490.1740, found [M + H] + 491.1803. Rf = 0.28 (EtOAc/n-Hex; 1:1, v / v ).
38 – 41 ) 4.7.1. Hydrazine Hydrate 3-Amino-4-substituted monocyclic ß-lactams with aromatic substituents: 2-(1-Benzyl-2-oxo-4-(4-(trifluoromethyl)phenyl)azetidin-3-yl)isoindoline-1,3-dione ( 19 ) (225 mg, 0.44 mmol, 1 EQ) was dissolved in dried methanol and placed under argon atmosphere. Hydrazine hydrate (0.046 mL, 0.76 mmol, 1.7 EQ) was added dropwise. The mixture was stirred for 2 h at room temperature. Then, the solvent was evaporated. Anhydrous methanol and 3 drops of concentrated aqueous HCl were added to the solid. After the solid was completely dissolved again, the solvent was evaporated. The solid was again dissolved in anhydrous methanol and stirred for 16 h at room temperature. The precipitate formed was filtered off and the solvent evaporated. The solid was dissolved in dichloromethane and washed with saturated aqueous NaHCO 3 . The aqueous phase was extracted three times with dichloromethane, the combined organic phases were dried over Na 2 SO 4 and the solvents evaporated. The deprotected amine was used directly in the next step without purification. 2-(2-((1-(2,4-Dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-yl)carbamoyl)benzoyl)hydrazin-1-ide ( 38 ), colorless amorphous solid. 1 H NMR (400 MHz, DMSO- d 6 ) δ 9.37 (s, 1H), 9.04 (d, J = 7.9 Hz, 1H), 8.15 (d, J = 8.8 Hz, 2H), 7.45 (d, J = 8.7 Hz, 2H), 7.39–7.27 (m, 2H), 7.20 (td, J = 7.5, 1.4 Hz, 1H), 7.09 (d, J = 8.0 Hz, 1H), 6.49–6.39 (m, 4H), 5.39 (dd, J = 7.8, 5.1 Hz, 1H), 4.91 (d, J = 5.0 Hz, 1H), 4.50 (d, J = 14.5 Hz, 1H), 4.30 (s, 1H), 4.17 (d, J = 14.4 Hz, 1H), 3.72 (s, 3H), 3.59 (s, 3H). MS (ESI + , m / z ) 518.1 ([M − H] − ). Rf = 0.05 (EtOAc). 1-Benzyl-2-oxo-4-(4-(trifluoromethyl)phenyl)azetidin-3-aminium chloride ( 39 ), yield: 91%, colorless amorphous solid. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.69 (s, 3H), 7.74 (d, J = 8.2 Hz, 2H), 7.59 (d, J = 8.2 Hz, 2H), 7.38–7.17 (m, 4H), 5.06 (d, J = 5.4 Hz, 1H), 4.92 (d, J = 5.4 Hz, 1H), 4.69 (d, J = 15.4 Hz, 1H), 4.17 (d, J = 15.4 Hz, 1H). 13 C NMR (100 MHz, DMSO- d 6 ) δ 163.03, 136.83, 135.32, 130.06, 129.63, 129.11, 128.75, 128.19, 125.84, 123.23, 58.90, 57.79, 45.10.HRMS (ESI+) m/z calc. for C 17 H 15 F 3 N 2 O 320.1136, found [M + H] + 321.1208. Rf = 0.61 (EtOAc). 1-(2,4-Dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-aminium chloride ( 40 ), yield: 94%, light brown oil. The reaction was carried out according to the General Procedure using 2-(1-(2,4-dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 24 ) (1.01 g, 2.1 mmol, 1 EQ) and hydrazine hydrate (0.213 mL, 3.5 mmol, 1.7 EQ). Product was used directly in the next step without purification. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.23–8.13 (m, 2H), 7.48–7.38 (m, 2H), 7.03 (dd, J = 11.5, 6.3 Hz, 1H), 6.48–6.37 (m, 2H), 4.68 (d, J = 5.1 Hz, 1H), 4.45 (d, J = 14.7 Hz, 1H), 4.43 (d, J = 5.3 Hz, 1H), 4.00 (d, J = 14.5 Hz, 1H), 3.71 (s, 3H), 3.56 (s, 3H). 13 C NMR (100 MHz, DMSO- d 6 ) δ 170.45, 160.89, 158.58, 147.23, 145.34, 131.29, 129.27, 123.55, 115.89, 105.07, 98.64, 71.70, 66.04, 62.52, 55.66, 55.61. HRMS (ESI+) m/z calc. for C 18 H 19 N 3 O 5 357.1325, found [M + H] + 358.1390. Rf = 0.51 (DCM/ iPrOH; 11:1, v / v ). 3-Amino-4-substituted monocyclic ß-lactams with aliphatic substituents: 2-(1-(2,4-Dimethoxybenzyl)-2-isobutyl-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 29 ) (200 mg, 0.47 mmol, 1 EQ) in was dissolved in dried methanol and placed under argon atmosphere. Hydrazine hydrate (0.088 mL, 1.42 mmol, 3 EQ) was added dropwise. The mixture was stirred for 2 h at room temperature. Then the solvent was evaporated, and the residue was redissolved in ethyl acetate. The organic phase was washed with saturated aqueous NaHCO 3 and brine, dried over Na 2 SO 4 and the solvent was evaporated. Mixture of isomers cis and trans 1-(2,4-dimethoxybenzyl)-2-isobutyl-4-oxoazetidin-3-aminium chloride ( 41 ), quantitative yield, colorless amorphous solid. 1 H NMR (400 MHz, CDCl 3 ) δ 7.34 (d, J = 14.6 Hz, 1H), 6.88 (d, J = 8.3 Hz, 1H), 6.80 (d, J = 8.4 Hz, 1H), 6.48–6.37 (m, 4H), 6.32 (d, J = 13.9 Hz, 1H), 5.09 (dd, J = 13.9, 7.3 Hz, 1H), 4.93 (dd, J = 14.6, 7.2 Hz, 1H), 4.78 (s, 2H), 4.55 (s, 2H), 3.83 (d, J = 6.3 Hz, 6H), 3.79 (d, J = 5.4 Hz, 6H), 3.64 (s, 2H), 3.43 (s, 2H), 2.34 (td, J = 13.7, 7.0 Hz, 1H), 2.26 (td, J = 13.5, 6.8 Hz, 1H), 1.02–0.95 (m, 6H), 0.95–0.89 (m, 6H). 13 C NMR (100 MHz, CDCl 3 ) δ 172.10, 171.86, 160.14, 159.82, 157.74, 157.36, 128.01, 126.41, 124.38, 123.70, 123.44, 119.97, 117.29, 116.10, 104.07, 104.01, 98.37, 98.22, 55.39, 55.34, 55.29, 55.26, 44.05, 43.74, 42.80, 42.13, 29.50, 29.45, 23.11, 22.92. MS (ESI+) m/z calc. for C 16 H 24 N 2 O 3 292.1787, found [M + H] + 293.1853. Rf = 0.07 (EtOAc). 4.7.2. Methylhydrazine 2-(1-(2,4-Dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-yl) isoindoline-1,3-dione ( 24 ) (250 mg, 0.51 mmol, 1 EQ) was dissolved in dry methanol and methylhydrazine (0.081 mL, 1.54 mmol, 3 EQ) was added. The reaction was stirred at room temperature. After 4 h, additional methylhydrazine (0.11 mL, 2.12 mmol, 4 EQ) was added and stirred overnight. As the reaction was not yet complete, further methylhydrazine (0.11 mL, 2.12 mmol, 4 EQ) was added, and the reaction was left at room temperature for an additional 72 h. The reaction was allowed to proceed to completion. The organic phase was washed with saturated NaHCO 3 solution and brine and dried over Na 2 SO 4 . The solvent was evaporated, and the product was purified by silica gel column chromatography using DCM:iPrOH = 11:1 as eluent. 4.7.3. Ethanolamine 2-(1-(2,4-Dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-yl) isoindoline-1,3-dione ( 24 ) (250 mg, 0.51 mmol, 1 EQ) was dissolved in ethyl acetate. Ethanolamine (0.46 mL, 7.7 mmol, 15 EQ) was added, and the reaction mixture was refluxed (80 °C) for 2 h. Then the reaction mixture was cooled to room temperature, and a saturated solution of NaHCO 3 and additional ethyl acetate were added. The organic phase was washed with brine and dried over Na 2 SO 4 . The solvent was evaporated, and the product was purified by column chromatography using DCM:iPrOH = 11:1 as eluent. 4.7.4. Ethylenediamine 2-(1-(2,4-Dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-yl) isoindoline-1,3-dione (24) (250 mg, 0.51 mmol, 1 EQ) was dissolved in ethyl acetate. Ethylendiamine solution (1 M in ethyl acetate; 0.67 mL in 10 mL of ethyl acetate, 10 mmol, 19.5 EQ) was added, and the reaction mixture was stirred overnight at room temperature. After 16 h saturated solution of NaHCO3 and additional ethyl acetate were added. The organic phase was washed with brine and dried over Na 2 SO 4 . The solvent was evaporated and the product purified by silica gel column chromatography using DCM:iPrOH = 11:1 as eluent.
3-Amino-4-substituted monocyclic ß-lactams with aromatic substituents: 2-(1-Benzyl-2-oxo-4-(4-(trifluoromethyl)phenyl)azetidin-3-yl)isoindoline-1,3-dione ( 19 ) (225 mg, 0.44 mmol, 1 EQ) was dissolved in dried methanol and placed under argon atmosphere. Hydrazine hydrate (0.046 mL, 0.76 mmol, 1.7 EQ) was added dropwise. The mixture was stirred for 2 h at room temperature. Then, the solvent was evaporated. Anhydrous methanol and 3 drops of concentrated aqueous HCl were added to the solid. After the solid was completely dissolved again, the solvent was evaporated. The solid was again dissolved in anhydrous methanol and stirred for 16 h at room temperature. The precipitate formed was filtered off and the solvent evaporated. The solid was dissolved in dichloromethane and washed with saturated aqueous NaHCO 3 . The aqueous phase was extracted three times with dichloromethane, the combined organic phases were dried over Na 2 SO 4 and the solvents evaporated. The deprotected amine was used directly in the next step without purification. 2-(2-((1-(2,4-Dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-yl)carbamoyl)benzoyl)hydrazin-1-ide ( 38 ), colorless amorphous solid. 1 H NMR (400 MHz, DMSO- d 6 ) δ 9.37 (s, 1H), 9.04 (d, J = 7.9 Hz, 1H), 8.15 (d, J = 8.8 Hz, 2H), 7.45 (d, J = 8.7 Hz, 2H), 7.39–7.27 (m, 2H), 7.20 (td, J = 7.5, 1.4 Hz, 1H), 7.09 (d, J = 8.0 Hz, 1H), 6.49–6.39 (m, 4H), 5.39 (dd, J = 7.8, 5.1 Hz, 1H), 4.91 (d, J = 5.0 Hz, 1H), 4.50 (d, J = 14.5 Hz, 1H), 4.30 (s, 1H), 4.17 (d, J = 14.4 Hz, 1H), 3.72 (s, 3H), 3.59 (s, 3H). MS (ESI + , m / z ) 518.1 ([M − H] − ). Rf = 0.05 (EtOAc). 1-Benzyl-2-oxo-4-(4-(trifluoromethyl)phenyl)azetidin-3-aminium chloride ( 39 ), yield: 91%, colorless amorphous solid. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.69 (s, 3H), 7.74 (d, J = 8.2 Hz, 2H), 7.59 (d, J = 8.2 Hz, 2H), 7.38–7.17 (m, 4H), 5.06 (d, J = 5.4 Hz, 1H), 4.92 (d, J = 5.4 Hz, 1H), 4.69 (d, J = 15.4 Hz, 1H), 4.17 (d, J = 15.4 Hz, 1H). 13 C NMR (100 MHz, DMSO- d 6 ) δ 163.03, 136.83, 135.32, 130.06, 129.63, 129.11, 128.75, 128.19, 125.84, 123.23, 58.90, 57.79, 45.10.HRMS (ESI+) m/z calc. for C 17 H 15 F 3 N 2 O 320.1136, found [M + H] + 321.1208. Rf = 0.61 (EtOAc). 1-(2,4-Dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-aminium chloride ( 40 ), yield: 94%, light brown oil. The reaction was carried out according to the General Procedure using 2-(1-(2,4-dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 24 ) (1.01 g, 2.1 mmol, 1 EQ) and hydrazine hydrate (0.213 mL, 3.5 mmol, 1.7 EQ). Product was used directly in the next step without purification. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.23–8.13 (m, 2H), 7.48–7.38 (m, 2H), 7.03 (dd, J = 11.5, 6.3 Hz, 1H), 6.48–6.37 (m, 2H), 4.68 (d, J = 5.1 Hz, 1H), 4.45 (d, J = 14.7 Hz, 1H), 4.43 (d, J = 5.3 Hz, 1H), 4.00 (d, J = 14.5 Hz, 1H), 3.71 (s, 3H), 3.56 (s, 3H). 13 C NMR (100 MHz, DMSO- d 6 ) δ 170.45, 160.89, 158.58, 147.23, 145.34, 131.29, 129.27, 123.55, 115.89, 105.07, 98.64, 71.70, 66.04, 62.52, 55.66, 55.61. HRMS (ESI+) m/z calc. for C 18 H 19 N 3 O 5 357.1325, found [M + H] + 358.1390. Rf = 0.51 (DCM/ iPrOH; 11:1, v / v ). 3-Amino-4-substituted monocyclic ß-lactams with aliphatic substituents: 2-(1-(2,4-Dimethoxybenzyl)-2-isobutyl-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 29 ) (200 mg, 0.47 mmol, 1 EQ) in was dissolved in dried methanol and placed under argon atmosphere. Hydrazine hydrate (0.088 mL, 1.42 mmol, 3 EQ) was added dropwise. The mixture was stirred for 2 h at room temperature. Then the solvent was evaporated, and the residue was redissolved in ethyl acetate. The organic phase was washed with saturated aqueous NaHCO 3 and brine, dried over Na 2 SO 4 and the solvent was evaporated. Mixture of isomers cis and trans 1-(2,4-dimethoxybenzyl)-2-isobutyl-4-oxoazetidin-3-aminium chloride ( 41 ), quantitative yield, colorless amorphous solid. 1 H NMR (400 MHz, CDCl 3 ) δ 7.34 (d, J = 14.6 Hz, 1H), 6.88 (d, J = 8.3 Hz, 1H), 6.80 (d, J = 8.4 Hz, 1H), 6.48–6.37 (m, 4H), 6.32 (d, J = 13.9 Hz, 1H), 5.09 (dd, J = 13.9, 7.3 Hz, 1H), 4.93 (dd, J = 14.6, 7.2 Hz, 1H), 4.78 (s, 2H), 4.55 (s, 2H), 3.83 (d, J = 6.3 Hz, 6H), 3.79 (d, J = 5.4 Hz, 6H), 3.64 (s, 2H), 3.43 (s, 2H), 2.34 (td, J = 13.7, 7.0 Hz, 1H), 2.26 (td, J = 13.5, 6.8 Hz, 1H), 1.02–0.95 (m, 6H), 0.95–0.89 (m, 6H). 13 C NMR (100 MHz, CDCl 3 ) δ 172.10, 171.86, 160.14, 159.82, 157.74, 157.36, 128.01, 126.41, 124.38, 123.70, 123.44, 119.97, 117.29, 116.10, 104.07, 104.01, 98.37, 98.22, 55.39, 55.34, 55.29, 55.26, 44.05, 43.74, 42.80, 42.13, 29.50, 29.45, 23.11, 22.92. MS (ESI+) m/z calc. for C 16 H 24 N 2 O 3 292.1787, found [M + H] + 293.1853. Rf = 0.07 (EtOAc).
2-(1-(2,4-Dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-yl) isoindoline-1,3-dione ( 24 ) (250 mg, 0.51 mmol, 1 EQ) was dissolved in dry methanol and methylhydrazine (0.081 mL, 1.54 mmol, 3 EQ) was added. The reaction was stirred at room temperature. After 4 h, additional methylhydrazine (0.11 mL, 2.12 mmol, 4 EQ) was added and stirred overnight. As the reaction was not yet complete, further methylhydrazine (0.11 mL, 2.12 mmol, 4 EQ) was added, and the reaction was left at room temperature for an additional 72 h. The reaction was allowed to proceed to completion. The organic phase was washed with saturated NaHCO 3 solution and brine and dried over Na 2 SO 4 . The solvent was evaporated, and the product was purified by silica gel column chromatography using DCM:iPrOH = 11:1 as eluent.
2-(1-(2,4-Dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-yl) isoindoline-1,3-dione ( 24 ) (250 mg, 0.51 mmol, 1 EQ) was dissolved in ethyl acetate. Ethanolamine (0.46 mL, 7.7 mmol, 15 EQ) was added, and the reaction mixture was refluxed (80 °C) for 2 h. Then the reaction mixture was cooled to room temperature, and a saturated solution of NaHCO 3 and additional ethyl acetate were added. The organic phase was washed with brine and dried over Na 2 SO 4 . The solvent was evaporated, and the product was purified by column chromatography using DCM:iPrOH = 11:1 as eluent.
2-(1-(2,4-Dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-yl) isoindoline-1,3-dione (24) (250 mg, 0.51 mmol, 1 EQ) was dissolved in ethyl acetate. Ethylendiamine solution (1 M in ethyl acetate; 0.67 mL in 10 mL of ethyl acetate, 10 mmol, 19.5 EQ) was added, and the reaction mixture was stirred overnight at room temperature. After 16 h saturated solution of NaHCO3 and additional ethyl acetate were added. The organic phase was washed with brine and dried over Na 2 SO 4 . The solvent was evaporated and the product purified by silica gel column chromatography using DCM:iPrOH = 11:1 as eluent.
43 , 51 – 53 ) In a flame-dried flask, 3-amino ß-lactam (1 EQ) was dissolved in dry dichloromethane. Triethylamine (1.1 EQ), di- tert -butyl dicarbonate (1.5 EQ) and 4-(dimethylamino)pyridine (catalytic amount) were added, and the solution was stirred overnight at room temperature. The solvent was evaporated, and the crude product was purified by silica gel column chromatography using EtOAc: Hex as eluent. tert -butyl (1-benzyl-2-isobutyl-4-oxoazetidin-3-yl)carbamate ( 43 ), yield: 51%, colorless amorphous solid. The reaction was carried out according to General Procedure using 3-amino-1-(benzyl)-4-isobutylazetidin-2-one (160 mg, 0.69 mmol, 1 EQ), triethylamine (0.11 mL, 0.76 mmol, 1.1 EQ), di- tert -butyl dicarbonate (226 mg, 1.03 mmol, 1.5 EQ) and 4-(dimethylamino)pyridine (catalytic amount). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:2 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 7.37–7.08 (m, 5H), 5.60–5.41 (m, 1H), 5.21–4.93 (m, 1H), 4.88–4.65 (m, 2H), 4.17–3.91 (m, 2H), 2.39–2.19 (m, 1H), 1.52–1.38 (m, 9H), 0.98–0.90 (m, 6H). 13 C NMR (100 MHz, CDCl 3 ) δ 167.47, 155.82, 136.71, 128.52, 127.10, 123.16, 79.76, 47.86, 42.96, 29.54, 28.35, 27.91, 23.01, 22.74. HRMS (ESI+) m/z calc. for C 19 H 28 N 2 O 3 332.2100, found [M + H] + 333.2166. Rf = 0.63 (EtOAc/n-Hex; 1:1, v / v ). tert -butyl (1-benzyl-2-oxo-4-(4-(trifluoromethyl)phenyl)azetidin-3-yl)carbamate ( 51 ), yield: 47%, colorless amorphous solid. The reaction was carried out according to General Procedure using 1-benzyl-2-oxo-4-(4-(trifluoromethyl)phenyl)azetidin-3-aminium chloride ( 39 ) (143 mg, 0.4 mmol, 1 EQ), triethylamine (0.061 mL, 0.44 mmol, 1 EQ), di- tert -butyl dicarbonate (130 mg, 0.6 mmol, 1.5 EQ) and 4-(dimethylamino)pyridine (catalytic amount). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:2 ( v / v ) as eluent. 1 H NMR (400 MHz, DMSO- d 6 ) δ 7.64 (d, J = 8.1 Hz, 2H), 7.55 (d, J = 8.3 Hz, 1H), 7.39 (d, J = 8.1 Hz, 2H), 7.34–7.25 (m, 3H), 7.22 (d, J = 6.6 Hz, 2H), 5.05 (dd, J = 8.3, 5.0 Hz, 1H), 4.85 (d, J = 4.9 Hz, 1H), 4.66 (d, J = 15.4 Hz, 1H), 4.13 (d, J = 15.4 Hz, 1H), 1.39 (s, 9H). 13 C NMR (100 MHz, CDCl 3 ) δ 166.14, 154.27, 138.46, 134.44, 128.99, 128.53, 128.20, 127.83, 125.56, 125.54, 80.39, 62.46, 60.92, 45.15, 27.86. HRMS (ESI+) m/z calc. for C 22 H 23 F 3 N 2 O 3 420.1661, found [M + Na] + 443.1550. Rf = 0.58 (EtOAc/n-Hex; 1:1, v / v ). tert -butyl (1-(2,4-dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-yl)carbamate ( 52 ), yield: 60%, colorless amorphous solid. The reaction was carried out according to General Procedure using 1-(2,4-dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-aminium chloride ( 40 ) (143 mg, 0.4 mmol, 1 EQ), triethylamine (0.061 mL, 0.44 mmol, 1 EQ), di- tert -butyl dicarbonate (130 mg, 0.6 mmol, 1.5 EQ) and 4-(dimethylamino)pyridine (catalytic amount). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:2 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 8.18 (d, J = 8.6 Hz, 2H), 7.32 (d, J = 8.7 Hz, 2H), 7.04 (d, J = 8.3 Hz, 1H), 6.39 (dd, J = 8.3, 2.3 Hz, 1H), 6.33 (d, J = 2.3 Hz, 1H), 5.15 (dd, J = 8.1, 5.0 Hz, 1H), 4.86 (d, J = 8.1 Hz, 1H), 4.77 (d, J = 4.9 Hz, 1H), 4.69 (d, J = 14.3 Hz, 1H), 4.13 (d, J = 14.3 Hz, 1H), 3.78 (s, 3H), 3.57 (s, 3H), 1.19 (s, 9H). 13 C NMR (100 MHz, CDCl 3 ) δ 165.89, 161.27, 158.50, 154.33, 147.63, 143.08, 131.45, 128.10, 123.40, 114.79, 104.28, 98.36, 80.46, 62.34, 61.42, 55.42, 55.03, 40.41, 27.90. HRMS (ESI+) m/z calc. for C 23 H 27 N 3 O 7 457.1849, found [M + H] + 458.1917. Rf = 0.57 (EtOAc/n-Hex; 1:1, v / v ). tert -butyl (1-(2,4-dimethoxybenzyl)-2-isobutyl-4-oxoazetidin-3-yl)carbamate ( 53 ), yield: 54%, colorless amorphous solid. The reaction was carried out according to General Procedure using 3-amino-1-(2,4-dimethoxybenzyl)-4-isobutylazetidin-2-one ( 41 ) (138 mg, 0.47 mmol, 1 EQ), triethylamine (0.072 mL, 0.52 mmol, 1.1 EQ), di- tert -butyl dicarbonate (155 mg, 0.71 mmol, 1.5 EQ) and 4-(dimethylamino)pyridine (catalytic amount). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 6.88 (d, J = 8.3 Hz, 1H), 6.47–6.22 (m, 3H), 5.52 (br s, 1H), 5.20–4.90 (m, 1H), 4.80–4.50 (m, 2H), 4.17–3.92 (m, 2H), 3.85–3.75 (m, 6H), 2.38–2.17 (m, 1H), 1.49–1.44 (m, 9H), 0.96 (d, J = 6.7 Hz, 3H), 0.93 (d, J = 6.7 Hz, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 167.35, 159.93, 157.42, 154.13, 128.19, 123.12, 104.07, 104.01, 98.26, 83.65, 55.26, 43.29, 43.01, 42.77, 42.40, 29.47, 27.94, 22.79. HRMS (ESI+) m/z calc. for C 21 H 32 N 2 O 5 392.2311, found [M + H] + 393.2384. Rf = 0.63 (EtOAc/n-Hex; 1:1, v / v ).
42 – 44 ) Birch Reduction In a flame-dried flask, Na dispersion in mineral oil (25 wt%, TCI, 6 EQ) and 15-crown-5 (6 EQ) were dissolved in dry tetrahydrofuran. The solution was warmed to room temperature under argon and stirred vigorously for 5 min. Then, the reaction mixture was cooled to 0 °C before a solution of ß-lactam (1 EQ), and isopropanol (6 EQ) in tetrahydrofuran was slowly added. After 15 min, the reaction was stopped by the addition of a saturated aqueous solution of NaHCO 3 and diethyl ether. The aqueous phase was extracted with diethyl ether (2 × 30 mL). The combined organic phases were dried (Na 2 SO 4 ), filtered, then concentrated in vacuo. The crude product thus obtained was purified by silica gel column chromatography using EtOAc:Hex as eluent. N-(2,4-dimethoxybenzyl)-3-(p-tolyl)propanamide ( 42 ), colorless amorphous solid. 1 H NMR (400 MHz, CDCl 3 ) δ 7.10 (d, J = 8.0 Hz, 1H), 7.05 (s, 4H), 6.46–6.37 (m, 2H), 5.76 (br s, 1H), 4.32 (d, J = 5.7 Hz, 2H), 3.80 (s, 3H), 3.77 (s, 3H), 2.94–2.85 (m, 2H), 2.47–2.38 (m, 2H), 2.30 (s, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 171.69, 160.48, 158.52, 137.86, 135.54, 130.53, 129.12, 128.21, 118.88, 103.88, 98.55, 55.41, 55.28, 38.86, 38.72, 31.29, 21.00. HRMS (ESI+) m/z calc. for C 19 H 23 NO 3 313.1678, found [M + H] + 314.1745. Rf = 0.51 (EtOAc/n-Hex; 1:1, v / v ). tert -butyl (2-isobutyl-4-oxoazetidin-3-yl)carbamate ( 44 ), yield: 89%, transparent oil. The reaction was carried out according to General Procedure using Na dispersion in mineral oil (25 wt%, TCI, 131 mg, 13.5 mmol, 6 EQ),15-crown-5 were (0.283 mL, 13.5 mmol, 6 EQ), tert -butyl (1-benzyl-2-isobutyl-4-oxoazetidin-3-yl) carbamate ( 43 ) (75 mg, 2.25 mmol, 1 EQ) and isopropanol (0,109 mL, 13.5 mmol, 6 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:1 ( v / v ) as eluent.. 1 H NMR (400 MHz, CDCl 3 ) δ 7.64 (br s, 1H), 6.69 (ddd, J = 14.3, 10.5, 1.2 Hz, 1H), 5.19 (dd, J = 14.3, 7.0 Hz, 1H), 5.15–5.07 (m, 1H), 3.82 (d, J = 5.9 Hz, 2H), 2.33 (dqd, J = 13.5, 6.8, 1.3 Hz, 1H), 1.47 (s, 9H), 1.01 (s, 3H), 1.00 (s, 3H). HRMS (ESI+) m/z calc. for C 12 H 22 N 2 O 3 242.1630, found [M + H] + 243.1685. Rf = 0.36 (EtOAc/n-Hex; 1:1, v / v ).
In a flame-dried flask, Na dispersion in mineral oil (25 wt%, TCI, 6 EQ) and 15-crown-5 (6 EQ) were dissolved in dry tetrahydrofuran. The solution was warmed to room temperature under argon and stirred vigorously for 5 min. Then, the reaction mixture was cooled to 0 °C before a solution of ß-lactam (1 EQ), and isopropanol (6 EQ) in tetrahydrofuran was slowly added. After 15 min, the reaction was stopped by the addition of a saturated aqueous solution of NaHCO 3 and diethyl ether. The aqueous phase was extracted with diethyl ether (2 × 30 mL). The combined organic phases were dried (Na 2 SO 4 ), filtered, then concentrated in vacuo. The crude product thus obtained was purified by silica gel column chromatography using EtOAc:Hex as eluent. N-(2,4-dimethoxybenzyl)-3-(p-tolyl)propanamide ( 42 ), colorless amorphous solid. 1 H NMR (400 MHz, CDCl 3 ) δ 7.10 (d, J = 8.0 Hz, 1H), 7.05 (s, 4H), 6.46–6.37 (m, 2H), 5.76 (br s, 1H), 4.32 (d, J = 5.7 Hz, 2H), 3.80 (s, 3H), 3.77 (s, 3H), 2.94–2.85 (m, 2H), 2.47–2.38 (m, 2H), 2.30 (s, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 171.69, 160.48, 158.52, 137.86, 135.54, 130.53, 129.12, 128.21, 118.88, 103.88, 98.55, 55.41, 55.28, 38.86, 38.72, 31.29, 21.00. HRMS (ESI+) m/z calc. for C 19 H 23 NO 3 313.1678, found [M + H] + 314.1745. Rf = 0.51 (EtOAc/n-Hex; 1:1, v / v ). tert -butyl (2-isobutyl-4-oxoazetidin-3-yl)carbamate ( 44 ), yield: 89%, transparent oil. The reaction was carried out according to General Procedure using Na dispersion in mineral oil (25 wt%, TCI, 131 mg, 13.5 mmol, 6 EQ),15-crown-5 were (0.283 mL, 13.5 mmol, 6 EQ), tert -butyl (1-benzyl-2-isobutyl-4-oxoazetidin-3-yl) carbamate ( 43 ) (75 mg, 2.25 mmol, 1 EQ) and isopropanol (0,109 mL, 13.5 mmol, 6 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:1 ( v / v ) as eluent.. 1 H NMR (400 MHz, CDCl 3 ) δ 7.64 (br s, 1H), 6.69 (ddd, J = 14.3, 10.5, 1.2 Hz, 1H), 5.19 (dd, J = 14.3, 7.0 Hz, 1H), 5.15–5.07 (m, 1H), 3.82 (d, J = 5.9 Hz, 2H), 2.33 (dqd, J = 13.5, 6.8, 1.3 Hz, 1H), 1.47 (s, 9H), 1.01 (s, 3H), 1.00 (s, 3H). HRMS (ESI+) m/z calc. for C 12 H 22 N 2 O 3 242.1630, found [M + H] + 243.1685. Rf = 0.36 (EtOAc/n-Hex; 1:1, v / v ).
46 – 50 ) Cerium Ammonium Nitrate ß-Lactam (1 EQ) was dissolved in acetonitrile (25 mL/mmol) and distilled water (20 mL/mmol) and placed under argon. The solution was cooled to −10 °C with a sodium chloride ice bath. Cerium ammonium nitrate (3 EQ) was dissolved in distilled water and added dropwise to the vigorously stirring reaction mixture. The reaction was stirred at −10 °C for 1–2 h and then transferred to a separation funnel containing diethyl ether and saturated aqueous NaHCO 3 . The aqueous phase was washed with diethyl ether. The combined organic phases were dried over Na 2 SO 4, and the solvents were evaporated. The solid was purified by silica gel column chromatography, using the gradient EtOAc: Hex as eluent. tert -butyl (1-(2,4-dimethoxybenzoyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-yl)carbamate ( 44 ), light orange amorphous solid. 1 H NMR (400 MHz, CDCl 3 ) δ 8.25 (d, J = 8.7 Hz, 2H), 7.55–7.51 (m, 1H), 7.52 (d, J = 8.4, 2H), 6.61–6.51 (m, 2H), 5.59 (d, J = 6.4 Hz, 1H), 5.36–5.27 (m, 1H), 4.70 (d, J = 8.4 Hz, 1H), 3.91 (s, 3H), 3.88 (s, 3H), 1.28 (s, 9H). 13 C NMR (100 MHz, CDCl 3 ) δ 164.61, 163.87, 160.10, 154.18, 147.80, 141.44, 132.04, 127.90, 123.80, 115.51, 105.18, 98.65, 81.20, 60.79, 59.56, 55.93, 55.58, 27.95. HRMS (ESI+) m/z calc. for C 23 H 25 N 3 O 8 471.1642, found [M + Na] + 494.1532. Rf = 0.43 (EtOAc/n-Hex; 2:1, v / v ). 2-(2-Oxo-4-(4-(trifluoromethyl)phenyl)azetidin-3-yl)isoindoline-1,3-dione ( 46 ), yield: 65%, colorless amorphous solid. The reaction was carried out according to General Procedure using 4-(1-(2,4-dimethoxybenzyl)-3-(1,3-dioxoisoindolin-2-yl)-4-oxoazetidin-2-yl)benzonitrile ( 21 ) (156 mg, 0.31 mmol, 1 EQ) and cerium ammonium nitrate (502 mg, 0.92 mmol, 3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 7.72–7.60 (m, 4H), 7.54–7.41 (m, 4H), 7.28 (br s, 1H), 5.71 (dd, J = 5.4, 1.8 Hz, 1H), 5.26 (d, J = 5.4 Hz, 1H). 13 C NMR (100 MHz, CDCl 3 ) δ = 166.66, 164.59, 138.82, 134.48, 131.00, 130.45, 130.13, 127.11, 125.43, 123.62, 60.37, 57.23. HRMS (ESI+) m/z calc. for C 18 H 11 F 3 N 2 O 3 360.0722, found [M + H] + 361.0791. Rf = 0.43 (EtOAc/n-Hex; 1:1, v / v ). 2-(2-(3-Bromo-4-fluorophenyl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 47 ), yield: 40%, light yellow amorphous solid. The reaction was carried out according to General Procedure using 2-(2-(4-bromo-3-fluorophenyl)-1-(2,4-dimethoxybenzyl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 22 ) (151 g, 0,29 mmol, 1 EQ) and cerium ammonium nitrate (473 mg, 0.86 mmol, 3EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 2:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 7.76–7.65 (m, 4H), 7.53 (dd, J = 6.4, 2.0 Hz, 1H), 7.29–7.23 (m, 1H), 7.03 (br s, 1H), 6.99 (t, J = 8.4 Hz, 1H), 5.63 (dd, J = 5.3, 1.9 Hz, 1H), 5.14 (d, J = 5.3 Hz, 1H). 13 C NMR (100 MHz, CDCl 3 ) δ 166.65, 164.37, 159.93, 157.46, 134.52, 132.15, 132.12, 131.98, 131.04, 127.46, 127.38, 123.69, 116.66, 116.44, 109.29, 109.07, 60.45, 56.58. HRMS (ESI+) m/z calc. for C 17 H 10 BrFN 2 O 3 387.9859, found [M + H] + 388.9936. Rf = 0.43 (EtOAc/n-Hex; 2:1, v / v ). tert -butyl (2-(4-nitrophenyl)-4-oxoazetidin-3-yl)carbamate ( 48 ), yield: 51%, light red amorphous solid. The reaction was carried out according to General Procedure using tert -butyl (1-(2,4-dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-yl)carbamate ( 52 ) (2.5 g, 5.4 mmol, 1 EQ) and cerium ammonium nitrate (9 g, 16.4 mmol, 3 EQ). Product was purified by silica gel column chromatography using the gradient EtOAc: Hex = 1:1 to 4:1 as eluent. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.71 (br s, 1H), 8.22 (d, J = 8.7 Hz, 2H), 7.46 (d, J = 8.6 Hz, 2H), 7.42 (d, J = 8.4 Hz, 1H), 5. 41–5.36 (m, 1H), 5.05–4.93 (m, 1H), 1.18 (s, J = 12.9 Hz, 9H). 13 C NMR (100 MHz, DMSO- d 6 ) δ 167.10, 155.14, 147.19, 145.99, 128.87, 123.40, 78.84, 63.68, 56.77, 28.27. HRMS (ESI+) m/z calc. for C 14 H 17 N 3 O 5 307.1168, found [M + H] + 308.1240. Rf = 0.23 (EtOAc/n-Hex; 2:1, v / v ). 2-(2-(4-(Methylsulfonyl)phenyl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 49 ), yield: 43%, yellow amorphous solid. The reaction was carried out according to General Procedure using 2-(1-(2,4-Dimethoxybenzyl)-2-(4-(methylsulfonyl)phenyl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 25 ) (72 mg, 0.14 mmol, 1 EQ) and cerium ammonium nitrate (228 mg, 0.42 mmol, 3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 2:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 7.81 (d, J = 8.4 Hz, 2H), 7.74–7.64 (m, 4H), 7.54 (d, J = 8.2 Hz, 2H), 7.14 (br s, 1H), 5.72 (dd, J = 5.5, 1.9 Hz, 1H), 5.27 (d, J = 5.5 Hz, 1H), 2.90 (s, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 166.58, 164.18, 141.21, 140.20, 134.62, 130.95, 127.79, 127.56, 123.74, 60.48, 57.18, 44.33. HRMS (ESI+) m/z calc. for C 18 H 14 N 2 O 5 S 370.0623, found [M + Na] + 393.0530. Rf = 0.33 (EtOA). 2-(2-(Furan-2-yl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 50 ), yield: 15%, light brown oil. The reaction was carried out according to General Procedure 2-(1-(2,4-dimethoxybenzyl)-2-(furan-2-yl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 26 ) (206 mg, 0.48 mmol, 1 EQ) and cerium ammonium nitrate (0.84 g, 1.43 mmol, 3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 2:1 ( v / v ) as eluent. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.04–7.80 (m, 4H), 7.02 (dd, J = 10.6, 3.7 Hz, 1H), 6.91 (d, J = 7.3 Hz, 1H), 6.08 (dd, J = 10.6, 1.4 Hz, 1H), 5.87 (d, J = 6.0 Hz, 1H), 5.64 (dd, J = 6.0, 3.7 Hz, 1H), 4.55 (d, J = 6.0 Hz, 1H). 13 C NMR (100 MHz, DMSO- d 6 ) δ 191.40, 167.88, 165.11, 146.54, 135.84, 131.53, 128.34, 124.56, 67.63, 56.54, 53.78. HRMS (ESI+) m/z calc. for C 15 H 10 N 2 O 4 282.0641, found [M + H] + 283.0721. Rf = 0.33 (EtOAc/n-Hex; 2:1, v / v ).
ß-Lactam (1 EQ) was dissolved in acetonitrile (25 mL/mmol) and distilled water (20 mL/mmol) and placed under argon. The solution was cooled to −10 °C with a sodium chloride ice bath. Cerium ammonium nitrate (3 EQ) was dissolved in distilled water and added dropwise to the vigorously stirring reaction mixture. The reaction was stirred at −10 °C for 1–2 h and then transferred to a separation funnel containing diethyl ether and saturated aqueous NaHCO 3 . The aqueous phase was washed with diethyl ether. The combined organic phases were dried over Na 2 SO 4, and the solvents were evaporated. The solid was purified by silica gel column chromatography, using the gradient EtOAc: Hex as eluent. tert -butyl (1-(2,4-dimethoxybenzoyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-yl)carbamate ( 44 ), light orange amorphous solid. 1 H NMR (400 MHz, CDCl 3 ) δ 8.25 (d, J = 8.7 Hz, 2H), 7.55–7.51 (m, 1H), 7.52 (d, J = 8.4, 2H), 6.61–6.51 (m, 2H), 5.59 (d, J = 6.4 Hz, 1H), 5.36–5.27 (m, 1H), 4.70 (d, J = 8.4 Hz, 1H), 3.91 (s, 3H), 3.88 (s, 3H), 1.28 (s, 9H). 13 C NMR (100 MHz, CDCl 3 ) δ 164.61, 163.87, 160.10, 154.18, 147.80, 141.44, 132.04, 127.90, 123.80, 115.51, 105.18, 98.65, 81.20, 60.79, 59.56, 55.93, 55.58, 27.95. HRMS (ESI+) m/z calc. for C 23 H 25 N 3 O 8 471.1642, found [M + Na] + 494.1532. Rf = 0.43 (EtOAc/n-Hex; 2:1, v / v ). 2-(2-Oxo-4-(4-(trifluoromethyl)phenyl)azetidin-3-yl)isoindoline-1,3-dione ( 46 ), yield: 65%, colorless amorphous solid. The reaction was carried out according to General Procedure using 4-(1-(2,4-dimethoxybenzyl)-3-(1,3-dioxoisoindolin-2-yl)-4-oxoazetidin-2-yl)benzonitrile ( 21 ) (156 mg, 0.31 mmol, 1 EQ) and cerium ammonium nitrate (502 mg, 0.92 mmol, 3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 1:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 7.72–7.60 (m, 4H), 7.54–7.41 (m, 4H), 7.28 (br s, 1H), 5.71 (dd, J = 5.4, 1.8 Hz, 1H), 5.26 (d, J = 5.4 Hz, 1H). 13 C NMR (100 MHz, CDCl 3 ) δ = 166.66, 164.59, 138.82, 134.48, 131.00, 130.45, 130.13, 127.11, 125.43, 123.62, 60.37, 57.23. HRMS (ESI+) m/z calc. for C 18 H 11 F 3 N 2 O 3 360.0722, found [M + H] + 361.0791. Rf = 0.43 (EtOAc/n-Hex; 1:1, v / v ). 2-(2-(3-Bromo-4-fluorophenyl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 47 ), yield: 40%, light yellow amorphous solid. The reaction was carried out according to General Procedure using 2-(2-(4-bromo-3-fluorophenyl)-1-(2,4-dimethoxybenzyl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 22 ) (151 g, 0,29 mmol, 1 EQ) and cerium ammonium nitrate (473 mg, 0.86 mmol, 3EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 2:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 7.76–7.65 (m, 4H), 7.53 (dd, J = 6.4, 2.0 Hz, 1H), 7.29–7.23 (m, 1H), 7.03 (br s, 1H), 6.99 (t, J = 8.4 Hz, 1H), 5.63 (dd, J = 5.3, 1.9 Hz, 1H), 5.14 (d, J = 5.3 Hz, 1H). 13 C NMR (100 MHz, CDCl 3 ) δ 166.65, 164.37, 159.93, 157.46, 134.52, 132.15, 132.12, 131.98, 131.04, 127.46, 127.38, 123.69, 116.66, 116.44, 109.29, 109.07, 60.45, 56.58. HRMS (ESI+) m/z calc. for C 17 H 10 BrFN 2 O 3 387.9859, found [M + H] + 388.9936. Rf = 0.43 (EtOAc/n-Hex; 2:1, v / v ). tert -butyl (2-(4-nitrophenyl)-4-oxoazetidin-3-yl)carbamate ( 48 ), yield: 51%, light red amorphous solid. The reaction was carried out according to General Procedure using tert -butyl (1-(2,4-dimethoxybenzyl)-2-(4-nitrophenyl)-4-oxoazetidin-3-yl)carbamate ( 52 ) (2.5 g, 5.4 mmol, 1 EQ) and cerium ammonium nitrate (9 g, 16.4 mmol, 3 EQ). Product was purified by silica gel column chromatography using the gradient EtOAc: Hex = 1:1 to 4:1 as eluent. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.71 (br s, 1H), 8.22 (d, J = 8.7 Hz, 2H), 7.46 (d, J = 8.6 Hz, 2H), 7.42 (d, J = 8.4 Hz, 1H), 5. 41–5.36 (m, 1H), 5.05–4.93 (m, 1H), 1.18 (s, J = 12.9 Hz, 9H). 13 C NMR (100 MHz, DMSO- d 6 ) δ 167.10, 155.14, 147.19, 145.99, 128.87, 123.40, 78.84, 63.68, 56.77, 28.27. HRMS (ESI+) m/z calc. for C 14 H 17 N 3 O 5 307.1168, found [M + H] + 308.1240. Rf = 0.23 (EtOAc/n-Hex; 2:1, v / v ). 2-(2-(4-(Methylsulfonyl)phenyl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 49 ), yield: 43%, yellow amorphous solid. The reaction was carried out according to General Procedure using 2-(1-(2,4-Dimethoxybenzyl)-2-(4-(methylsulfonyl)phenyl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 25 ) (72 mg, 0.14 mmol, 1 EQ) and cerium ammonium nitrate (228 mg, 0.42 mmol, 3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 2:1 ( v / v ) as eluent. 1 H NMR (400 MHz, CDCl 3 ) δ 7.81 (d, J = 8.4 Hz, 2H), 7.74–7.64 (m, 4H), 7.54 (d, J = 8.2 Hz, 2H), 7.14 (br s, 1H), 5.72 (dd, J = 5.5, 1.9 Hz, 1H), 5.27 (d, J = 5.5 Hz, 1H), 2.90 (s, 3H). 13 C NMR (100 MHz, CDCl 3 ) δ 166.58, 164.18, 141.21, 140.20, 134.62, 130.95, 127.79, 127.56, 123.74, 60.48, 57.18, 44.33. HRMS (ESI+) m/z calc. for C 18 H 14 N 2 O 5 S 370.0623, found [M + Na] + 393.0530. Rf = 0.33 (EtOA). 2-(2-(Furan-2-yl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 50 ), yield: 15%, light brown oil. The reaction was carried out according to General Procedure 2-(1-(2,4-dimethoxybenzyl)-2-(furan-2-yl)-4-oxoazetidin-3-yl)isoindoline-1,3-dione ( 26 ) (206 mg, 0.48 mmol, 1 EQ) and cerium ammonium nitrate (0.84 g, 1.43 mmol, 3 EQ). Product was purified by silica gel column chromatography using EtOAc: n-Hex = 2:1 ( v / v ) as eluent. 1 H NMR (400 MHz, DMSO- d 6 ) δ 8.04–7.80 (m, 4H), 7.02 (dd, J = 10.6, 3.7 Hz, 1H), 6.91 (d, J = 7.3 Hz, 1H), 6.08 (dd, J = 10.6, 1.4 Hz, 1H), 5.87 (d, J = 6.0 Hz, 1H), 5.64 (dd, J = 6.0, 3.7 Hz, 1H), 4.55 (d, J = 6.0 Hz, 1H). 13 C NMR (100 MHz, DMSO- d 6 ) δ 191.40, 167.88, 165.11, 146.54, 135.84, 131.53, 128.34, 124.56, 67.63, 56.54, 53.78. HRMS (ESI+) m/z calc. for C 15 H 10 N 2 O 4 282.0641, found [M + H] + 283.0721. Rf = 0.33 (EtOAc/n-Hex; 2:1, v / v ).
54 ) With Use of Trifluoroacetic Acid tert -butyl (2-(4-nitrophenyl)-4-oxoazetidin-3-yl)carbamate ( 48 ) (150 mg, 0.5 mmol, 1 EQ) was dissolved in dry dichloromethane (2 mL), anisole (0.49 mL, 4.5 mmol, 9 EQ) was added, and the solution was cooled to −5 °C on a sodium chloride ice bath. Trifluoroacetic acid (1.53 mL, 20 mmol, 40 EQ) was added, and the solution was slowly warmed to room temperature. After stirring for 1.5 h, the solvent and the excess trifluoroacetic acid were evaporated. The residue was precipitated from methyl tert -butyl ether. The solid was used in the next step without further purification. 2-(4-Nitrophenyl)-4-oxoazetidin-3-aminium trifluoroacetate ( 54 ), yield: 83%, brown oil. 1 H NMR (400 MHz, DMSO- d 6 ) δ 9.24 (s, 1H), 8.46 (br s, 3H), 8.30 (d, J = 8.8 Hz, 2H), 7.67 (d, J = 8.6 Hz, 2H), 5.19 (d, J = 5.4 Hz, 1H), 4.86 (dd, J = 5.4, 1.8 Hz, 1H). 13 C NMR (100 MHz, DMSO- d 6 ) δ 163.43, 148.06, 142.37, 129.63, 123.92, 59.90, 49.18. HRMS (ESI+) m/z calc. for C 9 H 9 N 3 O 3 207.0644, found [M + H] + 208.0716. Rf = 0.08 (EtOAc/n-Hex; 9:1, v / v ).
tert -butyl (2-(4-nitrophenyl)-4-oxoazetidin-3-yl)carbamate ( 48 ) (150 mg, 0.5 mmol, 1 EQ) was dissolved in dry dichloromethane (2 mL), anisole (0.49 mL, 4.5 mmol, 9 EQ) was added, and the solution was cooled to −5 °C on a sodium chloride ice bath. Trifluoroacetic acid (1.53 mL, 20 mmol, 40 EQ) was added, and the solution was slowly warmed to room temperature. After stirring for 1.5 h, the solvent and the excess trifluoroacetic acid were evaporated. The residue was precipitated from methyl tert -butyl ether. The solid was used in the next step without further purification. 2-(4-Nitrophenyl)-4-oxoazetidin-3-aminium trifluoroacetate ( 54 ), yield: 83%, brown oil. 1 H NMR (400 MHz, DMSO- d 6 ) δ 9.24 (s, 1H), 8.46 (br s, 3H), 8.30 (d, J = 8.8 Hz, 2H), 7.67 (d, J = 8.6 Hz, 2H), 5.19 (d, J = 5.4 Hz, 1H), 4.86 (dd, J = 5.4, 1.8 Hz, 1H). 13 C NMR (100 MHz, DMSO- d 6 ) δ 163.43, 148.06, 142.37, 129.63, 123.92, 59.90, 49.18. HRMS (ESI+) m/z calc. for C 9 H 9 N 3 O 3 207.0644, found [M + H] + 208.0716. Rf = 0.08 (EtOAc/n-Hex; 9:1, v / v ).
|
Digital cognitive behavioral therapy vs education for pain in adults with sickle cell disease | 1a3b5190-a7fd-4ec5-abf1-c8410aa78b55 | 11699089 | Patient Education as Topic[mh] | Sickle cell disease (SCD), a rare genetic blood disorder affecting ∼100 000 Black Americans, , is characterized by acute complications and chronic pain that impair daily functioning and quality of life. Recurrent vaso-occlusive episodes often require emergency care, and both acute and chronic pain and depression increase health care use and mortality risk. , , , , , Despite the substantial burden, current SCD pain management standards are inadequate, , with overreliance on opioids that have limited long-term efficacy data and known adverse effects. , , , , , Effective nonpharmacological pain treatments are needed. Cognitive behavioral therapy (CBT) is a type of psychotherapy that has shown effectiveness in treating various pain conditions. It focuses on identifying and changing negative thought patterns and behaviors that may contribute to a person’s pain experience. CBT has been proven effective in helping individuals develop coping strategies and more adaptive ways of thinking about and managing their pain. , , , , However, it remains underutilized for SCD, with reported barriers including limited availability or access, cost, and stigma related to seeking mental health services. , , , Emerging mobile health interventions could expand access to CBT and improve health outcomes. , , However, large-scale trials demonstrating digital CBT’s effectiveness and scalability, particularly in minority populations, are lacking. , , , Access barriers to CBT persist, but high-quality SCD and pain education materials are readily available through patient communities. Thus, there is benefit in comparing effectiveness between CBT and educational programs that patients may already be using routinely. Although limited data exist on psychoeducation for SCD pain specifically, evidence in other chronic pain conditions suggests that educational interventions can improve pain and coping. , Given the promising evidence and accessible resources, structured educational interventions have the potential to improve outcomes in SCD. Both CBT and education are viable nonpharmacological options, but evidence is needed to guide providers on feasible and acceptable nonpharmacological SCD pain management. The CaRISMA trial compared digital CBT and digital pain/SCD education (hereafter referred to as “Education”) mobile interventions for reducing pain interference (primary outcome) in adults with SCD at 6-month follow-up. We hypothesized that digital CBT would lead to greater improvement in pain interference, intensity, depression, anxiety, quality of life, and self-efficacy compared with Education. Secondary aims assessed baseline depression as a potential moderator of the effect on the primary outcome.
Participants English-speaking adults (aged ≥18 years) with SCD and chronic pain were recruited from 7 comprehensive sickle cell centers and partnering community-based organizations (CBOs) from 19 August 2020 through 12 April 2022. In-person and web-based recruitment identified eligible patients using screening tools, best practice alerts, and study/CBO websites. The inclusion criteria were (1) SCD diagnosis, (2) chronic pain (≥4 d/wk for ≥3 months) and/or prescribed daily opioids for pain, and (3) smartphone access. To verify SCD diagnoses, we required the virtual enrollees to provide recent official medical documentation (such as hospital records, laboratory tests, or doctors’ letters) or government forms that clearly stated their full name and confirmed their SCD diagnosis or treatment. The exclusion criteria were (1) cognitive dysfunction or low literacy, (2) inability to read, speak, and write English, and (3) unwillingness to participate in either intervention arm. Community organizations were integral partners throughout the CaRISMA study, contributing to protocol development, intervention creation, and health coach identification. They played a crucial role in recruitment efforts across various platforms including web, social media, and community events. Additionally, these organizations provided valuable feedback on study results throughout the project, ensuring that the research remained responsive to community needs and experiences. Design This multisite, randomized, pragmatic, comparative effectiveness trial assigned participants in a 1:1 ratio to either a digital CBT or an SCD and pain education (Education) intervention delivered via Facebook Messenger. Upon confirmation of eligibility, participants were randomly assigned to either digital CBT or Education. Permuted block randomization (varying block sizes of 4, 6, and 8) was stratified by study center to control for site-specific disease education and treatment approaches. The randomization schema was created using Stata (StataCorp) by the lead statistician (K.Z.A.) in the data coordinating center and integrated into the web-based data capture system to ensure allocation concealment. Upon confirmation of eligibility, the web-based system sent a unique code via an automated email and/or text to study participants to link them to their assigned chatbot intervention. Both interventions were cocreated with and tailored for adults with SCD, providing scripted chat interactions and video lessons. The digital CBT focused on behavioral coping skills, whereas Education emphasized self-management through learning about pain and SCD. Participants completed online assessments of pain, self-efficacy, quality of life, depression, and anxiety at 3-, 6-, and 12-month follow-ups, and daily pain diaries between follow-ups. Qualitative interviews were conducted with 48 participants at baseline and 6 months. The qualitative data and 12-month outcomes will be reported in a separate article. The study was approved by the Institutional Review Board of the University of Pittsburgh, and all participants provided written informed consent. Detailed methods are published in the trial protocol. Study outcomes and measures PROMIS Pain Interference - Short Form 8a The PROMIS Pain Interference - Short Form 8a measures the impact of pain on social, cognitive, emotional, physical, and recreational activities. Higher scores indicate greater pain interference. Raw scores (range, 8-40) are normalized to a T-score (mean = 50, standard deviation = 10) based on the general US population. Studies in pain populations show a minimally important difference ranging from 2.0 to 3.0 points. Electronic diary: daily pain intensity and mood Participants entered daily pain scores (0-10 scale), opioid use, and mood via a mobile web application (app). At each follow-up period (baseline, 3, 6, and 12 months), participants received daily reminders for 2 weeks to assess pain but could continue entering data throughout the 365-day study. The mean pain intensity during each 2-week assessment period was used for analyses. PHQ The Patient Health Questionnaire (PHQ)-8/9 assesses depression severity. The PHQ-9, administered only at clinical sites, includes an additional item on suicidality compared with the PHQ-8. Participants with PHQ-2 scores >0 completed the full PHQ-8 or PHQ-9. Items are scored on a 0 to 3 scale, with total scores ranging from 0 to 24 (PHQ-8) or 0 to 27 (PHQ-9). Higher scores indicate more severe depression, with cutoffs of 5 (mild), 10 (moderate), 15 (moderately severe), and 20 (severe). GAD-7 The Generalized Anxiety Disorder scale (GAD)-7 assesses generalized anxiety disorder severity. Participants with GAD-2 scores >0 completed the full GAD-7. Items are scored on a 0 to 3 scale, with total scores ranging from 0 to 21. Higher scores indicate more severe anxiety, with cutoffs of 5 (mild), 10 (moderate), and 15 (severe). ASCQ-Me The ASCQ-Me (Adult Sickle Cell Quality of Life Measurement Information System) assesses health care experiences, stress responses, and social functioning in SCD. Higher scores represent worse SCD-related quality of life. We examined the social functioning and emotional impact subscales. SCSES The 9-item Sickle Cell Self-Efficacy Scale (SCSES) measures self-perceived ability to manage day-to-day medical, mental, and emotional aspects of SCD. Items are scored on a 1 to 5 scale, with higher summed scores indicating greater self-efficacy.
English-speaking adults (aged ≥18 years) with SCD and chronic pain were recruited from 7 comprehensive sickle cell centers and partnering community-based organizations (CBOs) from 19 August 2020 through 12 April 2022. In-person and web-based recruitment identified eligible patients using screening tools, best practice alerts, and study/CBO websites. The inclusion criteria were (1) SCD diagnosis, (2) chronic pain (≥4 d/wk for ≥3 months) and/or prescribed daily opioids for pain, and (3) smartphone access. To verify SCD diagnoses, we required the virtual enrollees to provide recent official medical documentation (such as hospital records, laboratory tests, or doctors’ letters) or government forms that clearly stated their full name and confirmed their SCD diagnosis or treatment. The exclusion criteria were (1) cognitive dysfunction or low literacy, (2) inability to read, speak, and write English, and (3) unwillingness to participate in either intervention arm. Community organizations were integral partners throughout the CaRISMA study, contributing to protocol development, intervention creation, and health coach identification. They played a crucial role in recruitment efforts across various platforms including web, social media, and community events. Additionally, these organizations provided valuable feedback on study results throughout the project, ensuring that the research remained responsive to community needs and experiences.
This multisite, randomized, pragmatic, comparative effectiveness trial assigned participants in a 1:1 ratio to either a digital CBT or an SCD and pain education (Education) intervention delivered via Facebook Messenger. Upon confirmation of eligibility, participants were randomly assigned to either digital CBT or Education. Permuted block randomization (varying block sizes of 4, 6, and 8) was stratified by study center to control for site-specific disease education and treatment approaches. The randomization schema was created using Stata (StataCorp) by the lead statistician (K.Z.A.) in the data coordinating center and integrated into the web-based data capture system to ensure allocation concealment. Upon confirmation of eligibility, the web-based system sent a unique code via an automated email and/or text to study participants to link them to their assigned chatbot intervention. Both interventions were cocreated with and tailored for adults with SCD, providing scripted chat interactions and video lessons. The digital CBT focused on behavioral coping skills, whereas Education emphasized self-management through learning about pain and SCD. Participants completed online assessments of pain, self-efficacy, quality of life, depression, and anxiety at 3-, 6-, and 12-month follow-ups, and daily pain diaries between follow-ups. Qualitative interviews were conducted with 48 participants at baseline and 6 months. The qualitative data and 12-month outcomes will be reported in a separate article. The study was approved by the Institutional Review Board of the University of Pittsburgh, and all participants provided written informed consent. Detailed methods are published in the trial protocol.
PROMIS Pain Interference - Short Form 8a The PROMIS Pain Interference - Short Form 8a measures the impact of pain on social, cognitive, emotional, physical, and recreational activities. Higher scores indicate greater pain interference. Raw scores (range, 8-40) are normalized to a T-score (mean = 50, standard deviation = 10) based on the general US population. Studies in pain populations show a minimally important difference ranging from 2.0 to 3.0 points. Electronic diary: daily pain intensity and mood Participants entered daily pain scores (0-10 scale), opioid use, and mood via a mobile web application (app). At each follow-up period (baseline, 3, 6, and 12 months), participants received daily reminders for 2 weeks to assess pain but could continue entering data throughout the 365-day study. The mean pain intensity during each 2-week assessment period was used for analyses. PHQ The Patient Health Questionnaire (PHQ)-8/9 assesses depression severity. The PHQ-9, administered only at clinical sites, includes an additional item on suicidality compared with the PHQ-8. Participants with PHQ-2 scores >0 completed the full PHQ-8 or PHQ-9. Items are scored on a 0 to 3 scale, with total scores ranging from 0 to 24 (PHQ-8) or 0 to 27 (PHQ-9). Higher scores indicate more severe depression, with cutoffs of 5 (mild), 10 (moderate), 15 (moderately severe), and 20 (severe). GAD-7 The Generalized Anxiety Disorder scale (GAD)-7 assesses generalized anxiety disorder severity. Participants with GAD-2 scores >0 completed the full GAD-7. Items are scored on a 0 to 3 scale, with total scores ranging from 0 to 21. Higher scores indicate more severe anxiety, with cutoffs of 5 (mild), 10 (moderate), and 15 (severe). ASCQ-Me The ASCQ-Me (Adult Sickle Cell Quality of Life Measurement Information System) assesses health care experiences, stress responses, and social functioning in SCD. Higher scores represent worse SCD-related quality of life. We examined the social functioning and emotional impact subscales. SCSES The 9-item Sickle Cell Self-Efficacy Scale (SCSES) measures self-perceived ability to manage day-to-day medical, mental, and emotional aspects of SCD. Items are scored on a 1 to 5 scale, with higher summed scores indicating greater self-efficacy.
The PROMIS Pain Interference - Short Form 8a measures the impact of pain on social, cognitive, emotional, physical, and recreational activities. Higher scores indicate greater pain interference. Raw scores (range, 8-40) are normalized to a T-score (mean = 50, standard deviation = 10) based on the general US population. Studies in pain populations show a minimally important difference ranging from 2.0 to 3.0 points.
Participants entered daily pain scores (0-10 scale), opioid use, and mood via a mobile web application (app). At each follow-up period (baseline, 3, 6, and 12 months), participants received daily reminders for 2 weeks to assess pain but could continue entering data throughout the 365-day study. The mean pain intensity during each 2-week assessment period was used for analyses.
The Patient Health Questionnaire (PHQ)-8/9 assesses depression severity. The PHQ-9, administered only at clinical sites, includes an additional item on suicidality compared with the PHQ-8. Participants with PHQ-2 scores >0 completed the full PHQ-8 or PHQ-9. Items are scored on a 0 to 3 scale, with total scores ranging from 0 to 24 (PHQ-8) or 0 to 27 (PHQ-9). Higher scores indicate more severe depression, with cutoffs of 5 (mild), 10 (moderate), 15 (moderately severe), and 20 (severe).
The Generalized Anxiety Disorder scale (GAD)-7 assesses generalized anxiety disorder severity. Participants with GAD-2 scores >0 completed the full GAD-7. Items are scored on a 0 to 3 scale, with total scores ranging from 0 to 21. Higher scores indicate more severe anxiety, with cutoffs of 5 (mild), 10 (moderate), and 15 (severe).
The ASCQ-Me (Adult Sickle Cell Quality of Life Measurement Information System) assesses health care experiences, stress responses, and social functioning in SCD. Higher scores represent worse SCD-related quality of life. We examined the social functioning and emotional impact subscales.
The 9-item Sickle Cell Self-Efficacy Scale (SCSES) measures self-perceived ability to manage day-to-day medical, mental, and emotional aspects of SCD. Items are scored on a 1 to 5 scale, with higher summed scores indicating greater self-efficacy.
The CaRISMA interventions consisted of a chatbot app accessed through Facebook Messenger, and a health coach. The chatbot was fully scripted and followed prewritten conversation trees codeveloped with the research team and CBO partners. It provided customized responses, educational written content, both instructional and testimonial videos featuring adults with SCD, GIFs, and images based on participant responses. CBT arm The digital CBT arm taught participants how to recognize negative thoughts and emotions, use cognitive skills and problem-solving, and apply coping behaviors. It emphasized skills acquisition through practice, homework assignments, challenges, and check-ins with a health coach. Participants also had access to a study-associated Facebook page for peer support. Education arm The digital Education arm focused on pain and SCD education, teaching users about chronic pain, the physiology and neuroscience underpinning SCD pain, healthy lifestyle tips, and facts about SCD including genetic inheritance. It emphasized knowledge acquisition through brief quizzes and discussion with the health coach and the participant’s social network. Health coach Human support may enhance user engagement and optimize the impact of digital interventions. , Thus, both study arms had access to trained lay health coaches, primarily peers with SCD or caregivers, who provided weekly emotional and informational support for 12 weeks to reinforce skills and intervention use. Health coaches underwent an initial 8-hour intensive training session covering health coaching principles, study protocol, communication skills, motivational interviewing, CBT overview, and SCD education. Following this, they received ongoing support through weekly 1-hour supervision meetings and continued training sessions. Adherence to the protocol was maintained through weekly supervision by a master’s-level psychologist or behavioral specialist and the study principal investigator, and monitoring of text message communications. Training time varied but involved at least 6 hours of live video instruction. Statistical analysis Our planned sample size of 350 participants provided 80% power to detect a 0.37 standard deviation difference between study arms in 6-month changes, accounting for 15% attrition. Descriptive statistics were calculated for all variables. The missing data mechanism was characterized by comparing attrition rates between study arms and baseline characteristics of withdrawn and remaining participants, and was determined to be nonignorable. Intention-to-treat analyses were conducted using linear mixed models with time, study arm, their interaction, study site, and baseline depression level as fixed effects, and subject-level random effects. Contrasts assessed the impact on 6-month improvements. Sensitivity analyses of the primary outcome were restricted to participants with confirmed SCD diagnoses. For the depression outcome, sensitivity analyses omitted the ninth PHQ question for all participants. Subgroup analysis examined the heterogeneity of treatment effects on pain interference between high (PHQ-9 >10) and low (PHQ-9 ≤10) baseline depression levels using a 3-way interaction (time × study arm × baseline depression level). Exploratory analyses investigated the impact of intervention engagement on outcomes by restricting to participants who started at least 1 lesson and conducting “intensity-adjusted” analyses based on the proportion of completed chatbot sessions and health coach interactions. Primary and secondary hypotheses were tested using a 5% type I error rate. Analyses were conducted using SAS 9.4 (SAS Institute).
The digital CBT arm taught participants how to recognize negative thoughts and emotions, use cognitive skills and problem-solving, and apply coping behaviors. It emphasized skills acquisition through practice, homework assignments, challenges, and check-ins with a health coach. Participants also had access to a study-associated Facebook page for peer support.
The digital Education arm focused on pain and SCD education, teaching users about chronic pain, the physiology and neuroscience underpinning SCD pain, healthy lifestyle tips, and facts about SCD including genetic inheritance. It emphasized knowledge acquisition through brief quizzes and discussion with the health coach and the participant’s social network.
Human support may enhance user engagement and optimize the impact of digital interventions. , Thus, both study arms had access to trained lay health coaches, primarily peers with SCD or caregivers, who provided weekly emotional and informational support for 12 weeks to reinforce skills and intervention use. Health coaches underwent an initial 8-hour intensive training session covering health coaching principles, study protocol, communication skills, motivational interviewing, CBT overview, and SCD education. Following this, they received ongoing support through weekly 1-hour supervision meetings and continued training sessions. Adherence to the protocol was maintained through weekly supervision by a master’s-level psychologist or behavioral specialist and the study principal investigator, and monitoring of text message communications. Training time varied but involved at least 6 hours of live video instruction.
Our planned sample size of 350 participants provided 80% power to detect a 0.37 standard deviation difference between study arms in 6-month changes, accounting for 15% attrition. Descriptive statistics were calculated for all variables. The missing data mechanism was characterized by comparing attrition rates between study arms and baseline characteristics of withdrawn and remaining participants, and was determined to be nonignorable. Intention-to-treat analyses were conducted using linear mixed models with time, study arm, their interaction, study site, and baseline depression level as fixed effects, and subject-level random effects. Contrasts assessed the impact on 6-month improvements. Sensitivity analyses of the primary outcome were restricted to participants with confirmed SCD diagnoses. For the depression outcome, sensitivity analyses omitted the ninth PHQ question for all participants. Subgroup analysis examined the heterogeneity of treatment effects on pain interference between high (PHQ-9 >10) and low (PHQ-9 ≤10) baseline depression levels using a 3-way interaction (time × study arm × baseline depression level). Exploratory analyses investigated the impact of intervention engagement on outcomes by restricting to participants who started at least 1 lesson and conducting “intensity-adjusted” analyses based on the proportion of completed chatbot sessions and health coach interactions. Primary and secondary hypotheses were tested using a 5% type I error rate. Analyses were conducted using SAS 9.4 (SAS Institute).
Of the 453 participants screened, 359 (79.2%) were randomly assigned to the CBT (n = 181) and Education (n = 178) groups . As shown in , the sample was predominantly clinic-enrolled (n = 265; 73.8%), female (66.3%), and Black/African American (92.5%). The mean age was 36.3 years. Most (74.7%) had at least some college, 36.8% were employed, and 45.7% were receiving disability benefits. Apart from a slightly higher proportion of males in the CBT group, demographics appeared balanced between arms. No group differences in suicidal ideation were observed. At 3 months, 269 (75%) participants completed assessments; at 6 months, 250 (70%) completed assessments. In the CBT arm, 8 participants withdrew or were ineligible, and 41 (23.4%) and 48 (27.7%) missed 3- and 6-month follow-ups, respectively. In the Education arm, 8 withdrew or were ineligible, and 36 (21.1%) and 45 (26.5%) missed 3- and 6-month follow-ups, respectively. CBT and Education groups had similar withdrawal rates (CBT, n = 8 [4.4%] vs Education, n = 8 [4.5%]; P > .99) and 6-month retention rates (CBT, n = 125 [69.1%] vs Education, n = 125 [70.2%]; P = .82) . Primary and secondary outcomes There were no significant differences between CBT and Education arms in the effect of the intervention on pain interference (0.54; −1.30 to 2.37; P = .57) . Restricting the analysis to the 346 participants with confirmed SCD yielded negligible differences (0.50; −1.34 to 2.35; P = .59). Similarly, there were no between-arm differences in the effects on daily pain intensity; PHQ depression, GAD anxiety, and ASCQ-Me social functioning scores (all P > .30); or self-efficacy ( P = .12). For the ASCQ-Me emotional impact scores, the between-group difference approached significance, with CBT being associated with 1.72-point greater improvements (95% confidence interval [CI], −0.03 to 3.46; P = .05). When we restricted our primary analysis to the 346 participants with confirmed SCD diagnosis, the difference in results was negligible (0.50; −1.34 to 2.35; P = .59). Significant improvements in PROMIS Pain Interference scores were found in both CBT (−2.13 [95% CI −3.42 to −0.84]) and Education (−2.66 [−3.97 to −1.36]) arms at 6 months; however, this did not apply to daily pain intensity. Within-arm improvements were observed for PHQ depression (CBT, −1.33 [−2.28 to −0.39]; Education, −1.12 [−2.06 to −0.18]), GAD-7 anxiety (CBT, −0.85 [−1.68 to −0.01]; Education, −1.49 [−2.40 to −0.58]), ASCQ-Me social functioning (CBT, 2.45 [1.13-3.76]; Education, 2.34 [1.00-3.68]), and ASCQ-Me emotional impact scores (CBT, 3.51 [2.29-4.73]; Education, 1.79 [0.55-3.04]). For SCSES self-efficacy scores, only CBT showed significant improvement (1.45 [0.44-2.46]). Heterogeneity of treatment effects Baseline PHQ depression score (≥10 vs <10) did not moderate the effect of treatments on pain interference at 6 months. The effect of digital CBT was similar between participants with low (0.93 [95% CI, −1.42 to −3.28]) and high (0.10 [−2.87 to 3.07]) depression scores ( P = .52). Digital intervention engagement As shown in , a similar proportion of participants in the CBT (24%) and Education (23%) arms never connected to the chatbot. Reasons for not connecting included not wanting to be on Facebook, difficulty with passwords, and lack of reminders. A higher percentage of CBT participants (19%) connected but never started lessons compared with Education participants (9%). Completion rates for Lesson 1 were 45% (CBT) and 51% (Education), but these rates decreased for subsequent lessons, with 18% (CBT) and 32% (Education) completing all lessons. Health coach engagement Of all the randomly assigned participants, 286 (79.7%) completed at least 1 health coach session within 3 months of enrollment. The mean number of health coach sessions was 4.2 out of 12 overall, with 4.1 sessions in the CBT arm and 4.3 sessions in the Education arm. Health coach sessions were mostly conducted by telephone (62%-93%), with a mean session length of 25 minutes. For example, in Session 1, 93% of CBT contacts and 94% of Education contacts were conducted by telephone, with mean durations of 27 and 26 minutes, respectively. Text-based contacts accounted for 4% (CBT) and 3% (Education). The CBT and Education arms had similar numbers and durations of health coach contacts. Exploratory and post hoc analyses Treatment engagement intensity effect In the subset of participants who started at least 1 chat session, the intervention effect on pain interference was more attenuated (−0.05 [95% CI, −2.29 to 2.20]; P = .97) than in the overall cohort. Further, the proportion of lessons completed was not associated with changes in pain interference (−0.13 points per each 10% increase in lessons completed [95% CI, −0.37 to 0.12]; P = .31) , with estimated gains ranging from −1.83 points (0% completed) to −3.09 points (100% completed). No relationship was found between the proportion of completed health coach sessions and reductions in pain interference at 6 months (−0.24 points per each 10% increase in health coach sessions completed [95% CI, −0.59 to 0.10]; P = .17) . Estimated mean improvements ranged from −1.65 points among those completing 0% of health coach sessions to −4.10 points with 100% health coach session completion. After combining the chatbot and health coach engagement measures into an average value, no relationship was found between the proportion of sessions completed and reductions in pain interference at 6 months (−0.23 points per each 10% increase in average sessions/lessons completed [95% CI, −0.57 to 0.10]; P = .17) .
There were no significant differences between CBT and Education arms in the effect of the intervention on pain interference (0.54; −1.30 to 2.37; P = .57) . Restricting the analysis to the 346 participants with confirmed SCD yielded negligible differences (0.50; −1.34 to 2.35; P = .59). Similarly, there were no between-arm differences in the effects on daily pain intensity; PHQ depression, GAD anxiety, and ASCQ-Me social functioning scores (all P > .30); or self-efficacy ( P = .12). For the ASCQ-Me emotional impact scores, the between-group difference approached significance, with CBT being associated with 1.72-point greater improvements (95% confidence interval [CI], −0.03 to 3.46; P = .05). When we restricted our primary analysis to the 346 participants with confirmed SCD diagnosis, the difference in results was negligible (0.50; −1.34 to 2.35; P = .59). Significant improvements in PROMIS Pain Interference scores were found in both CBT (−2.13 [95% CI −3.42 to −0.84]) and Education (−2.66 [−3.97 to −1.36]) arms at 6 months; however, this did not apply to daily pain intensity. Within-arm improvements were observed for PHQ depression (CBT, −1.33 [−2.28 to −0.39]; Education, −1.12 [−2.06 to −0.18]), GAD-7 anxiety (CBT, −0.85 [−1.68 to −0.01]; Education, −1.49 [−2.40 to −0.58]), ASCQ-Me social functioning (CBT, 2.45 [1.13-3.76]; Education, 2.34 [1.00-3.68]), and ASCQ-Me emotional impact scores (CBT, 3.51 [2.29-4.73]; Education, 1.79 [0.55-3.04]). For SCSES self-efficacy scores, only CBT showed significant improvement (1.45 [0.44-2.46]). Heterogeneity of treatment effects Baseline PHQ depression score (≥10 vs <10) did not moderate the effect of treatments on pain interference at 6 months. The effect of digital CBT was similar between participants with low (0.93 [95% CI, −1.42 to −3.28]) and high (0.10 [−2.87 to 3.07]) depression scores ( P = .52).
Baseline PHQ depression score (≥10 vs <10) did not moderate the effect of treatments on pain interference at 6 months. The effect of digital CBT was similar between participants with low (0.93 [95% CI, −1.42 to −3.28]) and high (0.10 [−2.87 to 3.07]) depression scores ( P = .52).
As shown in , a similar proportion of participants in the CBT (24%) and Education (23%) arms never connected to the chatbot. Reasons for not connecting included not wanting to be on Facebook, difficulty with passwords, and lack of reminders. A higher percentage of CBT participants (19%) connected but never started lessons compared with Education participants (9%). Completion rates for Lesson 1 were 45% (CBT) and 51% (Education), but these rates decreased for subsequent lessons, with 18% (CBT) and 32% (Education) completing all lessons. Health coach engagement Of all the randomly assigned participants, 286 (79.7%) completed at least 1 health coach session within 3 months of enrollment. The mean number of health coach sessions was 4.2 out of 12 overall, with 4.1 sessions in the CBT arm and 4.3 sessions in the Education arm. Health coach sessions were mostly conducted by telephone (62%-93%), with a mean session length of 25 minutes. For example, in Session 1, 93% of CBT contacts and 94% of Education contacts were conducted by telephone, with mean durations of 27 and 26 minutes, respectively. Text-based contacts accounted for 4% (CBT) and 3% (Education). The CBT and Education arms had similar numbers and durations of health coach contacts.
Of all the randomly assigned participants, 286 (79.7%) completed at least 1 health coach session within 3 months of enrollment. The mean number of health coach sessions was 4.2 out of 12 overall, with 4.1 sessions in the CBT arm and 4.3 sessions in the Education arm. Health coach sessions were mostly conducted by telephone (62%-93%), with a mean session length of 25 minutes. For example, in Session 1, 93% of CBT contacts and 94% of Education contacts were conducted by telephone, with mean durations of 27 and 26 minutes, respectively. Text-based contacts accounted for 4% (CBT) and 3% (Education). The CBT and Education arms had similar numbers and durations of health coach contacts.
Treatment engagement intensity effect In the subset of participants who started at least 1 chat session, the intervention effect on pain interference was more attenuated (−0.05 [95% CI, −2.29 to 2.20]; P = .97) than in the overall cohort. Further, the proportion of lessons completed was not associated with changes in pain interference (−0.13 points per each 10% increase in lessons completed [95% CI, −0.37 to 0.12]; P = .31) , with estimated gains ranging from −1.83 points (0% completed) to −3.09 points (100% completed). No relationship was found between the proportion of completed health coach sessions and reductions in pain interference at 6 months (−0.24 points per each 10% increase in health coach sessions completed [95% CI, −0.59 to 0.10]; P = .17) . Estimated mean improvements ranged from −1.65 points among those completing 0% of health coach sessions to −4.10 points with 100% health coach session completion. After combining the chatbot and health coach engagement measures into an average value, no relationship was found between the proportion of sessions completed and reductions in pain interference at 6 months (−0.23 points per each 10% increase in average sessions/lessons completed [95% CI, −0.57 to 0.10]; P = .17) .
In the subset of participants who started at least 1 chat session, the intervention effect on pain interference was more attenuated (−0.05 [95% CI, −2.29 to 2.20]; P = .97) than in the overall cohort. Further, the proportion of lessons completed was not associated with changes in pain interference (−0.13 points per each 10% increase in lessons completed [95% CI, −0.37 to 0.12]; P = .31) , with estimated gains ranging from −1.83 points (0% completed) to −3.09 points (100% completed). No relationship was found between the proportion of completed health coach sessions and reductions in pain interference at 6 months (−0.24 points per each 10% increase in health coach sessions completed [95% CI, −0.59 to 0.10]; P = .17) . Estimated mean improvements ranged from −1.65 points among those completing 0% of health coach sessions to −4.10 points with 100% health coach session completion. After combining the chatbot and health coach engagement measures into an average value, no relationship was found between the proportion of sessions completed and reductions in pain interference at 6 months (−0.23 points per each 10% increase in average sessions/lessons completed [95% CI, −0.57 to 0.10]; P = .17) .
The CaRISMA study represents a significant milestone in SCD research, which, to our knowledge, is the largest behavioral intervention trial to date, with a strong emphasis on community engagement. Conducted during the COVID-19 pandemic, this comparative effectiveness study of digital CBT and education met its enrollment goal of 350 participants with a relatively low dropout rate, underscoring the potential of community-engaged approaches in underserved populations. Contrary to our hypothesis, this adequately powered study found that digital education with health coach support was as effective as digital CBT with the same support. Until now, there have been few large-scale trials of nonpharmacological SCD interventions. Although some studies suggest digital CBT’s effectiveness for pain and mental health in SCD, , , they are limited by small samples, heterogeneous interventions, lack of control groups, short follow-up periods, and reliance on self-report measures. , Only one-third of nonpharmacological SCD interventions have shown effectiveness, with inconclusive evidence for self-management interventions. However, the existing data on CBT specifically for SCD pain are encouraging. Schatz et al found that increased practice of digital CBT coping skills was associated with next-day pain reductions among pediatric participants. Sil et al found some CBT-related improvements in health care utilization among children and adolescents with SCD. Palermo et al demonstrated medium-sized effects on pain in adolescents with SCD. Although CBT is effective for pain compared with control or usual care, the current study is, to our knowledge, the first to compare digital CBT with another evidence-based nonpharmacological intervention such as health coach–supported psychoeducation in adults. The CaRISMA study addresses a key limitation of prior research by integrating the community at all stages, including intervention delivery. Previous SCD intervention research often lacked real community partnership in design, delivery, and trial conduct. Few studies involved people with SCD in developing self-management interventions, primarily in pediatrics, , with limited participation in needs assessment or treatment delivery. To our knowledge, this is the first study in adults with SCD to involve the target population in intervention cocreation, which is crucial for engagement. Our findings suggest that community engagement facilitates enrollment, and peer support from community members is a particularly attractive aspect of the intervention. This underscores the need for more community-engaged work in SCD research and intervention development. However, the study revealed important challenges in digital engagement. Despite successful enrollment, participants’ interaction with the digital components was lower than anticipated, which may explain the modest effect sizes and lack of significant differences between the CBT and Education arms. Trials typically find that less than half (44.2%) of participants complete all treatment modules, with poorer engagement among minority populations. Despite implementing codesign, tailoring, health coach support, and community involvement, engagement remained challenging. Factors potentially contributing to lack of intervention uptake include Facebook platform restrictions preventing push notifications; lack of trust in Facebook and shifts in social media popularity ; complex password and identification code verification process; inconsistent participant payments due to logistical challenges; and perceived redundancy of intervention information for some participants. Facebook was initially recommended by patients and family members during the study design phase, but subsequent data breaches and privacy concerns may have soured perceptions. Some participants refused to engage with Facebook, even when offered dummy accounts. Despite the barriers to engagement, the CaRISMA trial had 2 promising findings that should drive future research. First, this study highlighted the potential of peer health coaches in enhancing digital therapeutics’ effectiveness. Although digital engagement was low, participants showed higher interaction with peer health coaches, suggesting that this human element may be crucial in bridging the gap between digital interventions and patient needs. Second, CBT showed promise in improving self-efficacy, whereas the Education arm did not exhibit any change in self-efficacy. This suggests that CBT may be affecting the intended treatment mechanism, , potentially contributing to positive outcomes if engagement is optimized. Future studies should evaluate whether CBT is more effective than education alone for patients with low self-efficacy. The trial did not find differences between the study arms on the daily 0 to 10 numeric pain rating scale at 6 months. The discrepancy may be due to the limitations of this unidimensional measure, which only assesses pain intensity and fails to account for the affective and cognitive aspects of pain and the multidimensional nature of the pain experience. , Furthermore, the scale’s simplicity may lead to a lack of sensitivity in detecting subtle changes in pain perception over time. Limitations This study has some limitations. First, technical challenges may have contributed to poor intervention engagement and participants’ ability to use the study apps as intended. Second, variability in health coaching quality and adherence to the protocol was observed due to unexpected demand and the number of coaches involved. Health coach interventions that rely on volunteers may be less effective. Third the attrition rate was greater than the proposed 15% for which this study was powered. Finally, the inclusion of a standard-of-care control group would have helped explain improvements in symptoms. Conclusion In this study, which, to our knowledge, is the largest behavioral trial in SCD and one of the largest digital health trials in a minority population, we demonstrated the feasibility of a community-engaged, decentralized approach. Both health coach–supported digital interventions effectively improved pain interference, mental health, and quality of life in adults with SCD. Our findings suggest that digital CBT, or pain education, when combined with centralized health coach support, may serve as a scalable and low-cost method for delivering high-quality behavioral care to underserved communities, addressing the challenges of chronic pain and mental health in individuals with SCD and other marginalized populations.
This study has some limitations. First, technical challenges may have contributed to poor intervention engagement and participants’ ability to use the study apps as intended. Second, variability in health coaching quality and adherence to the protocol was observed due to unexpected demand and the number of coaches involved. Health coach interventions that rely on volunteers may be less effective. Third the attrition rate was greater than the proposed 15% for which this study was powered. Finally, the inclusion of a standard-of-care control group would have helped explain improvements in symptoms.
In this study, which, to our knowledge, is the largest behavioral trial in SCD and one of the largest digital health trials in a minority population, we demonstrated the feasibility of a community-engaged, decentralized approach. Both health coach–supported digital interventions effectively improved pain interference, mental health, and quality of life in adults with SCD. Our findings suggest that digital CBT, or pain education, when combined with centralized health coach support, may serve as a scalable and low-cost method for delivering high-quality behavioral care to underserved communities, addressing the challenges of chronic pain and mental health in individuals with SCD and other marginalized populations.
|
First genome-wide association study on rocuronium dose requirements shows association with | 832a42a2-59ae-4705-8a86-f7d863cf617f | 8132880 | Pharmacology[mh] | BrePainGen is a prospective study designed to examine the role of genetics in acute and persistent post-surgical pain, mood, and effects of drugs used in anaesthesia. One thousand women undergoing surgery for breast cancer at the Helsinki University Hospital were recruited between August 1, 2006 and December 31, 2010. The study protocol was approved by the coordinating ethics committee (136/E0/2006) and the ethics committee of the Department of Surgery (Dnro 148/E6/05) of the Hospital District of Helsinki and Uusimaa. Written informed consent was obtained from all patients. A flow chart of patient recruitment is provided in . After informed consent, the patients were interviewed for background information about medical conditions, medications, age, height, weight, previous operations, use of alcohol, and smoking. Those using drugs affecting the pharmacology of rocuronium were excluded. All patients were premedicated with diazepam 2.5–15 mg and paracetamol 1 g orally. Anaesthesia was induced with propofol 2–3 mg kg −1 , and remifentanil infusion of 0.2 mg kg −1 min −1 was started. Tracheal intubation was facilitated with rocuronium 0.6 mg kg −1 . During surgery, anaesthesia was maintained with a propofol infusion at 50–100 μg kg −1 min −1 to keep state entropy (M-Entropy S/5TM Module for Anesthesia Monitor; GE Healthcare Finland, Helsinki, Finland) at the level of 50 [5]. Remifentanil infusion was used at 0.05–0.25 μg kg −1 min −1 to keep systolic BP at [15%] of baseline minus 20 mm Hg. The neuromuscular block was maintained throughout the surgery with rocuronium boluses of 10 mg to keep the train-of-four ratio at 0–10% (E-NMT; GE Healthcare Finland). Mechanical ventilation was adjusted to normocapnia with 1:1 oxygen and nitrous oxide. During closure of the skin, remifentanil infusion was stopped, and boluses of fentanyl 1 μg kg −1 , ondansetron 4 mg, and droperidol 0.01 mg kg −1 were given intravenously. Neuromuscular block was reversed with neostigmine 2.5 mg and glycopyrrolate 0.5 mg. Before the patient woke from anaesthesia, a blood specimen was drawn for DNA isolation. DNA was extracted from peripheral blood using the Autopure LS™ automated DNA purification instrument (Gentra Systems, Inc., Minneapolis, MN, USA). Genotype data were produced at the Wellcome Sanger Institute (Hinxton, UK) on the HumanOmniExpress Illumina BeadChip (Illumina, Inc., San Diego, CA, USA) while blind to phenotypic information. Sample quality control procedures have been described in detail earlier. All single-nucleotide polymorphisms (SNPs) were filtered based on minor allele frequency (MAF >0.0005), Hardy–Weinberg equilibrium ( P >1 × 10 −6 ), and success rate (>0.97). The mean genotyping success rate was 0.997. After quality control, genotyping data were available for 926 of the 1000 participants. Eight patients were excluded for clinical reasons. The final participant population comprised 918 individuals with both genotype and clinical data available . Statistical analyses and data management were conducted using IBM SPSS software versions 23.0 and 24.0 and R version 3.6.2 (R Foundation for Statistical Computing, Vienna, Austria). To identify possible confounders that should be taken into account when performing the GWAS, we first performed univariate testing ( n =992) between clinical variables and the dose of rocuronium needed to maintain adequate neuromuscular block. We took the dose rates of rocuronium (in units of mg kg −1 min −1 ) using the natural logarithms of their numerical values to ensure normal distribution. The clinical variables to be tested for relevance to rocuronium dose requirements were age, height, BMI, total use of propofol during anaesthesia (mg kg −1 min −1 ), total use of remifentanil during anaesthesia (mg kg −1 min −1 ), ASA class, smoking (yes/no), alcohol use (yes/abstinent), type of axillary surgery (sentinel node biopsy/evacuation), and breast surgery (resection/mastectomy). We also tested associations between CYP2D6 copy numbers, CYP2D6 -predicted phenotype (poor, intermediate, extensive, or ultra-rapid metabolisers), CYP3A4 variant rs35599367 genotype (CC/CT/TT) and CYP3A5 variant rs776746 genotype (GG/AG/AA), and the natural logarithm of rocuronium dose rates (expressed in mg kg −1 min −1 ), as the CYP data were available from these patients. We tested these associations using t -test, Mann–Whitney U -test, analysis of variance, or Kruskal–Wallis test, depending on the distributions of variables. Pairwise comparisons were adjusted with Bonferroni corrections. After studying possible associations, we conducted multivariate linear regression modelling based on variables that had statistically significant ( P <0.05) associations with rocuronium requirements. We used a stepwise method to construct the final model. The final linear regression model contained only variables that remained significant in multivariate testing. These were used as covariates in the GWAS. In addition, the first five dimensions from multidimensional scaling of genotype data were also used as covariates to take into account a possible hidden population structure. The GWAS was conducted using an additive linear regression model with PLINK software. Associations between total dose of rocuronium and 653 034 genetic variants (SNPs) were tested. The standard threshold of genome-wide statistical significance, P <5 × 10 −8 , was used. After GWAS results became available, we performed another linear regression round, including four lead variants, to examine the impact of these on the total dose of rocuronium. The genomic region showing a significant association with rocuronium dosage was further examined to identify the most likely causal SNPs within the locus. For this, the genomic data were first pre-phased with Eagle software version 2.4. Subsequently, the genotypes were imputed using Beagle 4.1 and population-specific Sequencing Initiative Suomi panel as imputation reference. , Poorly imputed variants were excluded (INFO <0.7). The imputation reference panel consisted of 3775 Finns. To identify the number of independent association signals and the lead SNPs within the associated locus, the FINEMAP tool was used. The FUMA tool was used to examine the potential functional effect of each associated SNP. Genetic effects on gene expression across tissue types were studied using publicly available data from the Genotype-Tissue Expression project v8. FINEMAP 1.4 was run, allowing a maximum of K =5 causal variants. The credible sets, those containing the most likely causal variants, were reported, assuming either one or two causal variants.
To test associations with clinical variables, all 992 individuals with complete clinical data were used. For 918 of these, genome-wide genotype data were available and used for the GWAS. The characteristics of the subjects and of rocuronium requirements are shown in , . The total dose of rocuronium had a linear relationship ( R =0.412) with the duration of anaesthesia ( a). The median number of additional doses was 4 (inter-quartile range [IQR]: 2–5), range 0–20. Twenty-one patients did not receive additional doses of rocuronium after intubation. Of the continuous variables tested, age ( P <0.001), BMI ( P <0.001), and total doses of propofol (mg kg −1 min −1 ; P <0.001) and of remifentanil (mg kg −1 min −1 ; P <0.001) had significant correlations with the dose of rocuronium (in mg kg −1 min −1 ). Of the dichotomous and ordinal variables, ASA class ( P <0.001), use of alcohol compared with abstinence ( P =0.001), and breast resection vs mastectomy ( P <0.001) were statistically significant at P <0.05. Multiple linear regression models were run to understand the effects of the aforementioned variables on the dose of rocuronium (in mg kg −1 min −1 ). There was homoscedasticity as assessed by visual inspection of a plot of standardised residuals vs standardised predicted values, normality of the residuals being assessed by visual inspection of a normal probability plot: there were no significant outliers, as assessed by Cook's distance. The results of the linear model are presented in . The clinical variables in the model accounted for 35.3% of the variability in rocuronium dosage. The GWAS highlighted one locus on chromosome 12 showing genome-wide significant evidence of association with rocuronium dose . Eight genotyped SNPs reached the standard threshold of genome-wide statistical significance, P <5 × 10 −8 . The SNPs with the most significant evidence of association were all located in or near gene SLCO1A2 . LocusZoom plots of the area are shown in . Based on the genetic recombination patterns and the FINEMAP tool, the most likely scenario is that two SNPs are needed to explain the association. The top candidates are rs7967354 (P=5.3e −11 ; β=–0.143 for allele G; MAF=0.22) and rs11045995 ( P =1.4e −10 ; β=–0.147 for allele G; MAF=0.18), both located in gene SLCO1A2 and in moderate linkage disequilibrium (LD) with each other ( r 2 =0.26). The minor alleles of both variants (rs7967354-G and rs11045995-G) are associated with decreased rocuronium requirements. Patients with two minor alleles (G/G) of the variant rs7967354 needed significantly less rocuronium during anaesthesia ( n =47; median dose rate: 6.1 μg kg −1 min −1 ; IQR: 5.5–7.5) compared with patients with G/A genotype ( n =310; median dose rate: 7.7 μg kg −1 min −1 ; IQR: 6.4–9.5) and A/A genotype ( n =561; median dose rate: 8.2 μg kg −1 min −1 ; IQR: 6.3–9.1) ( b) in Kruskal–Wallis testing ( P <0.001). Similarly, patients with rs11045995 G/G genotype needed significantly lower doses of rocuronium during anaesthesia ( n =37; median dose rate: 6.3 μg kg −1 min −1 ; IQR: 5.6–7.0; P <0.001) than the patients with G/A ( n =254; median dose rate: 7.5 μg kg −1 min −1 ; IQR: 6.3–9.1) and A/A genotypes ( n =627; median dose rate: 8.3 μg kg −1 min −1 ; IQR: 6.8–10.1) ( c). We also found that patients with two minor alleles in both dose-altering variants (rs7967354-GG + rs11045995-GG; n =19; median dose rate: 5.9 [IQR: 5.4–6.5] μg kg −1 min −1 ) needed even less rocuronium during anaesthesia than with other genotype combinations in the Kruskal–Wallis test ( a and d; ). The final linear regression model included both the clinical variables and the two lead SNPs rs7967354 and rs11045995 . The variables in the model accounted for 41% of the variability (adjusted R 2 ) in the rocuronium dose. In this multivariate analysis, age ( P <0.001), BMI ( P <0.001), ASA 1 vs ASA 3 ( P =0.008), ASA 2 vs ASA 3 ( P =0.001), dose of propofol ( P <0.001), rs7967354 ( P =0.001), and rs11045995 ( P <0.001) remained significant at P <0.05. By including the two SNPs in the model, the proportion of the variance explained increased by 4 percentage points. As both rs7967354 and rs11045995 are located in the intronic regions of SLCO1A2 , we analysed the region further using imputed genome data, and FINEMAP and FUMA programs, to locate other variants possibly driving the association signal. On the imputed data, FINEMAP gives a probability of 74% to one causal variant and 26% for two causal variants. Assuming one causal variant, the 95% credible set (the set of variants containing the causal variant with 95% probability) contains 20 variants . Assuming two causal variants, the lead variants are rs7967354 and rs10743413 (the latter of which is highly correlated with rs11045995; r 2 =0.88), and the two credible sets contain, respectively, 13 and 5 variants with probability over 1% of being causal . FINEMAP estimates that the SLCO1A2 region explains 3.75% of the variance of the phenotype (95% credible interval: 1.77–6.29%). Next, we performed FUMA analysis to check whether the most likely causal variants were associated with tissue-specific changes in the expression level of SLCO1A2 or other genes. For the 20 potentially causal variants, FUMA analysis detected 39 expression quantitative trait loci (eQTL) for four genes at a false discovery rate of 5% . For SLCO1A2 , three variants were eQTL in the brain and three in the cerebellum. Other eQTL were found for the gene RECQL (expression measured in blood), PYROXD1 (in blood), and C12orf39 (in lymphocytes and in blood). Plots for tissue-wide expression results for rs7967354 and rs11045995 are in .
We explored clinical and genetic factors explaining variation in rocuronium requirement during surgery for breast cancer in 918 women. We showed that a locus containing gene SLCO1A2 affects the dose rate needed for maintaining adequate neuromuscular block. Of the clinical variables examined, age, BMI, total dose of propofol, and ASA class were associated with the rocuronium dose. Combined, these factors explained 41% of the dose rate variation. Our study provides further confirmation that neither CYP2D6 nor CYP3A4 plays a role in determining rocuronium requirements. The median rocuronium dose used is in line with previous reports. , In our study, advanced age, lower BMI, and higher ASA class decreased the amount of rocuronium needed. Previous reports on the effect of BMI on rocuronium requirements are conflicting. , Our patients were medicated according to their actual body weight, which might explain why lower BMI decreased rocuronium requirements. We observed that higher propofol doses were associated with increased need for rocuronium, whereas some previous studies suggested that propofol would have muscle-relaxing effects, reducing the required dose of neuromuscular blockers. However, the designs of these studies were very different from ours. There are no previous linear regression models evaluating rocuronium needs during propofol anaesthesia. Our GWAS identified one genome-wide significant association peak, on chromosome 12, in and around the SLCO1A2 gene, which encodes the OATP1A2. The signal was best explained by two lead SNPs, rs7967354 and rs11045995. Higher numbers of the minor alleles of these SNPs were associated with a lesser need for rocuronium. The biggest variation in rocuronium dosage was observed when we compared participants homozygous for both rs7967354 and rs11045995 minor alleles (G) with those homozygous for the major alleles (A) ( a and d; ). Organic anion transporters are cellular transmembrane proteins, important in the distribution, metabolism, and excretion of various drugs and expressed in pharmacokinetically important organs, such as liver, kidney, and intestine. Based on immunohistochemical staining, OATP1A2 transporters are located in cholangiocytes, where they have an important role in excretion of drugs into the bile. As rocuronium is a known substrate of OATP1A2 and is mainly excreted unchanged in the urine (10–25%) and bile (>70%), , the role of OATPs in its excretion is of interest. A study with Slco1a/1b–/– knockout mice showed that lack of functioning OATP1A2 leads to accumulation of the substrates of this transporter in plasma. Previous studies also indicate that OATPs are a target for drug interactions. Expression of OATPs, especially OATP-A, were significantly increased in patients treated with carbamazepine. Carbamazepine use is known to increase the required rocuronium dosage. Our results suggest that this would be attributable to induction of OATP1A2 rather than of CYP3A4. Few previous studies have addressed the pharmacogenetics of rocuronium. A candidate gene study based on only 30 patients, by Costa and colleagues, showed evidence of association between a variant –189_188InsA (rs3834939), located in the promoter region of the SLCO1A2 gene, and reduced clearance of rocuronium. Neither this SNP rs3834939 ( P =0.000026 in our study) nor the other variants tested by Costa and colleagues were amongst the SNPs showing genome-wide significant evidence of association (a standard threshold of genome-wide statistical significance is P <5 × 10 −8 ). The other two earlier rocuronium studies were also candidate gene studies, analysing only a few variants in small patient samples. Based on a sample of 105 Chinese patients, Qi and colleagues reported that SNPs rs12720464 and rs1055302 in the ABCB1 gene, coding for an ATP-dependent drug efflux pump, associate with prolonged spontaneous recovery after a single dose of rocuronium. In the study of Mei and colleagues, another ABCB1 SNP (rs1128503) and an SNP (rs2306283) in the OATP1B1 transporter gene SLCO1B1 showed association with the clinical action time of rocuronium in 200 patients. Our data do not provide support for the ABCB1 findings, whilst several variants within the SLCO1B1 gene, located right next to SLCO1A2 , show almost genome-wide significant evidence of association in our study . Further analyses showed that these SNPs were not independent from our SLCO1A2 lead variants (data not shown). The top SNPs in our study, rs7967354 and rs11045995, and the variants in high LD with them, are located in non-coding parts of the gene. The intronic rs7967354 is in high LD with SNPs rs4149005 (non-coding exon transcript variant) and rs875234 (3′ UTR variant). Our extensive eQTL analyses suggest that a possible mechanism for the genotype–rocuronium dose association is tissue-specific gene expression regulation. As OATP1A2 has a role in excretion of rocuronium into bile, changes in the expression of SLCO1A2 caused by polymorphisms in the gene might prolong the effect of rocuronium by reducing clearance. One unexplored option to explain variation in rocuronium dose requirement is that of inter-individual differences in neuromuscular junctions. Here, we can only speculate on the possibility of some underlying variation in neuromuscular junction in otherwise healthy patients that could explain the variation in dose needs. Interestingly, SLCO1A2 is highly expressed in neural tissues, including peripheral nerve tissue, and our eQTL analyses suggested that the lead variants affect the SLCO1A2 expression level in the brain. It is tempting to speculate that neural tissue also plays a role in the impact of SLCO1A2 variants on rocuronium requirements. Our study has some limitations. Creatinine or creatinine clearance values were not available to assess kidney function. However, patients with clinically relevant kidney failure were excluded from the study. Although the study cohort is the largest thus far examined for pharmacogenetic data suitable for rocuronium studies, it is small for the GWAS approach, which usually requires thousands of participants. As our results are based on only 918 participants, all female, it is likely that some of the more subtle genetic effects remain undetected. Our study suggests that genetic variation in the gene SLCO1A2 , encoding OATP1A2, is significantly associated with differences in rocuronium requirements. Our discovery offers one explanation for inter-individual differences in the duration of action of rocuronium. This variation was estimated to account for 4% of the variability in rocuronium dose. The most likely underlying mechanism is altered uptake of rocuronium by OATP1A2.
Study design: SA, RJ, KTO, MAK, EK Genetic data design: MAK, MP Recruitment and perioperative management of patients: RJ Whole BrePainGen study: EK, RJ, MAK Data analysis: SA, PB Genetic analysis: SA, LO, AA-O, MP, MAK Interpretation of results: SA, PB, RJ, KTO, MAK, EK Writing of paper: SA, PB, RJ, LO, AA-O, KTO, MAK, EK
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.